Real IG Followers

Instagram Will Flag Copies of Your Reels. It Won't Decide Who Owns Them

Nov 19, 2026  ·  27 min read
Instagram Will Flag Copies of Your Reels. It Won't Decide Who Owns Them

TL;DR

  • On November 17, Instagram opened Content Protection to eligible creators. It scans Instagram and Facebook for visual matches of your original videos and lets you track a copy, hide it, add an "original by" credit, or release the claim.
  • A separate, small test will look for AI videos that reuse a creator's spoken words while swapping the face and body, the trick behind the AI "comedians" that lifted stand-up sets this fall.
  • The tool finds copies. It does not decide who owns them, punish the copier, or move any money, and the first account to enroll can end up holding the claim. Treat it as a lost-and-found desk, not a court.

On October 3, the California comedian George Brett Olson posted two videos side by side. On one side was Olson, doing a bit about artificial intelligence threatening jobs. On the other was "Tiffany Sloane," an AI-generated woman with a large and fast-growing Instagram following, doing the same bit, with matching staging and nearly the same words. Only the punchline had been changed. Olson later called it "one of the most ironic things that's ever happened," which is a comedian's way of saying a machine had stolen his joke about machines stealing work.

Sloane was not a one-off. WIRED identified eight comedians whose videos had been replicated by scantily clad AI-generated women, many of whom followed each other. Sloane's account, according to WIRED and to Kotaku's reporting as summarized by the New York Post, passed 375,000 followers within weeks of its first posts in September, and its profile pointed to an OnlyFans page. Another persona, "Aylen Skye," performed a routine by the Los Angeles comic Julia Jasunas about renting a room to an attractive man, down to the verbal filler, according to the same reporting. AFP reported that Aylen Skye gathered more than 230,000 followers and a Meta verified checkmark before the account was removed.

When Jasunas reported the account, Instagram rejected her impersonation complaint, the New York Post reported, because Aylen Skye was not pretending to be her. That is the whole problem in one sentence. The copy did not steal her face. It stole her material and wore a new face on top of it, and none of the boxes in the reporting menu fit.

Six weeks later, Instagram answered. On November 17, Adam Mosseri announced that Content Protection, a copy-detection tool Meta launched on Facebook a year earlier, is now open to eligible Instagram creators, along with a limited test that tries to catch exactly what happened to Olson and Jasunas: AI videos that keep a creator's words and replace everything else. Most coverage has read this as Instagram finally standing up for original creators, and in part it is. What fewer people have asked is what kind of protection it actually is. Content Protection is very good at one job, finding copies, and deliberately silent on the three questions creators care about most: who owns the original, what happens to the thief, and where the money goes. This piece walks through how the system works, why its design choices matter more than its detection rate, and what the law, YouTube and Meta's own history say about where it is heading.

What Instagram actually switched on

Start with the mechanism, because it is more modest than the headlines suggest. According to Instagram's description, as reported by Digital Trends, TechWyse and Social Media Today, the tool "automatically looks out for visual matches of your original videos" across Instagram and Facebook. Once a creator enrolls, every new original video is protected automatically. Older videos are not swept in; eligible creators pick them one by one.

When the system finds a match, the creator gets an alert that shows the copy's view count and the follower count of the account that posted it, TechWyse reported. Then the creator chooses one of four responses:

  • Track, the default: the copy stays up and the creator watches how it performs.
  • Hide: the copy is blocked from view on both Instagram and Facebook.
  • Credit: Instagram adds an "original by" label that names the original creator and links back to the video or profile.
  • Release: the creator drops the claim and the copy stays up with no further monitoring.

There is also an allow list, so a second account, a collaborator or a brand partner can reuse a video without being flagged. Eligibility depends on meeting Instagram's "originality and integrity standards." Qualifying creators see a notification in their activity feed, and others can apply. Instagram has not published how many accounts qualify, what the thresholds are, or how accurate the matching is. TechWyse listed those gaps plainly, and none of the coverage we reviewed found numbers that fill them.

The engine underneath is not new. When Meta launched the Facebook version on November 17, 2025, TechCrunch and Tubefilter both reported that it uses the same matching technology as Rights Manager, the system Meta has offered large copyright holders for years. What changed in 2025 was the packaging: a mobile tool for ordinary creators instead of an enterprise dashboard. What changed this week is that a creator who posts only to Instagram no longer has to cross-post to Facebook to be covered. A year ago, TechCrunch noted, Instagram videos were protected only if the creator also pushed them to Facebook, a requirement the outlet suggested could nudge more creators into posting there.

Here is an image worth keeping for the rest of this piece. Content Protection works like the lost-and-found desk at a big train station. It is good at telling you where your umbrella turned up. It can put the umbrella behind the counter, or tie a tag to it with your name. What it does not do is decide whose umbrella it is when two people claim it, call the police on whoever walked off with it, or pay you for the days someone else used it in the rain.

