“I feel physically safe but not politically safe”:

Understanding the Digital Threats and Safety Practices of OnlyFans Creators

Ananta Soneji Vaughn Hamilton Adam Doupé

Arizona State University Max Planck Institute for Software Systems Arizona State University

Allison McDonald Elissa M. Redmiles

Boston University Georgetown University

Abstract creators must manage risks such as censorship of sexual ex-

pression, platform precarity [30], refusal of service from bank-
OnlyFans is a subscription-based social media platform with
ing institutions [77], and digital account loss [15]. Despite
over 1.5 million content creators and 150 million users world-
these risks, OnlyFans provides a level of flexibility, accessi-
wide. OnlyFans creators primarily produce intimate content
bility, and control that draws creators to the platform [47].
for sale on the platform. As such, they are distinctly positioned
Drawing on the contextual risk factors outlined by War-
as content creators and sex workers. Through a qualitative
ford et al. [99], OnlyFans creators are at-risk—“anyone who
interview study with OnlyFans creators (n=43), building on
experiences heightened digital-safety threats”—due to their
an existing framework of online hate and harassment, we shed
prominence as well as the stigma associated with sex work.
light on the nuanced threats they face and their safety prac-
With this work, we respond to the call of Thomas et al. [90]—
tices. Additionally, we examine the impact of factors such
“responding to each class of hate and harassment requires
as stigma, prominence, and platform policies on shaping the
a unique strategy”—by identifying the online risks to and
threat landscape for OnlyFans creators and detail the preemp-
protective measures taken by sexual content creators.
tive practices they undertake to protect themselves. Leverag-
Through semi-structured interviews with 43 OnlyFans cre-
ing these results, we synthesize opportunities to address the
ators,1 we investigate their online risks and the protective
challenges of sexual content creators.
strategies they employ to mitigate them. In particular, we
recruited both participants who were new to creating sexual
1 Introduction content and those with previous sex industry experience, cap-
turing a broad spectrum of experience and stages of learning
The exponential growth of OnlyFans since its founding in about digital safety. We investigate three research questions:
2016 has transformed it into a prominent social media plat- RQ1: How well does the existing online hate and harass-
form with over 1.5 million creators, who primarily create ment framework [90] capture the threats experienced or per-
erotic content, and 150 million “fans” [71]. OnlyFans creators, ceived by sexual content creators?
like other professional content creators, navigate a compli- RQ2: How do contextual risk factors [99]—stigma and
cated landscape of content creation, monetization strategies, prominence—relate to the threats faced by sexual content
and fan management, and need to engage in extensive off- creators?
platform advertisement to draw followers. The prominence
RQ3: What is the role of platforms in sexual content cre-
that comes with being a content creator, regardless of the
ators’ digital safety?
platform, brings specific security and privacy risks such as
obsessive fans, online harassment, and stalking [30, 91]. We analyze our interviews with OnlyFans creators through
OnlyFans creators face additional sources of risk stemming the lens of Thomas et al.’s online hate and harassment frame-
from sex work stigma [30]. Those who create and sell sexual work [90]. We do so for two primary reasons, to: (i) support a
content are sex workers; they exchange erotic services for growing effort to use unifying language and frameworks [99]
money [69]. However making, distributing, or even possess- across research of hate, harassment, and safety issues and (ii)
ing certain kinds of sexual content is legislated in a variety validate the framework in the context of the combination of
of ways across different countries [84]. Although millions contextual risk factors we study—prominence and stigma.
of people worldwide work in the sex industry, sex workers This approach further allows us to juxtapose our participants’
globally face significant stigma, discrimination, violence, and 1 Hereafter, when we use the term ‘OnlyFans creators,’ we refer to creators

