A Deepfake Is Not ‘Just a Fake Picture’: Synthetic Violence as a Tool for Silencing Women in the Media

At 7.12 on a Monday morning, a journalist opens a message from a colleague. The re is no greeting, only a warning: somebody has circulated a sexualised synthetic image bearing her face. By 7.20, three editors are discussing whether she should come off air ‘until things settle down’. At 7.35, a manager asks whether publishing a denial would amplify the material. Nobody has yet asked what she wants, whether she is safe, or who will take over the practical work of reporting, preserving evidence and contacting platforms.

This is a hypothetical scene. It is also a useful correction to the way deepfakes are commonly discussed. The standard conversation begins with detection: is the picture real or fake? For the person targeted, that may be only one small part of the problem. A sexualised deepfake can be recognised as fabricated and still function as violence. It can humiliate, invite harassment, supply new material for abuse, damage professional relationships and make the victim’s name searchable alongside an imposed sexual narrative.

Calling it ‘fake content’ describes the file. Calling it synthetic violence describes the act, the power relationship and the human cost.

The false picture and the real event

Disinformation frameworks are valuable when an altered image is used to deceive a public about an event. They are insufficient when the primary purpose is to punish a woman for being visible. In such cases, falsity is not necessarily a defect in the weapon. Perpetrators may know the image is fake. Recipients may know it too. The point can be to force the target to spend time denying, reporting and emotionally surviving something she never did.

Technology-facilitated sexual violence includes harms committed, enabled or amplified through digital systems. Research in this field emphasises that online and offline consequences cannot be neatly separated: image-based abuse can affect safety, employment, relationships and psychological wellbeing (Henry & Powell, 2018, pp. 195–208). Generative systems alter the cost and scale of production, but they do not invent the underlying entitlement to discipline women through sexual humiliation.

Emma A. Jane’s work on online misogyny is helpful here. Abuse is not merely a collection of unpleasant messages. It is a patterned communicative practice that marks who is supposedly entitled to speak and who must pay an extra toll for doing so (Jane, 2017). A synthetic image can become part of that toll. The journalist’s face, taken from the public work that establishes her authority, is repurposed to attack the same authority.

The result is a double bind. Visibility is necessary for the job, particularly for presenters, reporters and independent creators. The same visibility supplies material and distribution pathways for attack. Telling a targeted woman to become less visible may reduce one immediate exposure while completing the perpetrator’s political work.

What the 2026 UN Women study tells us

The UN Women report Tipping Point is unusually important because it measures not only exposure but consequences and attempts at redress. Its

2025 survey was distributed in five languages through trusted networks and analysed responses from 641 women in public life across 119 countries. Twelve per cent reported non-consensual sharing of personal images and 6% reported being targeted by deepfakes or manipulated imagery. Twenty-four per cent said they had been diagnosed with or treated for anxiety or depression linked to online violence; 13% reported post-traumatic stress disorder; and 41% said they self-censored on social media (Posetti et al., 2026, pp. 6–7).

The professional breakdown matters. Among women journalists and media workers who answered the relevant questions, 9.1% reported non-consensual sharing of personal images, 4.5% deepfake or manipulated-image abuse, and 23.8% unwanted sexual advances in private messages (Posetti et al., 2026, pp. 8–9). Almost one quarter reported treatment for anxiety or depression connected with online violence and 12.8% reported treatment for PTSD (pp. 10–11). Of 351 women journalists and media workers answering the self- censorship question, 45.3% said they self-censored on social media and 21.7% at work (pp. 11–12).

These figures require disciplined interpretation. The survey used purposive and snowball sampling, not a probability sample, and its authors explicitly state that the results are not generalisable. Questions were optional for ethical reasons, so denominators vary. Comparisons with the 2020 survey involve different samples, although the report tested the change in self-censorship and found it statistically significant within the samples used (Posetti et al., 2026, p. 16). The numbers identify serious patterns among respondents; they do not provide a global prevalence rate for all women journalists.

Methodological caution should sharpen the story, not erase it. Behind the percentages are careers and nervous systems. One Indian environmental journalist described the cumulative effect of attacks and intimidation in five exact words: ‘we are forced into silence’ (Posetti et al., 2026, p. 15).

The chilling effect is the product

Elisabeth Noelle-Neumann’s spiral of silence theory concerns how perceptions of opinion climate and fear of social isolation can inhibit public

expression (Noelle-Neumann, 1974, pp. 43–51). It should not be mechanically applied to every case of gendered abuse. It does, however, offer a useful frame for the chilling mechanism: speaking is not simply an internal choice; it has a perceived social cost.

Synthetic sexual violence raises that cost deliberately. A reporter may avoid a subject, decline a live appearance or close a public account not because her evidence has changed, but because the expected punishment has. Colleagues watching the attack also learn. The message travels beyond one target: this is what can be done to women who investigate, criticise or remain publicly visible.

The newsroom can accidentally finish the attack

Institutions often respond with protective instincts: remove the journalist from air, take her off the sensitive story, make her accounts private, ask her not to comment. Some temporary changes may be requested by the targeted person or justified by an immediate safety assessment. The danger lies in making disappearance the default remedy.

The UN Women report found that among women who reported online violence to police, 27% experienced reluctance or refusal to investigate and 24% perceived victim-blaming treatment. The same proportion said they were made to feel responsible for preventing further harm by leaving social media, speaking less publicly, moving to less visible work or taking career leave (Posetti et al., 2026, p. 15). A response that protects the institution’s discomfort by shrinking the survivor’s life reproduces this logic.

