When a fake video, image or audio clip of you or your company starts to spread, the first mistake is usually speed. People post a denial before they know what they are dealing with.
A detection tool or a gut feeling gives you a suspicion. A verified finding needs evidence you can show to someone else. Keep the two apart in every message you write.
Step 1: Preserve before you react
Save the original post, the account that posted it, the date and time, the caption and the link. Take full-page captures. Content gets edited or deleted, and you may need the original later.
Step 2: Separate what you know from what you suspect
- Known: where it appeared, who posted it, when, and how it is being shared.
- Suspected: whether it was generated or altered, and by what method.
Detection results are indicators. They do not prove how a file was made, and no tool is right every time. Write “appears to be manipulated” until someone qualified has reviewed the evidence.
What preserving evidence means in practice
Step 1 says to save before you react. Government guidance on online abuse explains why. Australia’s eSafety Commissioner says to resist the urge to delete messages or comments straight away, because if you later report the content to the platform, the regulator or the police, you will need evidence of where and when it was sent, posted or shared. It also warns that if the other person’s account is blocked, the evidence can disappear.
The eSafety list of what to capture is a useful base for any business: the service or platform, the web address of the content, the usernames and profile addresses of the accounts involved, the times and dates, and anything else that may matter. It also suggests recording how often you were targeted, since one post may look harmless alone but part of a pattern when viewed together. On a phone, where the address bar is hidden, it suggests using the share icon and choosing “copy link.”
We add one more practice for teams: save in one shared place and name one person as the keeper of the record. When three people each keep a partial folder, nobody can later say what was seen first.
Only where lawful and safe
There is a limit that matters more than speed. eSafety’s guidance says not to save nude or sexual images or videos of anyone under 18, because that is usually illegal, and not to take screenshots or photos of illegal and restricted content. In that situation the content is reported with its location (the webpage address and account details) and not copied. The rule is written for Australia, but the principle travels: if the material itself may be unlawful to hold, record where it is and do not make copies, and talk to a lawyer or the police before you do anything else. If anyone is in immediate danger, or a threat or a minor is involved, contact the authorities first. Reputation steps can wait.
Why we say “appears to be manipulated”
Step 2 recommends careful wording until a qualified person has reviewed the file. That caution has a technical basis. The Content Authenticity Initiative explains that Content Credentials, the open C2PA standard for provenance, use signed metadata that cannot be altered without leaving evidence of alteration, but it also says metadata of any kind can be removed deliberately or accidentally, and that invisible watermarks can be removed or spoofed. In other words, the absence of a credential does not prove a file is fake, and the presence of one does not make it true in every respect. Provenance is one signal. It is not a verdict.
For a business, that means your first statement should describe what you can show: “this clip did not come from us and we did not approve it,” not “this is an AI deepfake made by X.” The first can be stood behind today. The second may be wrong.
Step 3: Match the response to the situation
Different content calls for different routes: a copyright claim if your own work was copied, an impersonation report if someone is posing as you or your business, a privacy report for private or intimate content, a brand or trademark report for fake endorsements, and legal counsel for false statements of fact. Platforms decide for themselves, and outcomes are not guaranteed.
Step 4: Keep watching
Copies and reposts often appear after the first removal. Keep a simple log of links and dates so any follow-up request is specific.
Step 5: Tell people what you can stand behind
A short, factual statement works best: what the content is, that you did not create or approve it, and where the real information is. Avoid speculation about who is responsible.
Matching the route: three examples of what the sources say
Step 3 lists several routes. Here is what the primary sources say about three of them, and where each stops.