harassment [23, 67, 82]. Therefore, OnlyFans’ sexual content producing sexual content on the platform.
experiences with those of previously studied populations: gen- 2 OnlyFans Background
eral internet users [100], at-risk users [99] and professional
non-sexual content creators [81, 91]. We begin with an overview of the platform, creator interac-
tions with the platform and fans, and noteworthy platform
We find that OnlyFans creators contend with two exist-
policy changes impacting creator experiences.
ing threats from Thomas et al.’s online hate and harassment
framework—toxic content and content leakage—and two OnlyFans and creators. OnlyFans is a fast growing (70%
sources of risk not included in the Thomas et al. framework: per month [38]) social media platform with two types of user:
platform precarity and censorship (e.g., shadowbanning2 and creator and fan. All users must confirm that they are over
deplatforming3 ). Whereas prior work finds that non-sexual 18, with additional identity verification checks for creators5 —
content creators find toxic content to be the most promi- name, date of birth, address, government photo ID, selfie,
nent threat [91], our participants were more concerned about social media handles, payment details—and consent checks
platform-related risks than they were about toxic content and when content features more than one creator.
content leakage. In response, we find participants diversified Creators can host both free and paywalled pages. Users
the platforms they used and practiced self-censorship to avoid must have an OnlyFans account to view free pages. Cre-
being deplatformed. ators can also generate income through pay-per-view con-
tent, custom content requested by individual fans, paid direct
The OnlyFans creators we study face many of the same
messages—where creators can set a per-message fee or flat
toxic content attacks as non-sexual content creators [91].
fee to set up messages for their fans [14]—and fan tips. Cre-
While they use some of the same strategies to address
ators earn 80% of all payments. OnlyFans creators manage
such toxic content—“ignoring haters” and engaging in self-
their payouts and are responsible for chargebacks from fans,
censorship [91]—our participants also establish strong bound-
a model termed “digital patronage” [18]. OnlyFans limits
aries with their fans, extensively use platforms’ blocking fea-
the type of content creators can share and sets guidelines for
tures to counter harassment, and draw strength and support
language in private messages [72].
from the broader community of sexual content creators.
Notably, the platform lacks internal discoverability, neces-
Given the intimate nature of content produced as part of sitating creators to drive traffic internally (when creators
sexual content creation, our analysis emphasizes the preva- promote each other) or externally (from social media plat-
lent and seemingly unavoidable threat of intimate content forms or personal websites) [6], fostering “multi-platform
leakage—constituting both content theft and non-consensual practices” [44]. This boosts the platform’s public visibil-
distribution of intimate content where the creator did not in- ity, and OnlyFans’ brand recognition has further grown by
tend to share it [68]—a consequence our participants often “celebrity hype” [47] and extensive mainstream news media
associated with the online sharing of explicit content. coverage [78], particularly during the COVID-19 pandemic.
While prior work on non-sexual content creators finds a Additionally, Hamilton et al. argue that OnlyFans’ unique
reactive approach to digital safety [81], we find that sexual design has not only spurred its growth but has also attracted a
content creators are uniquely proactive and prepared to defend diverse range of sexual content creators, including those with
against digital threats. All of our participants adopted at least existing sex industry experience and newcomers [47].
one protective practice before joining OnlyFans, compared to Policy changes and known creator risks. Due to its associa-
only one-third of participants in prior work focused on non- tion with erotic content [42], OnlyFans creators are exposed
sexual content creators. This preparedness is vital due to the to the prevailing stigma attached to sex work [53] and the chal-
essential requirement for cross-platform promotion to earn lenges stemming from platform policies. Notably, Bhalerao
money on OnlyFans. Without internal promotion mechanisms et al. point out that platform policies, created without subject
on OnlyFans, creators must drive traffic to their page from matter knowledge or sufficient understanding of its effects,
other platforms. This increases creator visibility and subjects can inadvertently harm sex industry workers [13], as demon-
them to heightened risks such as outing and doxxing4 [81]. strated by a proposed sexual content ban by OnlyFans in Au-
Participants point to the role of sex work stigma in influencing gust 2021. The platform attributed the ban to banking-related
their proactive stance toward digital safety. issues [42] and despite being reversed, resulted in creators suf-
fering financial and subscriber losses [8,13]. The utilization of
online platforms by sex workers for both work and daily activ-
2 Shadowbanning refers to platforms silently suppressing a user and/or ities is significantly affected by policies that are stigmatizing
their content, reducing their visibility. Users may be unaware they have been and overly restrictive [13]. These policies may lead to de-
shadowbanned until they experience unusually low engagement.
3 Deplatforming refers to the removal of a creator’s account from a plat- platforming and reduced platform accessibility, highlighting
form, often without transparency or recourse [16]. the precarious nature of sex work platforms [5, 13, 17]. Such
4 Doxxing refers to revealing someone’s real name, home address, work- circumstances can have profound financial and emotional im-
place, phone number, or other personal information for harassment and fi-
nancial harm. 5 https://onlyfans.com/transparency-center/verification
pacts on marginalized content creators [4]. Furthermore, the safety, a more appealing labor environment, better wages, ef-
structural design of OnlyFans necessitates off-platform adver- ficient advertising, and client vetting methods [27, 53].
tising, subjecting creators to the policy changes and associated Digitally-mediated sex work includes both digital-
risks of other platforms. mediation of in-person work (e.g., digital communication with
Additionally, since OnlyFans uses third-party services for clients for in-person sex work) and online-only sex work like
user verification and monitoring for minor content on the OnlyFans or web-camming “in which cam models sell inter-
platform, breaches of OnlyFans or its third-party services are active computer-mediated sex online” [54]. Jones and Camp-
sources of risk for creators [85]. In a case study, Ford high- bell, respectively, discuss challenges in digitally-mediated
lights the absence of preventive measures against screenshots in-person sex work, namely lack of control on digital iden-
or screen recording on OnlyFans, creating a risk of unau- tities [53], under- or non-payment for services, online ha-
thorized distribution of creators’ material, constituting both rassment including persistent unwanted contact, verbal abuse,
service theft and image-based sexual abuse [34]. surveillance, and hesitance in seeking law enforcement help
against harms given the non-existent legalization of crimes
on digital platforms [23]. Sanders et al. discover that U.K.-
3 Related Work based sex workers mitigate such digital risks by using aliases,
filtering clients, and utilizing social media to connect with
We discuss prior literature on online risks and safety practices
friends and ensure their safety [82]. In a recent study, McDon-
of content creators and sex workers.
ald et al. highlighted that in-person sex workers faced risks
Content creators. Across social media platforms, content coming from clients, deficient legal protections, and hostile
creators produce content for financial benefits [56, 93], self- digital platforms, which participants mitigate by covering for
expression [19, 65, 70, 73], socialization [65, 73, 93], and pop- a friend or colleague, vetting clients, and managing digital
ularity (e.g., to create “media brands”) [26]. However, prior identities [67]. Strohmayer et al. studied how sex workers use
work explored the multifaceted challenges that content cre- digital technologies to report harassment such as boundary
ators face. For instance, Uttarapong et al. investigated the pushing and violence, and use such tools to proactively avoid
harassment experiences of women and LGBTQ streamers violent or disrespectful clients [87, 88].
on Twitch, revealing difficulties in coping real-time with In analyzing the largest UK-based adult entertainment plat-
negativity from fans [96]. While Jhaver et al. describe how form, Rand [75] finds that although digital platforms enable
YouTube creators benefit from using collaborative keyword flexibility and choice, sex workers become dependent on those
filtering to manage hate and harassment [52], community platforms. As a consequence, workers may pay higher inter-
self-moderation requires significant emotional and relational mediary fees and must constantly diversify and engage with
work [96]. Further, such labor may not fall equally. Haimson consumers to stay profitable in a competitive market.
et al. found that content moderation may lead to dispropor- The traffic to online adult websites grew [39] during the
tionate censorship of trans people and certain racial identities COVID-19 pandemic outbreak and associated lockdowns,
(e.g., Black), despite following platform policies [43]. providing a source of income for both sex workers and those
In a survey of 135 social media creators, Thomas et al. ex- without prior sex work experience [45, 47, 78]. Hamilton
plored online risks, including targeted attacks, hate speech, et al. found that the in-person sex workers who shifted to
threats of violence, impersonation, account hijacking, stalk- online-only sex work during the COVID-19 pandemic expe-
ing, and false reporting [91]. To cope, creators often resort rienced increased digital exposure and faced platform loss
to self-censorship or may choose to leave a platform entirely. (deplatforming), stalking, harassment, and content theft [45].
Samermit et al. in a recent interview study with 23 content Rubattu et al. performed a media analysis post COVID-19
creators, found that prominence, social norms, marginaliza- pandemic, finding that those new to sex work experienced
tion, and access to sensitive resources were reasons for online outing and harassment online [78], echoing pre-pandemic
risks, including toxic content, harms to family, stalking, pri- observations [53, 54, 82].
vacy risks, impersonation, and account hijacking or false re- Apart from the aforementioned challenges, sex workers
porting [81]. To cope, creators maintained privacy, employed also face risks stemming from different legislatures and so-
moderation techniques, and sought advice from other creators, cial norms across countries. Sex work legality varies from
but found platform support and legal aid hard to reach. country to country and even within individual countries. Prior
Sex workers. Sex work encompasses a diverse spectrum work finds that American legislation and social attitudes lead
of sexual labor, including stripping, pornography, and out- platforms (including payment processors) to censor sexual
door full-contact sex work [60]. Digitally-mediated sex work, expression and sex work to shield from liability and protect
defined as “the Internet-mediated exchange of sexual com- their corporate image—regardless of the legal status of the
modities and/or services” [53, 82], is performed by sex work- censored expression or labor in the U.S. or in the user’s coun-
ers who organize and market their labor through digital plat- try [1, 9]. To do so, platforms may use algorithmic shadow-
forms [12, 53, 75]. This form of sex work may offer increased banning2 [15, 92], place blanket bans in their terms of service
(ToS) against the use of their services for any sex work-related one. We therefore used the quota function to ensure demo-
activities [9, 13, 15, 16], or over-censor potentially objection- graphic diversity amongst those we recruited. Qualification
able content [4]. Stardust et al. theorize that sex work stigma for the study was limited to those over 18 and currently or
is a key driver for such policies and practices and argue for recently (within 6 months) creating content on OnlyFans. Par-
anti-discrimination protections for sex workers [86]. ticipants that qualified through the screening survey questions
Beyond censoring expression and work, prior work also were invited to self-schedule an interview through Calendly,
finds that sex workers themselves may be systematically iden- which allows for anonymous sign-up.
tified and de-platformed based purely on their identity as sex Participant demographics. We had a diverse sample of par-
workers, regardless of whether they are using the platform for ticipants across many identities with a median age of 30 years
that work [9, 16, 67]. Such censorship can lead to diminished (σ = 5.37); self-reported genders6 of woman (25), non-binary
financial stability, safety, and community access [7, 17]. (15), man (4), and trans (3); and self-reported race/ethnicity
OnlyFans, a UK-based platform, operates where many of: white alone (26), Black (5), Asian (4), Hispanic and white
types of sex work are legal and primarily facilitates porno- (3), Asian and white (1), and other (4). 26 of 43 participants
graphic services, which are legal in the US. Our work con- were disabled (3 choose not to answer). Of the first cohort
tributes to the broader literature on both professional content of participants, all were LGBTQ+ and we did not ask the
creation and online sex work as we investigate creator security second cohort to specify. We did not ask the first cohort about
and privacy on OnlyFans, a web-based social media platform their highest level of education or their income on OnlyFans.
focused primarily on sexual content and with a unique combi- Most of the second cohort had at least some college education
nation of attributes including lack of internal searchability that but no degree, with the rest having additional qualifications.
requires cross-platform advertising and interaction, a greater Among those who disclosed their OnlyFans income earned to
level of mainstream visibility than other sexual-content fo- date, the reported income ranged between $135 and $332,000
cused platforms [45,47], and a variety of subscription, paywall (σ = $83, 974.24). Out of 43 participants, 24 had prior sex
and direct interaction models of creator-fan interaction [79]. work experience, 19 did not, and three had recently left the
platform before the interview.
4 Methods Not all of our participants disclosed their number of fol-
lowers; but those who did disclose ranged from creators with
We conducted 43 semi-structured interviews with OnlyFans tens to those with tens of thousands of fans. The web-use
creators in two cohorts: 1) at the onset of the COVID-19 skills [48] of our 43 participants ranged from 1.33 to 5 with a
pandemic and 2) between September and October 2021. Ad- mean of 3.57 (σ = 1.21).
ditional questions regarding the impacts of policy changes
were incorporated into our questionnaire in response to Only- 4.2 Data Collection
