If you are asking “what should i do if someone makes a deepfake of me”, first save evidence without spreading the file, report the original upload to its host, and use the appropriate removal or safety service. A fabricated image or voice can be copied before you finish arguing that it is fake, so the strongest response is a sequence of containment decisions rather than a public debate about authenticity.
Deepfakes are not one legal or technical problem. A counterfeit advert using your face, a cloned voice asking relatives for money, and a fabricated sexual image all require different interventions. The most urgent question is not whether the content looks convincing. It is whether someone is being endangered, coerced, defrauded or exposed, and who can actually stop distribution.
In 2026, the US Federal Trade Commission began enforcing platform duties under the Take It Down Act. Covered services receiving a valid notice about non-consensual intimate imagery must remove it and known identical copies within 48 hours. That provision is powerful but narrow: it is not a worldwide promise that every fake video disappears in two days. Google likewise provides removal tools for Search results, not a universal delete button for every host.
This guide maps the first hour, the next two days and the longer recovery period. It draws on guidance from the FTC, Google, UK police, NCMEC, StopNCII and online-safety authorities. It also separates what an image hash can accomplish from what it cannot. Where rules depend on country, platform or the nature of the content, the distinction is explicit.
The First Hour: Protect Yourself Without Amplifying the File
Capture a Clean Evidence Record
Write down exactly where the fake appeared, when you noticed it, the publishing account, and any messages or threats. Preserve the post URL and a screenshot showing the username and date; capture the page address separately because a screenshot alone can conceal the origin. Store correspondence and takedown acknowledgements in a private folder. Police.uk advises recording links, usernames, user IDs, platform names, dates and what was said or depicted. If an image involves a child, do not download, forward or create new copies of explicit material merely to preserve evidence; use the platform or NCMEC reporting route.
Separate the evidence of publication from the harmful file itself. A private incident log can preserve dates and context without multiplying copies of intimate content. If immediate danger exists, contact local emergency services. If someone threatens to publish the image unless you pay or provide another image, preserve the demand and treat it as coercion, not a commercial negotiation.
Choose a Small Response Team
Ask one trusted person to help with reporting, screenshots of non-explicit context, and tracking replies. Repeatedly viewing disturbing content can be exhausting. If family or colleagues are likely to be targeted by a cloned voice, warn them through a trusted channel without attaching the fake. Avoid quote-posting an abusive upload: a rebuttal with a direct link can provide the attacker with precisely the circulation they wanted.
A useful rule is to create one secure record and several separate reports, not one viral rebuttal. That minimises amplification and makes later complaints easier to prove. recognising a manipulated video explains the limitations of visual verification; victims do not need a commercially produced detector report before filing a content-abuse complaint.
| First-hour action | Keep | Avoid |
| Identify source | URL, account, timestamp | Reposting the file |
| Preserve threats | Message records and case number | Responding with payment |
| Protect contacts | Independent warning by trusted channel | Forwarding the fake |
| Seek help | Platform, trusted adult or police | Uploading child sexual material to detectors |
Identify the Harm Before Choosing a Takedown Route
The term deepfake covers synthetic nude pictures, swapped faces, lip-synchronised videos, voice cloning and complete fabricated recordings. Platforms and laws classify those harms differently. Report the specific violation rather than selecting a generic “false information” category when a more precise impersonation, harassment, extortion or non-consensual intimate imagery option exists.
A Practical Triage Test
Ask four questions: Does it portray intimate or sexual conduct? Was the depicted person under 18 when the material supposedly shows them? Is anyone being threatened or asked for money? Is the file circulating publicly or only in private messages? These answers determine whether to involve child-protection specialists, intimate-image removal systems, fraud teams or law enforcement. If the image is synthetic, say so clearly; the harm does not vanish because the scene never occurred.
For reputation-only fakes, collect the original publication context and request removal based on impersonation or harassment policies. For commercial scams, notify people who may receive fraudulent instructions and report fake accounts or adverts. For suspected child sexual exploitation, use child-safety reporting mechanisms immediately rather than adult-only services.
