A deepfake is using my face. How should I respond?

Describe the false claim before debating the technology

A manipulated video may make you appear to recommend an investment or say something you never said. Start with the specific claim, the identity being used and the action viewers are encouraged to take. The use of AI alone does not explain the entire incident or automatically establish that every edited image is unlawful.

Separate what you know from what you suspect. You can document a false endorsement or impersonation without first identifying the model that produced the footage. Avoid presenting a guess about the production method as a confirmed finding. This distinction makes both the initial report and subsequent public communication more credible.

Capture the publication and the relevant sequence

Record the video URL, publishing profile, observation time and timestamps of the disputed statements. For an advertisement, include the ad identifier where available and the destination domain. Preserve the caption and offer as well as the clip: they may explain why the apparent endorsement is misleading.

Retain original files when lawfully available without bypassing access controls. Do not install an unknown downloader. If you have authentic source footage, note which passages have been altered. Keep originals separate from annotated or redacted sharing copies. Where legal action is contemplated, obtain advice about formal evidence preservation rather than relying solely on a cropped excerpt.

Do not treat an AI detector as a verdict

Odd lip movement, inconsistent lighting or an unusual voice may justify further investigation. They do not independently settle authenticity. An automated detector's score should likewise not be treated as certain proof. The false attribution, misleading offer and lack of the claimed relationship can be important regardless of the exact editing technique.

Do not upload private recordings or identity documents to an unfamiliar analysis service. If expert examination is necessary, agree the scope, access and retention arrangements. Technical authenticity analysis and immediate distribution reporting can proceed alongside one another, but they answer different questions and should have clearly assigned owners.

Map and report separate distribution points

List the video, advertisement, account, destination website and search result individually. Choose the relevant platform route for the facts involved. Meta reporting options and Google advertising reports concern different services; removing one object does not automatically remove every linked copy or campaign.

Removing a Google search result does not delete the source website. Nor is there a universal Google form that guarantees removal of every AI-generated endorsement. Procedures have specific eligibility criteria. For example, a route intended for intimate content is not appropriate for an ordinary commercial deepfake. Keep confirmations and check any appeal options provided with the decision.

Warn people without amplifying the clip

State clearly which offer you do not endorse and where your genuine channels can be found. Explain confirmed manipulation briefly, without reposting the complete video unnecessarily. A clear statement can help customers verify information without giving the fabricated footage additional distribution or making copies harder to track.

Prepare a consistent response for partners and customer support. Ask for links to new copies, not confrontations with suspicious accounts. People who have transferred money or disclosed banking information need prompt contact with their bank and appropriate authorities. Reporting the video is not a substitute for protecting funds or compromised accounts.

Follow recurring versions

Group related copies by distinctive statements, promoted products, visual sequences or domains, while preserving individual URLs and statuses. One removed video does not demonstrate that the campaign has ended. Cropping, new captions or another advertising account may make the next version look different while repeating the same misleading message.

Assign responsibility for reviewing audience reports and deciding the next step. State which public sources are checked and which areas remain inaccessible, such as private messages. The absence of a new alert should mean only that nothing was found in the agreed checks, not that no misuse exists anywhere.

Agree responsibilities before handing over the case

Evidence collection, reporting preparation, platform correspondence and repeated checks are operational work. Identifying an offender, pursuing compensation or conducting proceedings may require a lawyer or competent authority. Keep these roles explicit rather than treating a technical service as a substitute for every part of the response.

Agree who approves statements, what authority the representative has and how sensitive documents are shared. Never hand over an account password as a shortcut. A workable process combines clear authorisation with a concise case file and recorded decisions, so that speed does not come at the expense of privacy or accuracy.

Frequently asked questions

Must I identify the AI model before reporting?

No. You can begin documenting the false attribution or suspected fraud using verified facts. Specialist analysis may become relevant later.

Can every copy be removed permanently?

No provider can guarantee that outcome. A service can define searches, documentation, reporting and follow-up, while removal decisions remain with the relevant platforms or authorities.

Sources and reporting routes

Reporting guidance checked on 13 October 2026. Platform options and requirements may change.

This guide covers technical and organisational steps, not individual legal advice. Rights, claims and formal evidence requirements should be assessed by an appropriate legal professional.