Fans’ announcement and subsequent retraction of the ban
on explicit content in August 2021 (§ 2). Here, we provide Before starting the interview, interviewers explained the study
an overview of our participant recruitment, data collection, goals and answered any questions regarding confidentiality,
and data analysis approach, along with the ethical practices the research team and data use, and reminded participants that
followed throughout this multi-phase interview study. they could skip any question. We then re-affirmed consent to
record. We used a semi-structured interview protocol, starting
with background questions about their prior works and their
4.1 Participant Recruitment comparisons to OnlyFans experience, content creation, busi-
We recruited participants with a variety of backgrounds in- ness model, safety and privacy, impacts of changes in ToS,
cluding those with and without prior experience in sex work platform use, and overall OnlyFans experience. The relevant
on social media, through the researchers’ personal networks, interview questions for this paper can be found in Appendix A.
and for the second phase of recruitment, with printed flyers All interviews were conducted online in English via video,
in coffee shops, grocery stores and on university campuses in voice, or chat. They were transcribed and any identifying in-
multiple U.S. cities. formation was removed. Participants were compensated $50
The link and QR code on the recruitment graphic led par- through PayPal or Amazon gift card.
ticipants to a screening survey in Qualtrics (Appendix A). We
employed Qualtrics’ quota function with filters for age, gen- 4.3 Data Analysis
der, and ethnicity to achieve a diverse sample, encompassing
various identities that could impact participants’ experiences Study researchers employed thematic analysis [20] to gain
on the platform. Given that OnlyFans does not publish demo- insights into the online experiences of OnlyFans creators. Af-
graphic data about its creators and the inherent limitation of ter becoming familiar with the collected data, the primary
smaller sample sizes in interview studies, our aim was to pri- 6We report genders following the HCI Guidelines for Gender Equity and
oritize a diverse sample rather than pursuing a “representative” Inclusivity: www.morgan-klaus.com/gender-guidelines.html
coder inductively created an initial codebook [25] from a ran- important to note that due to the nature of the work and the
dom selection of five transcripts. This initial codebook was industry in general, some participants might have preferred
then cross-referenced with the online hate and harassment to omit relevant information. OnlyFans has a global reach,
framework [90] to add any missing themes and align the ter- and has creators and fans from many countries. We conducted
minology of the inductive codes with the framework where interviews with creators only from the Global North in En-
applicable. Two coders used this codebook to independently glish, which limits generalizations regarding other territories.
analyze five new transcripts and test the coverage of codes. By focusing on our participants’ shared (positive or nega-
They met to discuss code applications extensively and made tive) experiences, some insights might have been missed from
necessary refinements to the codebook, mainly developing those with contrasting opinions, however our participants’ var-
the codes into themes. After independently coding three more ied experiences on the platform might have balanced out the
transcripts, the coders achieved a substantial inter-coder re- bias. Lastly, the participant demographics include creators
liability score for Cohen’s Kappa (κ > 0.7) [58] and refined with genders or ethnicities which were not reflected in the
the codebook until any minor disagreements were resolved. interview or research team, limiting our insight into that data.
Coders used the final codebook to independently code the
remaining transcripts, evenly dividing them between both.
We primarily report our findings qualitatively using partici- 5 Results
pant quotes, shared experiences, and incidents to describe and
interpret the themes that emerged from the interviews. While After analyzing interviews with 43 OnlyFans creators through
reporting, we occasionally provide counts (i.e., how many the lens of the online hate and harassment framework [90],
participants reported each risk or safety practice) of overall we now delve into their perceived and experienced risks and
themes to demonstrate the prevalence of certain patterns and how they defend against those risks.
ensure transparency in our data analysis.
5.1 Platform Risks
4.4 Ethics Participants were most concerned (29 participants) about risks
from platform precarity and censorship.
The research procedures were approved by the institutional
ethics boards of the researchers. At every stage of the research, Platform precarity. The policy of banning sexual content
we undertook great care to provide the highest possible level from OnlyFans, which was later reversed (§ 2), had a lasting
of protections for participants’ data privacy and anonymity. impact on our participants. The impacts of this ToS change
We used end-to-end encrypted platforms for the interviews were not only financial: “people were not re-billing or re-
(paid Webex) and no systems required legal information or subscribing” (P11), but also emotional:
login. The interviews were scheduled using Calendly and we When you deplatform creators you take away their
provided advice on signing up with an encrypted email ad- community and that’s violence. You are isolating
dress if preferred. The payment options could both be anony- people and as we all know from this last year isola-
mous (Amazon gift card) or to any address (PayPal). tion leads to severe mental health consequences.
Research justice. We employed a sex worker to transcribe the P9 reported that the change in ToS and resulting media cov-
interviews. A copy of the published research will be sent to erage elevated the platform’s visibility, leading to increased
all participants who requested it. We also employed OnlyFans harassment toward creators: “we didn’t even get to find out
creators to work as consultants on designing the study (includ- from the platform itself, we found out from news outlets and I
ing the interview protocol) and they circulated the recruitment think it also exposed us to a lot of harassment on the internet
graphic within their networks. The authors’ commitment to from people who don’t like sex work.”
research justice contributes to their trust capital as researchers OnlyFans explained the ToS change as being a result of
of a population with research fatigue. This trust capital is pressure from financial institutions [77]. Some participants
essential in such research with a marginalized community. shared this view, explaining that “banking institutions are now
deciding what is moral and what is not moral to purchase...
Positionality. To design the study, recruit participants, and it directly impacts people on OnlyFans” (P11). Others as-
write the interview protocol, the research team included Only- cribed intentionality and responsibility to OnlyFans. 14 par-
Fans creators as consultants. The data analysis and writing of ticipants felt that OnlyFans “is uncomfortable with being a
this paper were carried out by researchers who are scholars of platform that hosts sex work” (P12), “is trying really hard to
technology and sex work but who are not OnlyFans creators. say no, there are no sex workers... it’s mostly chefs and fit-
The OnlyFans creator consultants who assisted in the study ness influencers” (P39), and was “cleaning up their image” by
design were not participants. banning sexual content (P26). P22 felt that “what happened
Limitations. OnlyFans is an extremely competitive market. with the ToS revealed that OnlyFans simply does not believe
Whilst we felt our participants were very honest with us, it is sex workers to be their primary creative force, which I think
they are.” After the ban, 14 participants—new to creating sex- kissing, and we were two Asian women, I’m South Asian,
ual content—expressed feeling less safe due to the increased she’s Asian, and that got deleted right away... similar photos
platform precarity. P6 stated: of my white counterparts weren’t censored... more explicit
In terms of like the decisions that the platform’s photos.” Conversely, P30 described his experience:
making... they just use sex workers and then will My risk of being censored is a lot lower as a male
sell them out. I feel physically safe you know, but performer, and as a white performer. So I’m quite
like not as politically safe as I did at first if that amazed my Instagram is still available and intact
makes sense. while other people have lost their account a thou-
Platform censorship. Since OnlyFans lacks internal discov- sand times in that same time period.
erability, creators work cross-platform to drive fans to their Participants tried to hypothesize about the cause of such cen-
accounts. These platforms may remove, block or restrict the sorship, pointing to co-moderation (P42: “I’m trying to figure
content they share, shadowban their accounts, or deplatform out if it’s someone reporting me or if it’s Instagram itself” and
creators, even if the platform’s ToS are not violated. P9: “TikTok is huge for that [censorship]... since they have a
Twitter does not ban pornography and other forms of user-based report system”) and suppression of sexual expres-
consensually produced adult content when marked “sensi- sion (P27: “I feel like the algorithm interprets that [pictures
tive” [35]. But P32 experienced shadowbanning on Twitter in panties] as little less human”).
for “overly promoting OnlyFans.” Although platforms by and Fear of violating platform norms can lead to creators proac-
large hide the criteria for shadowbanning content [4], our par- tively silencing themselves or leaving content creation [91].
ticipants have identified behaviors that seem to impact their P39 explained how the disproportionate impacts of platform
reach. They perceive that the likelihood of shadowbanning censorship can end up silencing some more than others:
while promoting on other platforms increases when creators It’s frustrating to be constantly living censored both
using automation tools are perceived as bots (P27), when they on Twitter and Instagram. It’s a systematic way of
use hashtags often (P25, P31, P33, P38), or send the same silencing people of color and queer folks so we just
message with an OnlyFans link too many times in a period give up and stop posting which essentially is what
of time (P23). Creators become aware of their content be- happens. It’s effective, because it’s exactly what
ing shadowbanned from platforms only when they observe happened. I gave up.
less engagement or income generation. P23 described how To make platform moderation more equitable for marginalized
fear of shadowbanning creates cascading promotional chal- social media users, Haimson et al. suggest platforms collabo-
lenges because “other colleagues don’t tend to retweet you rate with marginalized communities and develop censorship
when you’re promoting your OnlyFans, probably out of fear policies and enforcement mechanisms that better reflect the
of getting shadowbanned themselves.” voices and needs of the community [43].
Prior work on professional non-sexual content creators [81]
notes that creators are highly aware that violation of platform 5.1.1 Defenses
norms comes at a high cost—losing access to the platform
To defend against platform precarity and censorship, partici-
and its associated audience. Our participants similarly feared
pants diversified the platforms they used to sell content and/or
violating OnlyFans’ and other platforms’ ToS. 13 of the 22
engaged in self-censorship on the platforms they already used.
participants in our second cohort reported that they had either
read or skimmed through the ToS. P9 explained their proactiv- Platform diversification. 12 participants diversified the plat-
ity in carefully reading the ToS: “they [OnlyFans] often ban forms on which they created and sold content. While for
people and they [the creators] don’t get their pending balance. some, like P10, diversification was purely to defend against
It’s not a risk that I’m willing to take.” precarity (“the reason I’m building out on another platform is
Even when a post does not violate ToS, censorship oc- just in case something like that [sexual content ban] happens
curs due to factors such as co-moderation (where other users again, I definitely want to be diversified even more”), others
participate in moderation by blocking and reporting other peo- diversified to avoid violating ToS. P12 publishes “prohibited
ple) [15], discrimination against expression of sexuality, sex- content to another [sex work] platform [that allows it] instead.”
ual identity [15], poor implementation of not-safe-for-work Participants chose platforms for diversification based on their
(NSFW) moderation [92], or when creators post content that longevity (P18) or recommendations from other creators (P3,
is legally sensitive—deemed criminal in some countries [4] P4, P15). While some participants actively used the accounts
or perceived as advocating against policies and laws [15]. they created on other platforms, others username squatted—
In line with prior work on racial, gender and body size created accounts with their preferred username to reserve it
variance in platform censorship of sexual expression [36, 43, in case they were deplatformed or OnlyFans ceased to exist
66], participants underscored the variance in censorship across in the future—but did not actively use their other accounts.
creators. P39 explained: “I was so heavily censored... I had Self-censorship and Algospeak. Participants self-censored
a photo of me and my friend that was clothed, and we were on social media platforms to reduce platform harms (n = 24,
§ 5.1) and manage their identity (n = 22, § 5.3.2). provides messaging service on OnlyFans, was tired of people
P29 created a cheat-sheet to help them self-censor on Only- “pushing to learn more to get a little more familiar.” P13 expe-
Fans: “when I started I wrote everything on paper that we riences mental distress when fans repeatedly push boundaries
can’t use, so I’m always looking at that sheet for like okay, and “be disrespectful,” making it “complicated” (P25) for
this I can, this I can[’t].” On advertising platforms (e.g., so- creators to effectively manage fan interactions.
cial media), participants also invested significant effort to Boundary pushing can extend beyond requesting additional
ensure that they comply with all non-adult-content related information or content, reaching into demands for more of the
regulations and ToS (e.g., P33 avoids posting with music that creator’s time. For example, P39 said “people were expecting
“could potentially be flagged for copyright”). instant replies all the time from me” and P40 added “you had
Algorithm Speak (Algospeak) encompasses strategies of to be online 24-7, every time you’re not answering you’re
abbreviations, deliberate misspelling, or substitution of words missing out on money and I did not like the stress of that.”
to avoid automated content moderation systems [2]. This 8 participants mentioned experiencing other forms of ha-
approach is commonly employed by social media users to rassment and general bullying on the internet, on OnlyFans,
circumvent platform censorship [55]. To circumvent censor- and in-person as a result of their work on OnlyFans. P11 de-