The privacy dimension is wider than the image: a scammer may combine your face with public biography, address and job details. Our coverage of the broader privacy risks explores how separate fragments can create a convincing synthetic identity. That background matters when deciding which accounts and public details to secure after the fake appears.
| Scenario | Primary route | Additional step |
| Synthetic intimate image of adult | Host report; StopNCII if eligible | Search-result removal |
| Sexual image involving someone under 18 | NCMEC/child protection; platform | Do not circulate or download |
| Fake voice requesting money | Fraud reporting; warn recipients | Verify by a known number |
| Non-sexual impersonation clip | Host impersonation/harassment report | Preserve correction evidence |
| Threat to distribute fake | Police/appropriate emergency assistance | Preserve extortion messages |
Report the Original Host, Not Only the Search Result
Find the Platform’s Specific Abuse Form
Open the original post or account, use the reporting menu, and choose the closest applicable category. For intimate material, look for non-consensual intimate images, sexual deepfakes or intimate image abuse. Identify the person depicted, provide the exact URL and say explicitly that the depiction is fabricated and non-consensual. If the platform offers a dedicated legal notification workflow, use it as well as its ordinary user report where appropriate.
A concise report can say: “This page contains a fabricated image/video depicting me in an intimate or compromising situation without my permission. I request removal of this content and known identical copies. Here is the exact URL; I am the person depicted.” Avoid submitting more personal identity documents than the official form requires. Keep the case number, submission time, any automatic acknowledgement and the eventual response.
Escalate Precisely When the First Report Fails
If the post remains available, use the service’s appeal, legal or trust-and-safety escalation route. Show that a prior report was filed, note the timestamp, and identify any new mirror uploads. Do not assume reporting an account removes every post it has published. In the US, the FTC provides a dedicated route to report covered platforms that fail to comply with the Take It Down Act after a valid notice.
Our earlier reporting on deepfake detection and enterprise defence addresses why enterprise detection systems cannot replace a platform removal process. A detection signal may flag manipulated footage, but it does not itself create a legally sufficient victim notice or remove copies distributed by unrelated services.
Remove Search Results Separately from Hosted Content
A search result is a pointer. Removing that pointer can make abuse much harder to discover, but the image may remain on the source page and in private messages. This is why victims should normally pursue both the hosting platform and relevant search engines. Google’s February 2026 removal workflow for non-consensual explicit images includes an image-result menu: choose “Remove result”, then “It shows a sexual image of me”. The company also describes multi-image requests and optional protections intended to filter future similar explicit results.
When the Image Is Not Sexual
A non-sexual fake alleging criminal behaviour, professional misconduct or a made-up statement may not qualify for Google’s intimate-image pathway. Consider the platform’s defamation, impersonation, privacy or legal request process instead. Eligibility and outcomes depend on the service, jurisdiction and facts. Distinguish a platform removing a hosted post, a search engine delisting a result, and an investigator securing evidence: they are different actions with different objectives.
Be careful with commercial “internet erasure” services promising complete deletion. Ask which URLs they will contact, what legal authority they rely on, whether they access your accounts, what happens if removal fails, and whether the quote includes ongoing monitoring. No provider can credibly guarantee erasure of every screenshot, encrypted message or copy stored on a private device.
Use the Correct Image-Hashing Tool for Intimate Deepfakes
StopNCII for Eligible Adults
StopNCII is a free service operated by the Revenge Porn Helpline/SWGfL. Eligible adults select an intimate file already on their device; the system computes a digital fingerprint locally and distributes the hash, rather than the image, to participating platforms. These services can check for matches and act under their policies. StopNCII says it cannot remove content from the whole internet, only participating platforms. It also requires the depicted person to meet its age and possession criteria.
NCMEC Take It Down for Images Depicting Minors
NCMEC’s Take It Down helps people whose nude or sexually explicit images or videos were taken when they were younger than 18, including people who are now adults. It generates hashes on the user’s device; the images do not leave that device. NCMEC specifically warns people not to send, share or download an image just to submit it. The system is intended for participating public or unencrypted services; it is not an enforcement tool against every private or encrypted location.