The Polish NASK guide likewise identifies deepfakes used for reputational attacks, intimidation and sexualised depictions, including attacks on journalists. It recommends coordinated procedures involving editorial, legal and technical functions, reporting channels, verification and support for the person affected (Adamczyk et al., 2025, pp. 44–48, 55–57). The key word is coordinated. The journalist should not become incident commander, evidence archivist, platform liaison, public spokesperson and therapist to her colleagues while continuing her normal workload.

Nor should the newsroom republish the material merely to prove that it exists. Reproduction can increase discoverability, expose the survivor to repeated viewing and supply abusers with renewed attention. Reporting on the attack is sometimes in the public interest; showing the abusive artefact is a separate decision requiring necessity, consent and harm assessment.

Secure – Agree – Limit – Escalate – Return

The following protocol is designed for newsrooms, universities and creator organisations. It avoids one-size-fits-all instructions because legal routes, platform processes and security risks differ. For high-risk cases, involve qualified local professionals.

First assess immediate physical, digital and psychological safety. Ask whether the material includes threats, location information, family members, impersonation or signs of coordinated escalation. Assign one trained contact so the survivor does not have to retell the incident to every department.

Preserve the minimum evidence needed for reporting, legal advice and platform action without creating unnecessary copies. Record where and when the material appeared, relevant links and the chain of decisions. Store access in a restricted, approved location. Do not ask the survivor to keep the abusive material on her personal phone or repeatedly open it for colleagues.

‘Secure’ also means securing workload. Remove the incident-management burden, not automatically the journalist’s public role. Offer practical cover, protected time and, where appropriate, independent security assessment.

Agree a survivor-led plan. What outcome matters now: immediate removal, evidence preservation, public correction, no public statement, police advice, legal consultation, family safety, continued broadcasting? Preferences may change as facts develop, so consent is a process rather than a signature collected at 8.00 on day one.

Explain limits honestly. A newsroom cannot promise that every copy will disappear or that a platform or police force will act. It can promise who will do what, by when, and how decisions will be reviewed. Record permissions for internal sharing and public communication. Silence from a shocked person is not consent to a press strategy.

Limit circulation to what is necessary. Do not post the abusive image in an all- staff chat, embed it in a debunk or use it as a thumbnail. Share descriptive summaries where possible. Brief colleagues not to search, download or forward the material out of curiosity.

Limit public amplification too. A correction can state that synthetic sexualised material is false without reproducing it. Where visual evidence is genuinely necessary for public-interest reporting, use the least exposing form, obtain informed agreement from the survivor and reassess whether the evidential value outweighs renewed harm.

Limit internal speculation. The newsroom does not need a seminar on what the image ‘looks like’. It needs verified facts, assigned responsibilities and respect for privacy.

Run parallel, coordinated pathways rather than sending the survivor from desk to desk. The response may involve platform reporting, specialist digital- security support, legal advice, law enforcement where appropriate and desired, union or professional-association support, and trauma-informed mental-health care. One case lead keeps the map and communicates progress.

Set escalation thresholds in advance: credible physical threats, doxxing, contact with relatives, rapid cross-platform spread, impersonation of newsroom accounts, extortion, or acute distress require more senior and specialist response. Do not make access to care conditional on filing a police report. The UN Women findings show that formal reporting itself can expose survivors to reluctance and blame (Posetti et al., 2026, pp. 14–15).

Public statements should name the act and the institution’s support without turning the journalist into evidence on display. If she does not want to become the face of a campaign about deepfakes, that choice deserves respect.

Return means returning control, not returning to a fictional ‘normal’. Schedule check-ins after roughly 24 hours, one week and one month, adjusted to the person’s wishes and the incident. Review safety, workload, search visibility, platform responses, legal developments and wellbeing. Attacks can recur after public attention moves on.

Support continued professional participation. If duties were changed temporarily, agree criteria and a date for review. Do not let a precaution become a silent demotion. Offer options rather than assuming the survivor wants either immediate visibility or permanent withdrawal.

Finally, return to the system. Conduct a learning review that examines response times, unnecessary circulation, decision authority, staff conduct and gaps in support. The purpose is institutional improvement, not grading the survivor’s reaction.

The incident dossier

A concise dossier makes the protocol auditable. Record the survivor’s nominated contact, immediate risks, current preferences and sharing permissions, the minimum evidence register, actions with owners and deadlines, public-communication decisions, review dates, temporary duty changes and the criteria for reversing them. Restrict access and justify retention: the dossier must not become a second source of exposure. Senior leaders can review response times and recurring gaps without receiving the abusive material.

What this protocol cannot promise

No newsroom can guarantee complete removal from a networked environment. Laws and enforcement vary, and preserving legal evidence can

conflict with the desire to erase. A public rebuttal may reduce deception while increasing attention. The protocol cannot substitute for regulation, competent law enforcement, safer platform design or affordable care. Nor should one non-probability survey be treated as a universal prevalence estimate. Most importantly, survivors want different outcomes. Survivor-led practice means making decisions with the person who bears the most intimate cost, not merely about her.

Do not confuse disappearance with safety

A deepfake can be false in every pixel and true in every consequence. It can change a career, a family conversation, a search result and the willingness to investigate the next powerful subject. Treating it as a technical curiosity misses the social mechanism. Treating it as synthetic violence clarifies what protection is for.

The measure of a newsroom response is not how quickly the image vanishes from the managers’ screens. It is whether the journalist regains safety, choice and the ability to continue participating in public life. Secure the evidence, agree the plan, limit the spread, escalate support and return after the headlines have moved on. Above all, do not remove a woman’s voice in order to prove that you take an attempt to silence her seriously.