ship, our participants “try to be vague as possible to make scribed one of the rude comments: “some people just say
account[s] last as long as possible” (P25) or used acronyms: that... you’re just a cheap whore... why would anyone want
“if I tweet about OnlyFans, I’ll use the acronym OF...” (P8). to date you if they can just see you for free.” These negative
P41 told us: “there are many words that are forbidden [on experiences take a mental toll (P20) on creators and can do
OnlyFans]... obviously the words can be taken out of context psychological harm [10, 64, 83]: “online harassment is going
and not always mean that thing.” Therefore, our participants to affect [me] more mentally than anything else” (P11).
use different spellings (P41, P39, P28) or emojis, symbols, or Online hate. Six participants faced online hate based on their
synonyms for restricted words (P23, P26) to not get flagged appearance, race, or gender. P12 observed “people being trans-
or blocked on-platform when they are posting or chatting. phobic or racist” and like P31, P21 said “a lot of women of
In some cases, using Algospeak was too cumbersome and color and trans people who are not thin and white, experi-
participants instead engaged with clients off-platform: ence the brunt of hate.” P31 emphasized the mental toll of
Once, I got a request to write a dirty story for a increased hate online relating to more visibility:
client and I had to drop it in the chat and every With online work, because of how many more peo-
fourth word was a word that could not go through ple you’re seeing, you’re exposed to a lot more trolls
on OnlyFans... I ended up emailing the story. (P42) and time wasters than you would in person because
Finally, some tweaked their profile attributes to avoid bias it’s just the volume and number of people, especially
in the censorship process. P34, who intentionally changed if you’re not white or thin or cis.
their gender (to male) on Instagram explains how it reduced P21 noted varying types of hate on different social media
their experiences of censorship: “I’ve been kind of pushing platforms, attributing it to the demographics of both users
the envelope with Instagram but I haven’t had anything taken and creators. For example, they perceived Reddit to have an
down in a while... it’s really sad but [changing my gender] audience “dominated primarily by white American men” and
made a huge difference.” who preferred content from “thin white women” whereas
“Twitter is much more diverse with its user base.”
5.2 Toxic Content
5.2.1 Defenses
Like at-risk users [90] and other professional content cre-
ators [81, 91], 26 of our participants described dealing with To address hate and harassment, participants established clear
different forms of toxic content such as harassment and online rules (n = 18) or blocked (n = 13) attackers. Additionally,
hate from fans directly on OnlyFans as well as from other some participants call on community support (P22: “if some-
platforms that caused varying degrees of emotional harm. one posted something mean I would just retweet it and then
Echoing professional non-sexual content creators’ expres- the entire internet would ask to delete their account”) or re-
sions that such hate and harassment is unavoidable [91], our spond directly to their harasser themselves (P22: “I’m very
participants noted that toxic content is “just part of being on skilled at handling [hate and harassment] because I’m sassy”).
the platform” (P20) and there are “just clients that have no Others simply ignore the harassment: “I have a really really
idea how to message people” (P39). thick skin from working in this industry for a long time” (P11).
Harassment and bullying. Participants commonly (n = 19) This strategy is similar to how other populations manage ha-
experienced boundary pushing where fans become “overly rassment [81, 90, 91].
persistent”—exceeding or violating personal or professional Boundary setting with fans. To address the most common
limits set by the creators—and P17 indicates that “it’s a fine form of harassment, boundary pushing, participants made sig-
line between being overly insistent and obsession.” P1, who nificant efforts to establish strong boundaries with fans. P10
received uncomfortable messages and pricing negotiations something that unfortunately is part of the job” (P31).
on their free page, but when they observed that individuals Prior work on image-based sexual abuse finds that perpetra-
on their paid account exhibited comparatively more polite tors may leak content in order to raise their own social status,
communication, they deleted the free page. To avoid repeated using the content they non-consensually obtain and distribute
requests, P4 has added a statement to their bio: “you are wel- as a form of social currency [31, 49]. In line with this, P36
come to send requests but... I have every right to deny them.” describes their perception that:
During the interview, P2 said, “I mean people are rude for sure The usual porn user is kind of like a hoarder of porn,
sometimes but there isn’t a rating system like in so many gig they save all the porn and try to put them again on
work apps.” P38 increased their prices to “weed out assholes” platforms to share them around like if they were
and P36 shared that “they [fans] always wanted the cheaper, kind of football cards.
the best, even free... I don’t have to give them what the broke In the transactional context of OnlyFans, content theft de-
boys want.” P30 clearly communicates their boundaries to prives creators of income. Attackers may thus be motivated
fans, but they acknowledge that “expressing boundaries can to engage in content leaks to obtain content for free or to
be hard work and really difficult.” For P18, not having to ne- leverage fraudulently obtained content to profit financially, as
gotiate their established boundaries and saying that “these are detailed next.
boundaries for a reason” gives them a sense of privacy.
Blocking. 13 participants indicated that they employ the plat- 5.3.1 Harms of Content Leakage
form’s block feature when encountering spamming or harass-
Content leakage can cause harm in a variety of ways, and
ing clients. Upon receiving uncomfortable requests, some
enable additional attacks.
participants first tried to warn the fans about their boundaries
but would also “restrict them or block them depending on the Privacy violation & image-based sexual abuse. While Only-
severity... how aggressive they’re being with it” (P9). Simi- Fans creators consent to post and share their intimate content
larly, P4 does not react instantly to repeated requests but “ends within the platform’s transactional context, they do not con-
up blocking people who just kind of go out of hand.” sent to it being further shared nor viewed outside the context
P43 found it easier to handle fan issues online as compared of this platform and transaction. Thus, leaks of creators’ inti-
to in-person sex work: “[there] was nothing I couldn’t solve mate content are a form of image-based sexual abuse [49, 90].
by blocking.” P11 emphasized the significance of blocking, Such abuse is a violation in itself and has significant emo-
stating, “if someone is being really crappy towards you, just tional and psychological consequences [49, 80, 100]. The per-
block them... OnlyFans work is also the biggest privilege manence of intimate content leaks (e.g., P41 describes fearing
within sex work.” According to P19, “the only safe thing their leaked content “being on the internet forever”) and sever-
OnlyFans offers is blocking.” P10 further elaborated on the ity of consequences from such leaks make them a “severe”
convenience and practicality of the blocking feature: form of hate and harassment, as defined by prior work [90].
I’ve blocked way more accounts on these two Only- Outing. Attacker motivations for content leaks may vary. At-
Fans accounts I’ve had for a year and a half or two tackers may also non-consensually distribute others’ intimate
years... once they [fans] start pushing I’m just like content as a targeted form of abuse by doxxing creators and/or
no, blocked. I’m not even going to deal with you. outing them as members of stigmatized communities (e.g., the
It’s like no is a full sentence. sex worker and/or LGBTQ+ community). Attackers may do
When engaging in off-platform advertising, creators, like P18, so by leaking creators’ content and linking it to their personal
“made heavy use of block buttons” to manage heightened information [22], which they can discover through a variety
harassment due to sex work stigma. of means including using “facial recognition software... [to]
find your old [pictures] from high school and... your name
5.3 Content Leakage and where you grew up” (P9). Such attacks can cause rela-
tionship, reputational, and physical harms [83]. Similar to 13
Content leaks [90] were the next most common (n = 24) risk other participants, safety to P11 is:
our participants reported. Although creator content on Only- ...being able to do what I do and not worry about be-
Fans is paywalled (§ 2), it can be illicitly captured in violation ing doxxed and having stalkers and people showing
of the ToS and the creators’ consent by screenshotting and up at my house to potentially do harm to me.
downloading. Additionally, OnlyFans itself can be hacked: in P9 mentioned the potential “safety risk” of “families kicking
February 2020, an estimated 1.4 to 4 terabytes of OnlyFans [creators] out of the house” if they found out about their work,
videos and photos were shared across social media [28]. while P22 linked “getting outed as a sex worker” to “receiving
Multiple participants viewed content leaks as inevitable: professional retribution.” P1 fears how permanent content
“there’s always a possibility for that [content leak]” (P4), “no leaks could impact their ability to get other jobs in the future
matter what you do leaks are inevitable” (P21), “it’s [con- after they finish their schooling, because the “engineering
tent is] going to eventually get stolen”(P27), and “it’s just world does not look too highly upon OnlyFans.”
Chargebacks. Chargebacks happen when a purchaser falsely motivated by the fear of having their location discovered or
disputes a product charge, often by reporting credit card fraud having their non-OnlyFans professional progress hindered. P4
or claiming they do not recognize the payments on a bank and P7 expressed contentment with not having a significantly
statement. The sex industry has particularly high chargeback large following as the “stress from that is not great.” Given
rates due to the discreet and stigmatized nature of purchas- their limited prominence, P7 was confident that their content
ing sexual content [21, 86, 94, 105]. For OnlyFans creators, would not be widely shared, while bigger accounts have to
chargebacks are passed on directly to creators, meaning a “deal with large chargebacks and take the heat.” P23 was wor-
client gets access to their content without ultimately paying ried that increased visibility could impact their interpersonal
for it, violating their consent and impacting their revenue. relationships and privacy:
Four participants in our work described being victimized by It exposes me to much more... there’s much more ex-
chargebacks. P23 explains: posure, there’s not much privacy, any ex-boyfriend
So, if a guy gives you loads and loads of money and or stalker can look at my OnlyFans and they’re see-
then decides to report that he got his credit card ing my sex work in a way they couldn’t before.
stolen, then you don’t get that money back. On the other hand, because OnlyFans lacks internal discovery
P7 similarly described their experiences with “the people who and creators engage in cross-platform advertising, managing
come in and drop a whole bunch of money and then back off-platform prominence becomes challenging: “my advertis-
track it.” However, they proactively started charging extra for ing is everywhere, and I know TikTok advertises more wher-
content they particularly wanted to protect and told us: “it ever your SIM card is based and I’m just using my regular
would be nice if OnlyFans fixed that loophole.” phone so they’re pushing my videos out more locally” (P24).
Catfishing. Catfishing7 is an impersonation attack [90] in Internet presence monitoring. 17 participants actively im-
which an attacker creates a fake online profile to trick peo- plemented internet presence monitoring schemes to check
ple, usually for financial gain. In the context of OnlyFans, for leaked content. P40 said: “the likelihood of it [content]
catfishing means someone copying creators’ videos or posts getting leaked was giving me anxiety, OnlyFans leaks are
to impersonate their OnlyFans identify elsewhere and engage very common and I was constantly googling if my shit got
in fraud. P9 described that scammers create fake Instagram leaked.” P2 searches variations of their name on Google to
accounts using OnlyFans creators’ content to deceive fans, see “what’s out there and delete-able.” P35 does the same,
which damages creators’ reputations and harms their follow- and also checks PornHub and XVideos for stolen content to
ers. P12 and P22 shared instances where they found catfish “not have videos out there for free.” P9 uses image search to
profiles made using their images on Grindr. Additionally, P11 look for their leaked content on Reddit and Google. They also
encountered their stolen images on Reddit, where the account leverage a subscription service that searches for their name
attempted to defraud their fans. Fortunately, they were able and scrubs identifiable information from the internet.
to have these accounts taken down as they had watermarked 11 participants reported using DMCA takedown services8
their images (see below). to remove stolen content from the internet. However, some
participants hesitated to use these services either because of
5.3.2 Defenses skepticism (P27: “it’s not even like a guarantee”), the ser-
vice’s convoluted process and potential non-responsiveness
In response to the threat of content leakage and its cascading (P33: “this might be really complicated, what if they don’t
attacks, our participants employed a variety of protections. respond?”), or the financial burden (P40: “this would take a
P9 mentioned how an OnlyFans creator has to engage in lot of cash”).
adversarial thinking and “get in the minds of people who Two participants (P12, P22) who experienced catfishing
could potentially want to harm you.” They summarized some received help from others in reporting catfishers. For example,
of the technical protection strategies they were aware of: when P22’s photos were used to create a fake Grindr account,
There’s a lot of services out there that will get your they made their fans aware of the catfish by tweeting and
name scrubbed off of voter records or at least your asking people to report the account.
address so no one can show up at your house... be During the interview, P8 and P40 mentioned the feature on
very careful with not having your Snapchat location OnlyFans that supposedly prevents users from taking screen-
[on]... photos taken on iPhone have your informa- shots or recording content. P8 described the similarity to
tion encrypted inside of them. platforms like Netflix, where attempting to screenshot results
Limiting prominence (and profit). Seven of our less promi- in a black screen, adding “I never actually tried it out to see
nent participants intentionally aimed to limit their popularity, if it works, I sort of just trusted it.” However, OnlyFans does
not currently have any features in place to prevent screen
7 Prior literature has sometimes referred to similar attacks using an attacker