Content Protection answers "where is my video?" It was never built to answer "whose video is this?"

The joke without the face

The visual matcher has a blind spot that the AI comedians found almost immediately. Matching works by comparing what a video looks like. An account that re-performs a stand-up set with an AI-generated body, a new room and a new outfit produces a video that looks nothing like the original. The part that was copied, the writing, is invisible to a system that compares pictures.

That is why the second half of the announcement matters more than the first. Instagram says it will test a system that flags videos reusing the spoken words of a protected video even when the visuals are fully replaced. Social Media Today described it as transcript detection. According to TechWyse and The Hype Magazine, which cited TechCrunch, the test starts with a small group of creators "in the coming weeks" and is meant to expand later to everyone enrolled in Content Protection on Instagram and Facebook. A Meta spokesperson told The Hollywood Reporter, as quoted by The A.V. Club and Chortle, that it is a "limited test, distinct from the broader protection rollout." Instagram also says it is exploring ways to limit the reach of accounts that repeatedly use AI tools to reproduce other people's work, with no threshold and no date.

Mosseri did not oversell it. "We will not get everything right at first, but we'll keep improving as fast as we can," he wrote, in a line quoted by both The A.V. Club and Chortle. He had been more candid a month earlier. Replying on Threads to an artist whose work had been copied by AI, he wrote, according to Social Media Today: "To be clear, I am very worried about this trend." He added that the company was "actively working out what to do," that existing copyright tools were not enough because many comedians and dancers had never copyrighted their material, and that "We need a way to help those creatives without letting bad actors make more fake claims (happens all the time unfortunately)."

Moving from pictures to words is a real step, and it is aimed at the right target. It is also a step into much harder territory, because words are where originality gets genuinely contested. Two comics can arrive at the same premise independently. A creator can quote a famous line on purpose. A whole genre of Reels is built on people speaking the same trending audio. A picture match asks a fairly narrow question: is this the same footage? A script match has to ask a much broader one: is this the same work? Platforms have historically struggled with the second question, and Instagram has not said how its test will draw the line.

A face can be swapped in seconds. The writing is the part that was always worth stealing.

The turn: a registry nobody voted for

The common framing of this story is copyright enforcement finally arriving on Instagram. That framing is understandable, and it is wrong in a way that matters for anyone who plans to rely on the tool.

Look at what the four buttons do. None of them removes a copier's account. On the Facebook version, TechCrunch reported in 2025, blocking a reel reduced its distribution, but the account that reposted it faced no disciplinary action; Meta reportedly wanted to prevent the tool from being used to target specific accounts. TechWyse reported the same design on Instagram: hiding reduces distribution "without directly penalizing the publishing account." None of the buttons moves money either. Every option Instagram describes is about visibility: show the copy, hide it, label it, or leave it alone. Not one of them sends the copy's earnings, if it has any, to the person who made the original. That stands out in a year when Meta has been busy adding ways for creators to earn, from long-form Features deals to Amazon affiliate links in Reels.

That last point is a choice, not a technical limit, and Meta's own history proves it. In 2017, TechCrunch reported that Facebook had added a "claim ad earnings" option to Rights Manager, so a rights owner could collect ad revenue from other people's uploads of their video. The same underlying matching technology now powers Content Protection. The money button exists in the building. It simply is not on the creator's panel.

Then look at ownership. TechWyse reported that creators can file an ownership dispute through Content Protection when they believe another account has claimed their work, and named the common case directly: another account protecting someone's work before the original creator enrolls. The outlet also made the point that most coverage skipped: detecting similar content does not establish who owns it or whether a use was authorized.

Put those pieces together and the turn comes into view. Content Protection is not a copyright system. It is a private registry of who got there first, run on Meta's eligibility rules, with remedies that cost Meta very little. Hiding a copy or tagging it with a credit changes what people see. It does not change who gets paid, and it does not end the business model of the account that copied. Jasunas understood the logic before the tool existed. She told Chortle she started posting her stand-up online to create a timestamp that attaches her to her work. The lost-and-found desk, it turns out, trusts the oldest ticket.

There is a fair defense of this design, and it deserves a hearing. Mosseri's own worry was fake claims, and he is right that they are common. A tool that let any enrolled account delete other accounts with one tap would become a weapon within weeks. Hiding instead of punishing, and requiring an eligibility bar, are reasonable ways to limit the damage a false claim can do. The trouble is that the same caution that protects the accused also caps what the tool can do for the wronged.

The tool changes what viewers see. It does not change who gets paid.

What US law already says about jokes and dances

You do not need to register to own it. The US Copyright Office says a work "is under copyright protection the moment it is created and fixed in a tangible form," and that registration is generally voluntary. A recorded set posted to Instagram is fixed. Registration becomes necessary when you want to sue over a US work.