Intimate images and digital forgeries (United States). The TAKE IT DOWN Act, now codified at 47 U.S.C. § 223a, required covered platforms to set up a notice and removal process by May 19, 2026. According to the statute, a valid request is in writing and includes a signature of the identifiable individual or an authorized person, enough information to locate the depiction, a brief statement of a good faith belief that it was published without consent, and contact information. On receiving a valid request, the platform must, as soon as possible and no later than 48 hours afterward, remove the depiction and make reasonable efforts to remove known identical copies. The law covers a defined category, intimate visual depictions, on defined platforms: those serving the public that primarily host user-generated content, or that regularly host such depictions. It excludes email and broadband providers. It does not cover every fake, and Inside Privacy’s June 16, 2026 summary notes that the FTC enforces it and has launched a complaint site for platforms that fail to comply.
Impersonating a business (United States). The FTC’s Trade Regulation Rule on Impersonation of Government and Businesses, published March 1, 2024 and effective April 1, 2024, prohibits impersonating governments, businesses and their officials or agents in interstate commerce. In its statement accompanying the rule, the FTC said it revised the text to track the FTC Act’s language on unfair or deceptive practices, with materiality and interstate commerce requirements, so it should not reach parody or non-commercial speech. It also said it was still considering rules on impersonating individuals and sought comment through a supplemental notice. So a fake that poses as your business is a stronger match for this rule than a fake that targets a named person. Whether a given case qualifies is a legal question for counsel, and the FTC, not a private business, enforces the rule.
Your face on YouTube. YouTube’s help page describes a likeness detection tool for eligible creators, and it sets conditions: the creator must be over 18, be a channel owner or manager, and complete verification with a government-issued ID and a short selfie video. The tool runs a one-time search of newly uploaded videos for the enrolled person’s face and lets the creator review matches and request removal through the privacy complaint process. It is limited. It currently matches faces of enrolled creators, not voices (YouTube says it aims to extend this), and it may also surface genuine footage of the person, which cannot be removed under the privacy policy, though a copyright request may apply if the use is unauthorized. This is a tool for one platform, and enrollment takes up to five days according to the page.
The common thread is that each route has its own conditions, and the best route depends on what the content is, where it sits and who is affected. That is why Step 3 begins by asking what kind of content you are dealing with.
A short response plan your team can fill in today
The closing action in this article asks you to write down where your name, face or brand appears online and who would own a response. Here is a fuller version of that exercise. It takes about an hour and does not need any tool.
- List the assets. Your business name, logo, key people, voices and the pages where they appear.
- Name the owner. One person leads the response, and one person keeps the evidence record.
- Write the escalation points. When does a fake become a legal matter? When it contains false statements of fact, intimate content, a threat or a minor, the answer is to involve counsel or the authorities.
- Draft the holding statement. Two sentences: what the content is, that you did not create or approve it, and where the real information lives. Leave the cause blank.
- Set up a log. A simple table with columns for URL, platform, date first seen, date reported, reference number and status.
- Decide who gets told. Staff, key clients and partners may hear about a fake from someone else. Decide in advance who hears it from you.
None of this guarantees a particular outcome. Platforms and courts make the decisions. A prepared team simply loses less time on the first day.
Sources
- eSafety Commissioner (Australia), “How to collect evidence”
- Content Authenticity Initiative, “Durable Content Credentials,” April 8, 2024
- 47 U.S.C. § 223a (TAKE IT DOWN Act, Pub. L. 119-12, § 3)
- Inside Privacy, “The TAKE IT DOWN Act’s Notice and Removal Requirements Enter Into Effect,” June 16, 2026
- Federal Trade Commission, Trade Regulation Rule on Impersonation of Government and Businesses, 89 FR 15017 (March 1, 2024)
- YouTube Help, “Likeness detection on YouTube”
Where we fit
We now offer synthetic content monitoring and removal support as a new service. It covers four areas: platform takedown support, ongoing monitoring, legal evidence preparation and escalation to counsel, and investigation of where a piece of content came from and how it spread. We prepare the evidence and support the request. Platforms and courts make the final decisions, and we cannot promise a result or a timeline.
Take one step today: write down where your name, face or brand appears online, and who on your team would own a response if a fake appeared.