term of art—“eWhoring”—to specify catfishing fraud conducted using “sex- 8 A DMCA takedown happens when the owner of content claims a

ualized images” [51]. The use of the term “whore” is however disrespectful copyright violation and requests a platform remove the content, https:
and therefore we opt not to use it [97]. //www.dmca.com/
recording. This highlights the misconceptions that can exist ing information online, as P7 mentioned: “Sex workers are
in participants’ mental models of platform-provided security more at risk as soon as they put nudity online... I do more to
tools. keep my identity hidden on my OnlyFans account rather than
Watermarking. Six participants relied on OnlyFans’ in- my personal accounts.”
platform content watermarking to deter content theft ( § 5.3). To that end, 10 participants concealed their faces in their
However, 14 participants practiced proactive watermarking to posts to maintain anonymity. For instance, P4 angles their
protect their content from theft and unauthorized use, as well camera to exclude their face in nude content, but reveals their
as to trace specific people who stole the content via person- face when portraying a cosplay character. P1 hesitates to use
alized watermarks. OnlyFans does not challenge intellectual blurring software to conceal their face due to concerns about
property infringement on behalf of its creators, and so the potential de-blurring, and instead opts to physically cover
responsibility for recovery is left to individuals. Prominence and crop their face before posting for “double security.” P33
also influences creators’ decisions about how to protect their shared an incident where they did not hide their face in the
content. P43 deems the extensive efforts of watermarking “most clever way”:
“probably not worth it at my level of visibility.” P11, P20, and Sometimes I hide the top part of my face and some-
P37 recognize that leaked watermarked content also serve times I hide the bottom half of my face, and I had a
as a kind of promotion, as people can identify the creators guy who had stitched together the top half and the
through the username on the content, which potentially helps bottom half and he sent it to me like look, I’ve made
recover lost income. a picture of your face, this is what you look like!
P3 avoids posting identical content on their OnlyFans and
Keeping online identities separate. Similar to online account NSFW Twitter accounts to mitigate the risk of reverse image
management strategies of in-person sex workers [67], many searches linking pictures of their face across platforms.
participants (n = 22) kept their personal social media accounts 16 participants proactively used aliases to protect their legal
separate from their OnlyFans accounts or the social media name. P13, a cosplayer, mentioned: “...with cosplayers, it’s
accounts used to promote OnlyFans. P25 keeps their social so easy to get people who are just creepy and trying to find
media accounts separate to prevent platforms from suggesting my Facebook account. Then, they were finding out where
their profile to unwanted individuals. P33 reduced their posts I worked, where I lived... going by a different name adds a
on social media after an incident: layer of security to anything that I do.”
While I was on holiday I posted a picture of my- 18 participants concealed their location on OnlyFans using
self on the beach on my main Instagram which my security-focused tools such as VPNs. P17 felt “super secure”
friends and family see and on my porn Instagram, using VPNs because “just in case somebody was to get a
and I tagged my location, and then someone found weird idea to try to figure out my location it wouldn’t even
the main, and I was like that was so dumb, I can’t ping off to what state I am in.” P18, on using a VPN, said “I’m
believe I did that. assuming I’m not the only one who’s heard horror stories. So
Participants’ efforts to separate their online identities exist on like just making sure to have like a level of semi-anonymity
a spectrum. Seven participants who were already open about in there.” Some participants further hid their location by, e.g.,
their sex work in their personal lives did not worry about the avoiding taking photos near identifiable landmarks (P9, P10),
risk of getting outed. For example, P32 said “I was already turning off geo-location from images and videos (P10, P17,
face out, so I wasn’t necessarily concerned about clients tak- P19), and removing image meta-data (P16).
ing screenshots of private photos that were revealing my face
and sharing them and outing me.” One participant (P34) even Managing outness offline. 16 participants proactively
switched their personal Instagram account to a sex work ac- avoided disclosing (“a polite omission” - P12) their Only-
count and deliberately outed themselves. Fans work with certain people in their life, primarily because
of the stigma around sex work (P1: “this person has no dig-
Saving username on different platforms. To minimize im- nity, no morals”) or to reduce relationship harm (P11: “I don’t
personation risk, five participants strategically saved their think that they [parents] would even understand what is going
usernames on other platforms. To that end, P33 said: “I did on”) [83]. Maintaining secrecy posed challenges for some;
worry that someone would want to impersonate me on there P10 consistently faced stress while upholding a “cover story.”
which is why I thought it’d make sense to get the username Similar to 13 participants, P8 opted to identify as a content
[on another site].” Creators sometimes encounter challenges creator to navigate interpersonal and professional relation-
in securing their usernames, as highlighted by P16: “someone ships without revealing their OnlyFans association.
took my username, some cis guy from UK who doesn’t even
do anything, he just watches people, he doesn’t post anything,
he doesn’t stream. It’s just really annoying.” 5.4 Adopting Digital Safety Practices
Limiting identifying information on OnlyFans. Our study In contrast to prior work on non-sexual professional content
participants employed various strategies to hide any identify- creators [81], where two-thirds of participants adopted protec-
tive practices only reactively, we observe that before joining posture, they also engaged in ongoing learning. To that end,
OnlyFans, all of our participants adopted one or more protec- 13 participants actively monitor Twitter and Reddit to learn
tive practices to avoid content leakage (§ 5.3.2). While most about safety tips on OnlyFans: “I follow like keywords for
participants focused on protective practices to preemptively safety and security on OnlyFans” (P13), “there’s a subreddit
keep their online identities separate, some also preemptively and there’s a lot of really knowledgeable creators on there
limited their prominence, set up internet presence monitoring that can tell you what to do and how to stay safe” (P9), and
tools, watermarked content, limited identifying information “I check reddit, check top content creators, see how they feel
on OnlyFans, and managed outness about OnlyFans work of- about safety... they will post scammer IDs and stuff like that
fline. Additionally, some participants became early adopters so you can go ahead and pre-block them” (P7).
of strong boundaries to defend against toxic content (§ 5.2.1).
Learning sources. Participants learned these initial practices
6 Discussion
from fragmented sources of digital safety advice [76] and
from their own “common sense” (P10) and technical skills, Through 43 semi-structured interviews with sexual content
like “tech nerd” P16 and “Google queen” P20. For exam- creators on OnlyFans, we documented creators’ online threats
ple, P8 engaged in preliminary research on OnlyFans’ safety and safety strategies. In this section, we synthesize our re-
mechanisms to assess the platform’s security features: sults using our research questions as a guide. We conclude
So the types of like encryption that they use or how with an exploration of the opportunities for better security
they watermark photos... definitely didn’t research affordances.
in depth but like very slightly like legality and mak-
ing a claim in your account about like if people
were to redistribute it [content]. 6.1 Creators’ Threat Landscape: A Hate and
Friends also served as sources of knowledge. For example, Harassment Framework Perspective
P13 learned their safety practices through a friend’s doxxing
experience in addition to their own stalking experience as a We used Thomas et al.’s online hate and harassment frame-
younger cosplayer, and P22 received information from friends work [90] as a lens to understand the threat landscape of
about general internet safety. P4 on the other hand learned OnlyFans creators (RQ1). Our analysis reveals that OnlyFans
to watermark any content they share from observing their creators experience digital risks common to other “at-risk”
friend’s experience of content leaks, and seeing their friend users—toxic content and content leakage—as well as more
use watermarks to help trace who had leaked their content. unique risks from platform censorship and precarity.
As reported in prior work [7, 87, 88, 95], our participants Similar to non-sexual content creators [81, 91], our par-
also learned certain safety mechanisms from OnlyFans and on- ticipants experienced a high volume and expressed a high
line sex-work-specific communities. 16 participants derived degree of concern about hate and harassment attacks. In con-
reassurance and support from being an active member of a trast with non-sexual content creators who prior work finds
strong and closely-knit online community of content creators, predominantly face attacks related to toxic content [91], our
where they learned about safety and privacy, content creation participants experienced the most frequent and concerning
and marketing strategies, and other OnlyFans-related experi- threats from platform censorship and precarity.
ences. P13 felt from the beginning that they could establish Prior work finds that some non-sexual content creators
boundaries about the type of content they create when they worry about violating platform ToS and resulting platform
saw others opting out of creating very explicit content. From censorship [81]. Our participants expressed such concerns
being part of the OnlyFans community, P36 learned tips to in greater volume and with greater concern, and additionally
grow as an online sex worker, perseverance, as well as bound- raised concern about platform censorship as a result of biased
ary setting (“they taught me that I don’t have to cater to what moderation of ToS-compliant content due to the explicit and
the boys want”). Likewise, P21 learned about “basic things stigmatized nature of the content they share. Further, our
such as knowing your boundaries, to not feel uncomfortable, participants expressed significant concern about the precarity
how to protect yourself online, and a little bit of anonymity” of OnlyFans as a platform. Combined with threats creators
through a sex work industry video. face from people—of toxic content and content leakage—the
Preparedness. Participants mentioned that they learned to be threats of platform censorship and precarity foster a strong
proactive in their defensive approach because of the societal sense of insecurity among creators (§ 5.1).
stigma around sex work: “there’s definitely a little bit of a
stigma... I can’t exactly go on my Instagram and start broad- 6.2 Role of Stigma and Digital Prominence in
casting or I’m gonna get a label [for] myself” (P1). Recog- Creators’ Threat Landscape
nizing potential risks, P9 advocated for proactive preparation,
stating, “ignorance is not bliss in this situation.” Our results show that awareness of sex work stigma prompts
While participants were very proactive in their security OnlyFans creators to proactively anticipate and prepare for
online risks (§ 5.4), in contrast to non-sexual content creators provided them a sense of empowerment9 and avoided reliance
who tend to approach digital safety in a more reactive fash- on the platform for justice. We hypothesize that our partic-
ion [81]. Participants often turned to their community, either ipants prefer not to rely on the platform due to experiences
passively through following community-members on social of platform precarity. For example, when OnlyFans suddenly
media, or actively by seeking out support from friends and banned explicit content (this ban was later reverted, see § 2),
community members. Our findings support prior work illus- creators were left with a sense of helplessness (§ 5.1).
trating the influence of stigma on digital communities, where Additionally, some platform policies increased participants’
shared experiences of stigma foster a strong sense of solidar- apprehension that the platform would adequately safeguard
ity and collective support [7, 33, 63]. Together this highlights their interests. For example, chargebacks pose a significant
the role of stigma in digital security education and offers challenge in the sex industry. OnlyFans passes such risk on to
evidence for its influence on differential threat models and creators directly: the community guidelines on chargebacks
protective behaviors between otherwise similar groups of end state that “any amounts that users seek to refund or chargeback
users (e.g., between digital content creators who do and do will be reported to the creator and will be removed from the
not do stigmatized work). creator’s income” [24]. This places a considerable financial
Stigma also intersects with prominence, influencing cre- burden on the creators, treating them as de-facto e-commerce
ators’ online risks. Prior research revealed that social media merchants without affording them control over payments—
content creators, even without the added layer of sex work typical to an e-commerce merchant—or the ability to contest
stigma, faced increased risks with growing popularity [81]. chargebacks effectively.
Experts advise general Internet users to approach online par- Beyond OnlyFans itself, our participants experienced
ticipation with a minimalist mindset [100]. However, the con- and/or perceived that mainstream platforms may be biased
tent creation industry inherently requires information sharing toward shadowbanning their content or deplatforming them
for monetary gain, rendering a minimalist mindset impracti- due to the stigmatized nature of their labor and the content
cal. Particularly on OnlyFans, which lacks internal searcha- they share (§ 5.1). Consequently, they might be reluctant to
bility, off-platform information sharing becomes a necessity depend on reporting tools provided by platforms they perceive
for creators to promote their content, attract subscribers, and as indifferent to their safety.
generate revenue.
OnlyFans creators actively strive to maintain separate on- 6.4 Suggestions for Solutions
line identities to overcome challenges posed by stigma and
balance the necessity of increased prominence with preserva- Affordances for diversification. Our participants practiced
tion of personal privacy. However, effectively maintaining pri- platform diversification strategies (e.g., saving usernames on
vacy while prominently engaging online remains a challenge. other platforms) to avoid sole dependence on a single source,
Future work may seek to address this challenge by develop- such as OnlyFans, for income (§ 5.1.1). The precarious nature
ing privacy-preserving technology to better support persona of sex work platforms also means that an account and all data
management [46]. Such work can help non-sexual profes- within—photos, videos, messages, connections with fans—
sional content creators [81] who must manage their identities could be deleted arbitrarily and without notice [4]. However,
due to prominence, recreational content creators who are es- diversification can be extremely challenging. It requires a
pecially at-risk because of a stigmatized identity [61], and significant time commitment and effort to maintain multiple-
professional sexual content creators such as the OnlyFans platform account activities effectively [44]. For instance, dur-
creators we study who are at risk due to both prominence and ing OnlyFans’ ban on sexual content (§ 2), participants ex-
stigmatized labor. pressed a desire for sufficient time to retrieve pending deposits
and “safely delete content or transfer it [content] to other plat-
form” (P4). Moving to alternative platforms requires critical
evaluation based on factors such as popularity (in comparison
6.3 Platforms’ Impacts on Creator Risks
to OnlyFans) and fan engagement. Only half of P11’s fans
Experts recommend using platform-provided tools to mod- were willing to move to other platform and P12 said “cre-
erate abusive messages online [100]. However, both sexual ators did not experience the same success that they had on
and non-sexual content creators are skeptical about the sup- OnlyFans” when creating new accounts, attributing it to the
port offered by platform tools. Although non-sexual content differing audience presence.
creators used keyword blocklists and manual reviewing to There have been various attempts to increase consumer
address hate and harassment, they expressed skepticism about control over personal data. For example, the General Data Pro-
the effectiveness of reporting users [91]. tection Regulation (GDPR) Article 20 [37] asserts individuals’
right to “receive their personal data in a machine-readable
Similarly, while OnlyFans offers reporting options for ad-
format and transfer that directly from one data controller to
dressing hate and harassment, our participants predominantly
opted for blocking fans in response to such incidents, which 9 “If someone is being really crappy towards you, just block them” - P11
another (wherever possible).” This right to data portability, genuine social media accounts and report fraudulent accounts.
in theory, offers consumers significantly reduced costs for These mechanisms align with established phishing detection
switching between digital service providers [57], and could and mitigation techniques: content classification and block-
play a crucial role in enabling OnlyFans creators to diver- listing [62, 101, 103, 104], user training and education [29],
sify their online presence and mitigate risks associated with content take-down [3], and direct reports from users [40].
platform dependence.
Despite the fact that the GDPR has been in force for several Chargebacks and financial harm reduction. Although only
years, the right to data portability is still, in practice, largely four participants discussed the unique risk of chargebacks
unusable [89]. Requests can be cumbersome, data can be in in our study (§ 5.3), chargebacks on OnlyFans are common
unusual formats, and new platforms rarely have a method for largely because of sex work stigma [11]. For example, a fan
ingesting data from another. With further advancements in may request a refund for their OnlyFans transactions to hide
enforcement and usability, data portability can offer creators it from their spouse on their bank statement. To free cre-
stability and facilitate seamless content and data transfer. ators from bearing the brunt of chargeback liability, OnlyFans
could challenge chargebacks on behalf of their creators or
Internet monitoring. OnlyFans’ ToS recognize redistribution
implement a platform currency: fans purchase tokens from
of content by fans on other platforms as a violation, but it
OnlyFans, and then use the tokens to pay creators (who are
neither prevents users from recording their screen nor enforces
then, in turn, paid proportional to the tokens they earn). The
intellectual property rights on behalf of creators [34].
streaming platform Twitch has already implemented a similar
To search for and take down leaked content, participants
system to challenge chargebacks [32].
practiced extensive “internet monitoring” (§ 5.3.2). This prac-
tice is time-consuming and emotionally draining, taking cre-
ators’ time away from running a profitable business and lim-
iting freedom of self-expression. Although the platform pro- 7 Conclusion
vides a free watermarking service to creators and actively
scans the internet to identify external websites hosting large We interviewed 43 OnlyFans creators with and without prior
amounts of stolen content from its creators [74], there re- sex work experience to understand their online risks on- and
main opportunities for OnlyFans to proactively prevent unau- off-platform. Creators were most concerned about platform
thorized content capture. For instance, OnlyFans could im- risks of shadowbanning and deplatforming from platform
plement features that mimic the behavior of streaming web- censorship and precarity which affected their creativity and
sites like Netflix, which detect screen recording and cause the emotional and financial safety. Despite facing a substantial
screen to turn black. Additionally, leveraging existing digital amount of online hate and harassment, they demonstrated
rights management approaches, such as content fingerprint- preparedness to handle toxic content by proactively setting
ing [59,98], could help automate the detection of unauthorized strong boundaries with fans and blocking abusive fans. Cre-
content use. Such advancements would also streamline the ators found intimate content leaks prevalent and unavoidable
process of issuing automated takedown notices [41]. and employed a range of protective measures from limiting
Catfish prevention. Prior work has described various types prominence to keeping their online identities separate. Our
of catfishing attacks. For example, romance scams involve study emphasizes the significance of platform risks which are
an attacker using stolen photographs and false identities to concerning for sexual, non-sexual, and marginalized creators.
develop a fake romantic relationship with their victim to de- By incorporating platform-specific risks, we can better under-
fraud them of large sums of money [50, 102]. Other attacks stand the challenges faced by at-risk populations and enhance
involve selling photos and videos with sexual content of an- security measures to promote safer online environments.
other person to third parties, sometimes while pretending to
be the subject [51].
Preventative mechanisms from this area of research could
be translated to help OnlyFans creators cope with catfishing Acknowledgements
(§ 5.3): 1) automatic detection and blocking of impersonators
through a hashed image dataset shared across social media and We thank the anonymous reviewers as well as our major re-
dating sites; 2) verification processes where creators could vision shepherd for their valuable feedback. We extend our
apply their verified status on OnlyFans to other platforms, sincere gratitude to the participants of this study. This work
or adding a layer of security through physical assets (one- was supported in part by the Max Planck Institute for Soft-
time-passwords on cellular devices); 3) active and regular ware Systems, the U.S. National Science Foundation awards
reporting of catfishers (by verified accounts and OnlyFans CNS-2206950 and 1663651, and a Google Security & Privacy
automatically generating reports to platforms like Twitter, Research Award for Research into at-risk users and creator
Facebook, or Grindr, requesting investigation into reported safety. A portion of this work was completed while the last
catfish or scam accounts); and 4) asking creators to link their author was at the Max Planck Institute for Software Systems.
Author Contributions [10] Corey H Basch, Lorie Donelle, Joseph Fera, and
Christie Jaime. Deconstructing TikTok videos on men-
VH, AM, and EMR developed the interview protocol and col- tal health: cross-sectional, descriptive content analysis.
lected the interview data. AS and VH analyzed the interview JMIR formative research, 6(5):e38340, 2022.
data. AS, VH, AD, AM, and EMR wrote the paper.
[11] Bianca Beebe. “shut up and take my money!”: Revenue
chokepoints, platform governance, and sex workers’
References financial exclusion. International Journal of Gender,
Sexuality and Law, 2(1):140–170, 2022.
[1] Kendra Albert, Emily Armbruster, Elizabeth Brundige,
Elizabeth Denning, Kimberly Kim, Lorelei Lee, Lind- [12] Elizabeth Bernstein. Temporarily yours: Intimacy, au-
sey Ruff, Korica Simon, and Yueyu Yang. Fosta in thenticity, and the commerce of sex. University of
legal context. CoLUM. Hum. RTs. L. REV., 52:1084, Chicago Press, 2010.
2020.
[13] Rasika Bhalerao and Damon McCoy. An Analysis of
[2] What Does ALGOSPEAK Mean? https://www.cy Terms of Service and Official Policies with Respect to
berdefinitions.com/definitions/ALGOSPEAK. Sex Work. In 2022 IEEE International Symposium on
(Accessed on 06/05/2023). Technology and Society (ISTAS), volume 1, pages 1–14.
[3] Eihal Alowaisheq, Peng Wang, Sumayah Alrwais, Xi- IEEE, 2022.
aojing Liao, XiaoFeng Wang, Tasneem Alowaisheq,
[14] Marta Biino. How to make money on OnlyFans, ac-
Xianghang Mi, Siyuan Tang, and Baojun Liu. Cracking
cording to creators. https://www.businessinside
the wall of confinement: Understanding and analyzing
r.in/tech/news/how-to-make-money-on-onlyf
malicious domain. In Proceedings of the 28th Network
ans-according-to-creators/articleshow/89
and Distributed System Security Symposium (NDSS),
727848.cms, March 2023. (Accessed on 05/30/2023).
2019.