Jokes can be protected, but proving it is costly. In 2015, the comedy writer Robert Alexander Kaseberg sued Conan O'Brien's production company, alleging that five of his jokes had been lifted. A federal judge let claims over three of them move toward trial before the case settled in May 2019, weeks before trial, on undisclosed terms. O'Brien denied taking anything. The case showed that jokes can be protected, and how expensive it is to find out.

Most social dances are not protected. The Copyright Office's guidance on choreography says "social dances, simple routines, and other uncopyrightable movements cannot be registered" as separate works. Composed choreography is a different matter: in November 2023, a federal appeals court revived choreographer Kyle Hanagami's suit over a Fortnite emote, ruling that a short excerpt of a dance is not unprotectable just because it is brief. The case was resolved in 2024.

Ideas are free, expression is not. A premise can be shared by anyone. The specific words and performance are what the law protects, which is why a word-for-word AI re-performance is a far stronger case than a similar idea.

Mosseri's copyright line, examined

The sidebar matters because of one sentence in Mosseri's October post. He said existing copyright claims were not enough because many comedians and dancers had not copyrighted their material. That framing has been repeated without much pushback, and it deserves some.

In the United States, a comedian who records a set and posts it already holds a copyright in that recording and, in most cases, in the written material, whether or not they ever file a form, according to the Copyright Office's own guidance. The gap Mosseri describes is real, but it is not a paperwork gap. It is a cost gap and a category gap. Olson consulted an intellectual property lawyer who, according to WIRED and the New York Post, told him legal action probably was not worth it unless the account was earning millions. Jasunas's complaint failed because Instagram's reporting menu asked whether she had been impersonated, and she had not. Her writing had been taken, and Instagram's menu had no clean box for that.

Dancers are a fairer case. Under the Copyright Office's guidance, most social dance steps and simple routines are not protectable at all, and a huge share of dance on Instagram falls in exactly that category. For those creators, no amount of registration would help, and a platform tool is the only remedy available. For comedians, the situation is almost the opposite. The law already covers the words. What has been missing is a cheap, fast way to enforce it, and a platform willing to treat stolen writing as a problem in its own right.

Seen that way, the script test is less a gift than a correction. It points Instagram's enforcement at the thing the law already recognized and the old reporting menu ignored.

The same company sells the scissors

There is a second thing nobody seems to be questioning, and it sits right next to the first. Meta is not a neutral bystander to AI re-skinning. It builds and promotes tools that make it easy.

In June 2025, Meta added Restyle to its Edits app and the Meta AI app: preset prompts that change a person's outfit, location or visual style in short videos, PetaPixel and Social Media Today reported. In October 2025, a version arrived in Instagram Stories, where users can type a prompt to add, remove or change elements of a photo or video, according to TechCrunch and Gulf News. In September 2025, Meta launched Vibes, a feed of AI-generated videos inside the Meta AI app where users can remix clips from the feed and cross-post the results to Instagram and Facebook, according to Meta's own announcement and TechCrunch. And this month Instagram launched Effect Builder, a tool for making AI effects from text prompts.

None of these tools is designed to steal anyone's work, and most of what people make with them is harmless. But together they describe a company that wants as much AI video on its platforms as possible, and has said so. In his October Q&A, Mosseri said Instagram does not limit the reach of AI content simply for being AI, Social Media Today reported: "We do not throttle the reach of AI content just for being AI content." Tubefilter, covering the Facebook launch a year ago, pointed out the irony directly: Meta's push into generative AI had made it easier for bad actors to imitate popular content cheaply.

That context changes how to read the "exploring ways to limit reach" line. Instagram is promising to slow down a behavior that its own product strategy speeds up. That is not hypocrisy so much as a tension Meta will have to manage indefinitely, and creators should expect the line between "remix" and "rip-off" to be drawn in a place that keeps plenty of AI video flowing.

The lock and the lock pick are coming out of the same factory.

What YouTube learned in 18 years of matching

Instagram is not the first platform to face this, and the comparison is instructive. YouTube's Content ID, which Tubefilter described in 2025 as an 18-year-old fingerprinting system, processed about 2.5 billion copyright claims in 2025, up 14% from about 2.2 billion the year before, according to YouTube's copyright transparency report as tallied by TorrentFreak. The most striking number is what rights holders chose to do with those claims: they monetized more than 90% of them rather than taking the videos down. Uploaders disputed about 0.51% of claims.

Read that carefully, because it says something Instagram's design does not. When rights holders are given the choice, they overwhelmingly prefer to get paid for a copy rather than make it disappear. Content ID turned copying from pure theft into a revenue stream for the owner. Instagram's tool, as described, offers no equivalent, so its most valuable outcome for a creator is the credit link, which may send some viewers back to the original. Nobody has published data on how much traffic an "original by" label actually delivers, so for now that value is a reasonable hope, not a measured effect.