Hash matching and ordinary takedown forms are complements. A form identifies a known live URL; a hash can help participating services recognise subsequent copies. Cropping, alteration, conversion and non-participating hosts create limits. Do not mistake a successful case submission for proof that every variation has been removed.
| Service | Best fit | What it does not guarantee |
| StopNCII | Eligible adult intimate imagery | Whole-internet deletion |
| NCMEC Take It Down | Explicit content depicting minors | Removal on non-participating services |
| Google image-removal request | Qualifying Search results | Deletion from original host |
| Platform abuse report | Specific hosted posts/accounts | Deletion of off-platform copies |
Understand Your Legal Options in 2026
United States: The Take It Down Act
According to FTC business guidance, Section 3 of the Take It Down Act took effect on 19 May 2026. Covered platforms must provide a notice-and-removal process and remove non-consensual intimate imagery and known identical copies within 48 hours of a valid request. The obligation is not universal to every website or every offensive deepfake. If an eligible platform has no workable reporting mechanism or fails to respond on time, the FTC provides its TakeItDown.ftc.gov reporting channel. A complaint to the FTC is different from a criminal report about the offender.
United Kingdom: Intimate-Image Offences and Platform Duties
The UK government’s September 2026 Crime and Policing Act factsheet describes measures including duties addressing timely takedown of non-consensual intimate imagery, offences concerning nudification-tool supply, and image deletion orders. The precise commencement and applicability of a provision should be checked against official legislative updates before claiming a specific deadline applies to an individual case. UK police guidance supports reporting threatening or abusive deepfakes and preserving identifying information even where the perpetrator is not known.
European Union and Other Jurisdictions
The EU AI Act imposes particular transparency duties around synthetic content and deepfakes, but disclosure of an AI-generated image is not consent to publish a person’s intimate likeness. Privacy, harassment, copyright and criminal remedies may be separate. Our analysis of European AI transparency rules explains why product-provider and publisher obligations should not be confused. Outside these jurisdictions, local cybercrime, harassment and privacy laws may offer different procedures; a local legal or victim-support adviser can establish which apply.
Legality also differs between making, possessing, publishing, threatening to publish and profiting from a deepfake. A universal statement that “all deepfakes are illegal” is inaccurate. The safer approach is to describe the actual harmful behaviour, the jurisdiction and the applicable reporting mechanism.
When It Is a Voice Clone, Payment Scam or Fake Endorsement
A synthetic voice recording can be an instrument of impersonation without showing any intimate image. Treat demands for wire transfers, gift cards, verification codes or account resets as a security incident. Call the supposed sender on a telephone number you already trust. Do not use the contact details embedded in the suspicious message, and do not rely on voice familiarity alone.
Secure the Account and the Audience
Change compromised passwords, revoke suspicious sessions and enable multifactor authentication if account takeover is suspected. Tell a small group of relevant colleagues or relatives that any unusual request must be independently verified. If the fake is an advert endorsing an investment, supplement or product, report the advertisement and impersonating account, preserving the sponsor identity and ad reference where accessible.
A victim should not have to prove which model made a fake before asking for assistance. The difference between detection, attribution and identity proof is explained in how AI-authorship detection works. What matters operationally is that a false message is circulating, someone is impersonating you and people may act on it.
Why Deepfake Detectors Are Supporting Evidence, Not a Verdict
Detection products inspect signals such as audio cadence, mouth movement, compression artefacts or statistical patterns associated with synthesis. These clues can help an investigator prioritise cases, but a high “AI probability” score rarely establishes who made the file or whether the original underlying scene actually happened. Authentic media can be edited deceptively; synthetic media can also be edited to remove easy-to-spot artefacts.
Prefer Provenance and Independent Corroboration
Search for the earliest credible version, request source files when safe, verify the purported recording location and compare the claim against independently documented events. Provenance credentials may show authenticated editing history when correctly applied, but absence of credentials does not prove fakery, and a credential cannot automatically establish whether a spoken claim is true.