[4] Carolina Are and Pam Briggs. The emotional [15] Danielle Blunt, Stefanie Duguay, Tarleton Gillespie,
and financial impact of de-platforming on cre- Sinnamon Love, and Clarissa Smith. Deplatforming
ators at the margins. Social Media+ Society, sex: A roundtable conversation. Porn Studies, 8(4):420–
9(1):20563051231155103, 2023. 438, 2021.

[5] Carolina Are and Susanna Paasonen. Sex in the shad- [16] Danielle Blunt and Zahra Stardust. Automating Whore-
ows of celebrity. Porn Studies, 8(4):411–419, 2021. phobia: sex, technology and the violence of deplat-
forming: An interview with Hacking//Hustling. Porn
[6] Y. Aydinalp. Struggle for Power and Representation: Studies, 8(4):350–366, 2021.
The Impact of Digital Platform Structures on Sex Work-
ers’ Agency. Master’s thesis, Frei University Brussels, [17] Danielle Blunt and Ariel Wolf. Erased: The impact of
2022. FOSTA-SESTA and the removal of Backpage on sex
workers. Anti-trafficking review, (14):117–121, 2020.
[7] Hanna Barakat and Elissa M Redmiles. Community
under surveillance: Impacts of marginalization on an [18] Ross Bonifacio and Donghee Yvette Wohn. Digital
online labor forum. In Proceedings of the International patronage platforms. In Conference companion publi-
AAAI Conference on Web and Social Media, volume 16, cation of the 2020 on computer supported cooperative
pages 12–21, 2022. work and social computing, pages 221–226, 2020.

[8] Eloise Barry. Why OnlyFans Suddenly Reversed its [19] David R Brake. Are we all online content creators now?
Decision to Ban Sexual Content. https://time.c Web 2.0 and digital divides. Journal of Computer-
om/6092947/onlyfans-sexual-content-ban/, Mediated Communication, 19(3):591–609, 2014.
August 2021. (Accessed on 05/16/2023).
[20] Virginia Braun and Victoria Clarke. Using thematic
[9] Catherine Barwulor, Allison McDonald, Eszter Har- analysis in psychology. Qualitative research in psy-
gittai, and Elissa M Redmiles. “Disadvantaged in the chology, 3(2):77–101, 2006.
American-dominated internet”: Sex, Work, and Tech-
nology. In Proceedings of the 2021 CHI Conference [21] Nadeem Brown. Dirty Little Secret: What No One Ever
on Human Factors in Computing Systems, pages 1–16, Tells You About Internet Adult Entertainment Industry.
2021. Iuniverse, 2005.
[22] Vicky Bungay and Adrian Guta. Strategies and chal- social norm formation and enforcement. Proceed-
lenges in preventing violence against Canadian in- ings of the ACM on Human-Computer Interaction,
door sex workers. American journal of public health, 3(GROUP):1–24, 2019.
108(3):393–398, 2018.
[34] Amy Ford. “Id be Mortified if Someone Thought
[23] Rosie Campbell, Teela Sanders, Jane Scoular, Jane I was Putting my Bum Online for Some Fucking
Pitcher, and Stewart Cunningham. Risking safety Gucci Sandals”: Liminality, Ownership and Identity–A
and rights: online sex work, crimes and ‘blended Case-study of Sex Work on OnlyFans. DEARCADH,
safety repertoires’. The British journal of sociology, page 27, 2022.
70(4):1539–1560, 2019.
[35] X (formerly Twitter). Non-consensual nudity policy.
[24] OnlyFans Help Center. Chargebacks / Credit Card https://help.twitter.com/en/rules-and-pol
Disputes. https://onlyfans.com/help/2/12/45. icies/intimate-media, December 2021. (Accessed
(Accessed on 06/05/2023). on 12/12/2023).
[25] Juliet Corbin and Anselm L. Strauss. Basics of Qual-
[36] Gabriella Garcia. [SHIFT+CTRL+END+DELETE]
itative Research: Techniques and Procedures for De-
Encoded Stigma. PhD thesis, New York University,
veloping Grounded Theory. SAGE Publications, USA,
2020.
fourth edition, 2014.

[26] David Craig. 27. Creator management in the social [37] Art. 20 GDPR – Right to data portability. https:
media entertainment industry. In Making Media, pages //gdpr-info.eu/art-20-gdpr/. (Accessed on
363–374. Amsterdam University Press, Amsterdam, 06/05/2023).
Netherlands, 2019.
[38] Deyan Georgiev. 18 Shocking OnlyFans Statistics to
[27] Scott Cunningham and Todd D Kendall. Prostitution Show How Big It Is [2023]. https://techjury.net
2.0: The changing face of sex work. Journal of Urban /blog/onlyfans-statistics/#gref, March 2023.
Economics, 69(3):273–287, 2011. (Accessed on 04/19/2023).

[28] Anthony Cuthbertson. OnlyFans leak: Huge file of [39] Steven M Gillespie, Andrew Jones, Kasia Uzieblo,
stolen porn dumped online. https://www.independ Carlo Garofalo, and Eric Robinson. Coping using sex
ent.co.uk/tech/onlyfans-leak-porn-hack-v during the coronavirus disease 2019 (covid-19) out-
ideos-photos-file-a9365406.html, February break in the united kingdom. The Journal of Sexual
2020. (Accessed on 06/05/2023). Medicine, 18(1):50–62, 2021.