Technical analysis becomes more valuable in a legal dispute when its methods, files, timestamps, and limitations are documented. If the stakes are high, use a qualified forensic professional and preserve the original evidence chain rather than uploading private or explicit files to multiple unvetted detector sites. Our guide to limits of automated content detection explains why a detector score should never be treated as conclusive proof.
Decide Whether to Make a Public Statement
Publicly calling a file fake can be necessary when an impersonation scam is actively misleading customers, an employer is investigating a fabricated statement, or journalists are preparing coverage. Yet a reaction can also transform a small abusive upload into a search trend. Choose the smallest audience that needs accurate information.
A Minimal Correction That Does Not Promote the Fake
For colleagues: “A manipulated recording claiming to feature me is circulating. It is not authentic. Please do not forward it or act on any requests attributed to me; verify directly through our normal channels.” Avoid including the abusive clip, naming an attacker without reliable evidence, or speculating about technology used. If press attention is foreseeable, establish one contact point and keep a written chronology of what you can verify.
In harassment or intimate-image cases, silence in public does not mean inaction. It may be sensible to work privately with platforms, specialist services and legal advisers. A public rebuttal is not a prerequisite for reporting abuse or obtaining support.
Build a 48-Hour Case Log and Follow-Up Plan
The most useful operational improvement is a simple case ledger: each harmful URL or account gets its own entry; each report receives a timestamp and reference; each outcome is checked separately. This avoids the common failure where victims believe a report is complete because one duplicate vanished. It also gives an adviser a clean chronology without requiring repeated descriptions of the incident.
| Field | Example of what to record |
| Item | Post URL or account handle |
| Nature of harm | Intimate deepfake / impersonation / fraud |
| First observed | Date and timezone |
| Report submitted | Platform form and case ID |
| Response deadline | Applicable policy or legal timing, if verified |
| Outcome | Live / removed / appeal submitted |
| Next action | Escalate, notify contact, document copy |
During the first day, prioritise safety, evidence, platform reports and appropriate hash services. In the second day, track acknowledgements and look only for known mirrors rather than endlessly searching graphic content. Where the US Take It Down Act applies, the statutory 48-hour removal clock starts from a valid request received by a covered platform; it should not be confused with an informal support ticket submitted through an unrelated form.
Longer-term monitoring should be proportionate. Set a manageable review interval, ask a trusted helper to handle distressing material, and retain your reporting records securely. Search engines and hash tools reduce exposure, but continuing to inspect every corner of the internet is neither feasible nor required to make a strong case.
What You Should Not Do
Do not pay an extortionist merely because they promise to delete a file. There is no enforceable guarantee that copies will disappear, and payment can invite additional demands. Do not retaliate by posting the suspect’s private information, threaten anyone, or submit the file to an unknown “AI checker” that may retain uploads. Do not assume a synthetic intimate image is harmless because no real photograph was stolen. Do not infer that a fake is authentic because a detector cannot flag it.
Be cautious about advice to request a copyright takedown in every case. Copyright ownership is a technical question: being the person portrayed does not automatically make you the copyright owner of someone else’s video or photograph. Platform harassment, impersonation, privacy and intimate-image routes can be more directly relevant. A solicitor or qualified legal adviser can help if a counter-notice, litigation or disputed ownership is involved.
Finally, avoid inventing certainty. The precise generation model, uploader identity, reach and number of copies may remain unknown. Document what can be established, label hypotheses as such, and update the case when credible new evidence appears.
Our Editorial Verification Process
This guide cross-checked victim reporting steps against the UK Police deepfake guidance and RAINN’s evidence-preservation recommendations; US platform obligations against FTC guidance dated May and September 2026; Google Search removal against its February 2026 product announcement and support form; and hash-service scope against StopNCII and NCMEC official instructions. We also consulted UK government and Ofcom statements on intimate-image regulation. No live testing of a victim’s intimate media, platform case submissions or forensic detector evaluations was carried out for this article.