[29] Ronald C Dodge Jr, Curtis Carver, and Aaron J Fergu- [40] Google. Report Phishing Page. https://safebrowsi
son. Phishing for user security awareness. computers ng.google.com/safebrowsing/report_phish/.
& security, 26(1):73–80, 2007. (Accessed on 05/30/2023).
[30] Gwyn Easterbrook-Smith. OnlyFans as gig-economy [41] Joanne E Gray and Nicolas P Suzor. Playing with
work: A nexus of precarity and stigma. Porn studies, machines: Using machine learning to understand au-
10(3):252–267, 2023. tomated copyright enforcement at scale. Big Data &
[31] Asia A Eaton and Clare McGlynn. The psychology Society, 7(1):2053951720919963, 2020.
of nonconsensual porn: Understanding and addressing
[42] Andrew Griffin. As OnlyFans bans ‘sexually-explicit
a growing form of sexual violence. Policy Insights
content’, what else is actually on there? https://ww
from the Behavioral and Brain Sciences, 7(2):190–197,
w.independent.co.uk/tech/onlyfans-porn-b
2020.
an-other-content-b1905849.html, August 2021.
[32] Monica Eaton-Cardone. Chargeback Challenges On (Accessed on 05/16/2023).
The Rise For Twitch Streamers. https://www.forb
es.com/sites/forbestechcouncil/2020/02/2 [43] Oliver L Haimson, Daniel Delmonaco, Peipei Nie, and
8/chargeback-challenges-on-the-rise-for-t Andrea Wegner. Disproportionate removals and differ-
witch-streamers/, February 2020. (Accessed on ing content moderation experiences for conservative,
06/05/2023). transgender, and black social media users: Marginal-
ization and moderation gray areas. Proceedings of the
[33] Casey Fiesler and Amy S Bruckman. Creativity, copy- ACM on Human-Computer Interaction, 5(CSCW2):1–
right, and close-knit communities: a case study of 35, 2021.
[44] Lee Hair, Ross Bonifacio, and Donghee Yvette Wohn. [56] Susanne Kopf. “Rewarding good creators”: Corpo-
Multi-platform practices among digital patronage cre- rate social media discourse on monetization schemes
ators. Convergence, 28(5):1438–1456, 2022. for content creators. Social Media+ Society,
6(4):2056305120969877, 2020.
[45] Vaughn Hamilton, Hanna Barakat, and Elissa M Red-
miles. Risk, resilience and reward: Impacts of shift- [57] Sophie Kuebler-Wachendorff, Robert Luzsa, Johann
ing to digital sex work. Proceedings of the ACM on Kranz, Stefan Mager, Emmanuel Syrmoudis, Susanne
Human-Computer Interaction, 6(CSCW2):1–37, 2022. Mayr, and Jens Grossklags. The right to data porta-
bility: Conception, status quo, and future directions.
[46] Vaughn Hamilton, Gabriel Kaptchuk, Allison McDon-
Informatik Spektrum, 44:264–272, 2021.
ald, and Elissa M Redmiles. Safer digital intimacy for
sex workers and beyond: A technical research agenda. [58] J Richard Landis and Gary G Koch. The measurement
IEEE Security & Privacy, 2023. of observer agreement for categorical data. biometrics,
[47] Vaughn Hamilton, Ananta Soneji, Allison Mcdonald, 33(1):159–174, 1977.
and Elissa M Redmiles. “Nudes? Shouldn’t I charge [59] Sunil Lee and Chang D Yoo. Robust video finger-
for these?”: Motivations of New Sexual Content Cre- printing for content-based video identification. IEEE
ators on OnlyFans. In Proceedings of the 2023 CHI Transactions on Circuits and Systems for Video Tech-
Conference on Human Factors in Computing Systems, nology, 18(7):983–988, 2008.
pages 1–14, 2023.
[60] Carol Leigh. Inventing sex work. In Whores and other
[48] Eszter Hargittai and Yuli Patrick Hsieh. Succinct sur- feminists, pages 225–231. Routledge, 2013.
vey measures of web-use skills. Social Science Com-
puter Review, 30(1):95–107, 2012. [61] Ada Lerner, Helen Yuxun He, Anna Kawakami, Sil-
via Catherine Zeamer, and Roberto Hoyle. Privacy and
[49] Nicola Henry, Clare McGlynn, Asher Flynn, Kelly
activism in the transgender community. In Proceed-
Johnson, Anastasia Powell, and Adrian J Scott. Image-
ings of the 2020 CHI Conference on Human Factors
based sexual abuse: A study on the causes and con-
in Computing Systems, pages 1–13, 2020.
sequences of non-consensual nude or sexual imagery.
Routledge, 2020. [62] Bin Liang, Miaoqiang Su, Wei You, Wenchang Shi,
and Gang Yang. Cracking classifiers for evasion: A
[50] JingMin Huang, Gianluca Stringhini, and Peng Yong.
case study on the google’s phishing pages filter. In
Quit playing games with my heart: Understanding on-
Proceedings of the 25th International Conference on
line dating scams. In Detection of Intrusions and Mal-
World Wide Web, pages 345–356, 2016.
ware, and Vulnerability Assessment: 12th International
Conference, DIMVA 2015, Milan, Italy, July 9-10, 2015, [63] Jessa Lingel and Danah Boyd. “Keep it secret, keep
Proceedings 12, pages 216–236. Springer, 2015. it safe”: Information poverty, information norms, and
[51] Alice Hutchings and Sergio Pastrana. Understanding stigma. Journal of the American Society for Informa-
ewhoring. In 2019 IEEE European Symposium on Se- tion Science and Technology, 64(5):981–991, 2013.
curity and Privacy (EuroS&P), pages 201–214. IEEE, [64] Sonia Livingstone. Taking risky opportunities in youth-
2019. ful content creation: teenagers’ use of social network-
[52] Shagun Jhaver, Quan Ze Chen, Detlef Knauss, and ing sites for intimacy, privacy and self-expression. New
Amy X Zhang. Designing word filter tools for creator- media & society, 10(3):393–411, 2008.
led comment moderation. In Proceedings of the 2022 [65] Janne Tapani Matikainen. Motivations for content
CHI Conference on Human Factors in Computing Sys- generation in social media. Participations: Journal of
tems, pages 1–21, 2022. Audience and Reception Studies, 2015.
[53] Angela Jones. Sex work in a digital era. Sociology
[66] Samuel Mayworm, Michael Ann DeVito, Dan Del-
Compass, 9(7):558–570, 2015.
monaco, Hibby Thach, and Oliver L Haimson. Content
[54] Angela Jones. Camming: Money, power, and pleasure moderation folk theories and perceptions of platform
in the sex work industry. NYU Press, 2020. spirit among marginalized social media users. ACM
Transactions on Social Computing, 2023.
[55] Daniel Klug, Ella Steen, and Kathryn Yurechko. How
Algorithm Awareness Impacts Algospeak Use on Tik- [67] Allison McDonald, Catherine Barwulor, Michelle L
Tok. In Companion Proceedings of the ACM Web Mazurek, Florian Schaub, and Elissa M Redmiles. “It’s
Conference 2023, pages 234–237, 2023. stressful having all these phones”: Investigating Sex
 Workers’ Safety Goals, Risks, and Practices Online. [80] Nithya Sambasivan, Amna Batool, Nova Ahmed, Tara
In 30th USENIX Security Symposium, pages 375–392. Matthews, Kurt Thomas, Laura Sanely Gaytán-Lugo,
USENIX, 2021. David Nemer, Elie Bursztein, Elizabeth Churchill, and
Sunny Consolvo. “They Don’t Leave Us Alone Any-
[68] Clare McGlynn and Erika Rackley. Image-based sex- where We Go” Gender and Digital Abuse in South Asia.
ual abuse. Oxford Journal of Legal Studies, 37(3):534– In proceedings of the 2019 CHI Conference on Human
561, 2017. Factors in Computing Systems, pages 1–14, 2019.
[69] Jill Nagle. Whores and Other Feminists. Routledge, [81] Patrawat Samermit, Anna Turner, Patrick Gage Kelley,
New York, 2013. Tara Matthews, Vanessia Wu, Sunny Consolvo, and
Kurt Thomas. “Millions of people are watching you”:
[70] Bonnie A Nardi, Diane J Schiano, Michelle Gumbrecht,
Understanding the digital-safety needs and practices
and Luke Swartz. Why we blog. Communications of
of creators. In 32nd USENIX Security Symposium
the ACM, 47(12):41–46, 2004.
(USENIX Security 23), pages 5629–5645, 2023.
[71] OnlyFans About. https://onlyfans.com/about. [82] Teela Sanders, Jane Scoular, Rosie Campbell, Jane
(Accessed on 06/05/2023). Pitcher, and Stewart Cunningham. Internet sex work:
[72] OnlyFans Terms of Service. https://onlyfans.c Beyond the gaze. Springer, 2018.
om/terms. (Accessed on 11/30/2023). [83] Morgan Klaus Scheuerman, Jialun Aaron Jiang, Casey
[73] Bahiyah Omar and Wang Dequan. Watch, share or Fiesler, and Jed R Brubaker. A framework of severity
create: The influence of personality traits and user mo- for harmful content online. Proceedings of the ACM on
tivation on TikTok mobile video usage. International Human-Computer Interaction, 5(CSCW2):1–33, 2021.
Journal of Interactive Mobile Technologies (iJIM), [84] Sexual Rights Database. https://sexualrigh
14(04):121–137, 2020. tsdatabase.org/page/welcome. (Accessed on
06/05/2023).
[74] OnlyFans. Helping Creators Protect Their Copyright.
https://onlyfans.com/transparency-center/ [85] Aria Slippert. Platform Pimps: The Puppet Masters of
helping. (Accessed on 01/12/2024). the Porn-opticon. The iJournal: Student Journal of the
Faculty of Information, 7(2):43–46, 2022.
[75] Helen M Rand. Challenging the invisibility of sex work
in digital labour politics. Feminist Review, 123(1):40– [86] Zahra Stardust, Danielle Blunt, Gabriella Garcia,
55, 2019. Lorelei Lee, Kate D’Adamo, and Rachel Kuo. High
Risk Hustling: Payment Processors Sexual Proxies and
[76] Elissa M Redmiles, Sean Kross, and Michelle L Discrimination by Design. City University of New York
Mazurek. Where is the digital divide? a survey of Law Review, 26(1):57–138, 2023.
security, privacy, and socioeconomics. In Proceedings
of the 2017 CHI Conference on Human Factors in [87] Angelika Strohmayer, Jenn Clamen, and Mary Laing.
Computing Systems, pages 931–936, 2017. Technologies for social justice: Lessons from sex work-
ers on the front lines. In Proceedings of the 2019 CHI
[77] Adi Robertson. OnlyFans CEO on why it banned conference on human factors in computing systems,
adult content: ‘the short answer is banks’. https: pages 1–14, 2019.
//www.theverge.com/2021/8/24/22639356/on
lyfans-ceo-tim-stokely-sexually-explici [88] Angelika Strohmayer, Janis Lena Meissner, Alexander
t-content-ban-banks, August 2021. (Accessed on Wilson, Sarah Charlton, and Laura McIntyre. " we
12/12/2023). come together as one... and hope for solidarity to live
on" on designing technologies for activism and the
[78] Valeria Rubattu, Alicja Perdion, and Belinda Brooks- commemoration of lost lives. In Proceedings of the
Gordon. ‘Cam Girls and Adult Performers Are En- 2020 ACM Designing Interactive Systems Conference,
joying a Boom in Business’: The Reportage on the pages 87–100, 2020.
Pandemic Impact on Virtual Sex Work. Social Sci-
ences, 12(2):62, 2023. [89] Emmanuel Syrmoudis, Stefan Mager, Sophie Kuebler-
Wachendorff, Paul Pizzinini, Jens Grossklags, and Jo-
[79] Jorunn Birgitte Sætre. Performing sexuality, mone- hann Kranz. Data portability between Online services:
tizing intimacy and branding the self–an analysis of an empirical analysis on the effectiveness of GDPR Art.
onlyfans content creators and their interactions with 20. Proceedings on Privacy Enhancing Technologies,
subscribers. Master’s thesis, 2023. 2021(3):351–372, 2021.
[90] Kurt Thomas, Devdatta Akhawe, Michael Bailey, Dan [99] Noel Warford, Tara Matthews, Kaitlyn Yang, Omer
Boneh, Elie Bursztein, Sunny Consolvo, Nicola Dell, Akgul, Sunny Consolvo, Patrick Gage Kelley, Nathan
Zakir Durumeric, Patrick Gage Kelley, Deepak Kumar, Malkin, Michelle L Mazurek, Manya Sleeper, and Kurt
et al. SoK: Hate, Harassment, and the Changing Land- Thomas. Sok: A framework for unifying at-risk user
scape of Online Abuse. In 2021 IEEE Symposium research. In 2022 IEEE Symposium on Security and
on Security and Privacy (SP), pages 247–267. IEEE, Privacy (SP), pages 2344–2360. IEEE, 2022.
2021.
[100] Miranda Wei, Sunny Consolvo, Patrick Gage Kel-
[91] Kurt Thomas, Patrick Gage Kelley, Sunny Consolvo, ley, Tadayoshi Kohno, Franziska Roesner, and Kurt
Patrawat Samermit, and Elie Bursztein. “It’s common Thomas. “There’s so much responsibility on users right
and a part of being a content creator”: Understanding now:” Expert Advice for Staying Safer From Hate and
How Creators Experience and Cope with Hate and Harassment. In Proceedings of the 2023 CHI Confer-
Harassment Online. In Proceedings of the 2022 CHI ence on Human Factors in Computing Systems, pages
Conference on Human Factors in Computing Systems, 1–17, 2023.
pages 1–15, 2022. [101] Colin Whittaker, Brian Ryner, and Marria Nazif.
Large-scale automatic classification of phishing pages.
[92] Katrin Tiidenberg. Sex, power and platform gover-
Google Research, 2010.
nance. Porn Studies, 8(4):381–393, 2021.
[102] Monica T Whitty and Tom Buchanan. The online
[93] Maria Törhönen, Max Sjöblom, Lobna Hassan, and romance scam: A serious cybercrime. CyberPsychol-
Juho Hamari. Fame and fortune, or just fun? A study on ogy, Behavior, and Social Networking, 15(3):181–183,
why people create content on video platforms. Internet 2012.
Research, 30(1):165–190, 2019.
[103] Guang Xiang, Jason Hong, Carolyn P Rose, and Lor-
[94] Natasha Tusikov. Censoring sex: Payment platforms’ rie Cranor. Cantina+ a feature-rich machine learning
regulation of sexual expression. In Media and law: framework for detecting phishing web sites. ACM
Between free speech and censorship, volume 26, pages Transactions on Information and System Security (TIS-
63–79. Emerald Publishing Limited, 2021. SEC), 14(2):1–28, 2011.

[95] Jirassaya Uttarapong, Ross Bonifacio, Rae Jereza, and [104] Yue Zhang, Jason I Hong, and Lorrie F Cranor. Cantina:
Donghee Yvette Wohn. Social support in digital pa- a content-based approach to detecting phishing web
tronage: OnlyFans adult content creators as an online sites. In Proceedings of the 16th international confer-
community. In CHI Conference on Human Factors ence on World Wide Web, pages 639–648, 2007.
in Computing Systems Extended Abstracts, pages 1–7, [105] Matthew A Zook. Underground globalization: Map-
2022. ping the space of flows of the internet adult industry.
Environment and Planning A, 35(7):1261–1286, 2003.
[96] Jirassaya Uttarapong, Jie Cai, and Donghee Yvette
Wohn. Harassment experiences of women and LGBTQ
live streamers and how they handled negativity. In Appendix
ACM international conference on interactive media
experiences, pages 7–19, 2021. A Study Materials:
[97] Jan Pieter Van Oudenhoven, Boele de Raad, Francoise The interview protocol for this study and the relevant recruit-
Askevis-Leherpeux, Pawel Boski, Geir Scott Brunborg, ment materials can be accessed from https://osf.io/zbh
Carmen Carmona, Dick Barelds, Charles T Hill, Boris 7c/?view_only=657f69cccc17447f9be89cacaadd73f0.
Mlačić, Frosso Motti, et al. Terms of abuse as expres-
sion and reinforcement of cultures. International Jour-
nal of Intercultural Relations, 32(2):174–185, 2008.

[98] Sviatoslav Voloshynovskiy, Farzad Farhadzadeh, Olek-

siy Koval, and Taras Holotyak. Active content finger-
printing: a marriage of digital watermarking and con-
tent fingerprinting. In 2012 IEEE International Work-
shop on Information Forensics and Security (WIFS),
pages 175–180. IEEE, 2012.