The comparisons are procedural, not a benchmark of service success rates. Eligibility may depend on user age, jurisdiction, the depiction itself and the platform. Where an exact price is relevant, StopNCII and NCMEC describe their services as free; no commercial detector or paid lawyer is required by this article. The website sitemap endpoints did not return accessible data during research, so relevant internal article URLs were independently confirmed from live indexed pages rather than represented as a complete sitemap inventory.
This article was researched and drafted with AI assistance for review by the Awais Khalid editorial desk at Perplexity AI Magazine. Named editorial review and independent verification must be completed by the publisher before publication; neither is claimed to have occurred through this draft alone.
Conclusion
When someone makes a deepfake of you, the solution is rarely one dramatic counterattack. It is a sequence of narrower actions: protect yourself, document what exists, report the original host, reduce search visibility, limit re-uploads where a suitable hash service exists, and escalate coercion or fraud through the correct channels. Keeping those processes separate makes a distressing incident easier to manage.
The 2026 changes in US law and major-platform reporting tools strengthen the options available to people targeted by fabricated intimate content. They do not create worldwide deletion, guarantee that every variation is detected or settle the legal position of every non-sexual impersonation. The boundaries matter: victims should not be sold a promise that technology and legislation cannot fulfil.
A disciplined case record, clear reporting language and a limited circle of support are often more useful than a viral rebuttal. Over time, better provenance standards and platform prevention may reduce the burden on individuals, but their effectiveness remains uneven. The immediate priority is to keep control of the next action, not to prove every technical detail of the fake.
Frequently Asked Questions
What Should I Do If Someone Makes a Deepfake of Me?
Preserve the URLs, publication details and threats; report the original upload to its host, then request search-result removal where eligible. For intimate content, use StopNCII or NCMEC Take It Down if eligible and involve police where threats, extortion or child abuse are involved.
Can I Get a Deepfake Removed from Google?
Google offers removal requests for qualifying non-consensual explicit depictions, including sexual deepfakes. Approval removes relevant Search results, not necessarily the image from the original hosting website.
Is It Illegal to Make a Deepfake of Someone?
It depends on the country and the behaviour involved. Creating or sharing non-consensual intimate images, threatening distribution, stalking, fraud and harassment may trigger separate rules. A harmless labelled parody is not legally identical to intimate-image abuse.
Can StopNCII Remove Every Copy?
No. StopNCII works by sharing digital fingerprints with participating platforms. It cannot delete files from the entire internet, private devices or every messaging service.
What If I Was Under 18 in the Image?
Use specialist child-protection reporting such as NCMEC’s Take It Down or CyberTipline. Do not download, forward or upload sexual material of minors to unknown services to establish proof.
Should I Pay Someone Threatening to Post a Deepfake?
Do not assume payment will prevent distribution. Preserve the threats, stop direct negotiations where safe, report extortion to the appropriate authorities and seek specialist support.
Can an AI Detector Prove Who Created It?
Usually not. A detector may flag characteristics of synthetic content, but attribution to a specific uploader or model generally requires independent evidence, provenance or a formal investigation.
How Long Does It Take to Remove an Intimate Deepfake?
It depends on the platform and jurisdiction. Under the US Take It Down Act, covered platforms have 48 hours after a valid notice to remove qualifying content and known identical copies. Other processes do not necessarily share that deadline.
References
Federal Trade Commission. (2026). Complying with the Take It Down Act. Official source
Federal Trade Commission. (2026). Was your intimate image shared online without your consent?. Official source
Google. (2026). A simpler way to remove explicit images from Search. Official source
National Center for Missing & Exploited Children. (n.d.). Take It Down. Official source
StopNCII. (n.d.). How StopNCII.org works. Official source
Police.uk. (n.d.). Deepfakes: Reporting it to us. Official source
UK Home Office. (2026). Crime and Policing Act 2026: Intimate image abuse factsheet. Official source
Ofcom. (2026). Platforms should use detection technology to stop spread of illegal intimate images. Official source
RAINN. (2025). How to report tech-enabled sexual abuse. Official source