The comparison has limits, and they cut both ways. Content ID serves a relatively small group: TorrentFreak counted 4,454 rights holders actively using it in 2025, mostly large companies, while ordinary creators use simpler tools. In that sense, Meta's tool is more democratic, since it is built for everyday creators from the start. YouTube has also found that newer kinds of detection catch less than people hoped. Its likeness detection, which launched for YouTube Partner Program creators on October 21, 2025, according to MediaPost and Business Today, looks for AI copies of a creator's face. When YouTube expanded it in March 2026, TechCrunch reported that the tool had reached roughly 4 million creators and that removals so far had been "very small." Building a detector, in other words, is not the same as getting copies off the platform.

YouTube's experience also carries a warning about abuse. TorrentFreak reported that more than 6% of webform removal requests in 2025 were judged by YouTube's review team to be a likely false assertion of copyright ownership. Any system that hands out power to claim content will attract people who claim content that is not theirs.

The fake-claim economy is already here

That warning is not hypothetical on Instagram. In September, the BBC reported on scammers filing fake copyright complaints against Instagram accounts and demanding payment to withdraw them, a scheme covered by Malwarebytes, PetaPixel and Bitdefender. According to that reporting, Meta automates much of its copyright processing and does not verify complaints when they are filed, and repeated complaints can lead to temporary takedowns. The owner of one history account paid $50 in cryptocurrency to get his account back and, per the BBC's account as summarized by Malwarebytes, was hit with more claims immediately afterward. Meta told the BBC it fights deceptive behavior meant to scam people, and in a separate case acknowledged in court that 13 copyright strike notices against one Instagram user had been fraudulent.

This is the "bad actors make more fake claims" problem Mosseri named, and it explains the shape of Content Protection better than any press release. A system that cannot verify who owns what tends to fall back on proxies: an eligibility bar, an originality score, a first-to-enroll record, and remedies mild enough that a mistake does limited harm. Those proxies are sensible. They also mean that the creators best placed to benefit are the ones Meta already trusts, while a newer comic whose clip goes viral on someone else's account may find the copier enrolled first.

Jasunas described a version of that inversion to Chortle in October. She said some comedians had been falsely accused of plagiarism and had accounts removed even though they were the original creators, and that she had struggled to get verified as a working comedian because she performs under a stage name, while a fake AI account was verified easily. Her account is a single interview, but it describes exactly the failure a first-come registry can produce.

A registry that trusts the earliest claim works only if the real creator gets there first.

A two-year campaign for originality

Content Protection is the newest piece of a longer effort, and the earlier pieces show which way Instagram is leaning. In April 2024, Instagram said it would start replacing reposted content with the original in recommendations, and that accounts which repeatedly post unoriginal content would lose eligibility for recommendations, as Engadget and Instagram's creators blog described at the time. On April 30, 2026, Instagram extended that rule from Reels to photos and carousels. According to Instagram's own post, accounts that primarily post unoriginal photos, carousels or reels are no longer shown in recommendation surfaces, and they can become eligible again when most of their recent posts in a 30-day period count as original. Followers still see them. The rule sits alongside the other Instagram algorithm changes this year.

At the same time, Instagram has made reuse easier, not harder. In August 2025 it launched a native repost button, which credits the original poster and can surface reposts in friends' and followers' feeds, according to Meta's announcement and coverage by ABC7. The message across all of this is consistent: Instagram does not mind your work traveling, as long as Instagram controls the route and the label. Content Protection extends that logic to copies made outside the repost button. The copy is allowed to exist; the platform just wants the right name attached and the original favored when it ranks, in line with how Instagram's recommendation system already claims to work.

That approach is coherent, and for many creators it will be better than the old choice between filing a report, complaining in public and giving up. It is also, again, a visibility answer to what is often a money problem. The AI comedians were not after credit. They were using borrowed jokes as free advertising for paid adult subscriptions, according to WIRED, AFP and Chortle. A credit label on a clip does not touch the subscription page it points to.

The numbers, and why they are thin

It is worth being precise about what is known, because the figures in this story come from different sources counting different things.

On the copying: WIRED identified eight comedians whose videos were replicated by AI personas. Tiffany Sloane passed 375,000 followers within weeks, per WIRED and the New York Post citing Kotaku, and one of the account's clips drew more than 7 million views in under two weeks, the Post reported. Aylen Skye reached more than 230,000 followers by AFP's count and about 250,000 by the New York Post's. The comic Olivia Carter found her joke about asking out a personal trainer, which she had performed on television, repurposed by AI accounts; the Post, citing a Straight Arrow News investigation, reported that two of those accounts had about 180,000 and 240,000 followers, against Carter's roughly 113,000 at the time. Follower counts on accounts like these swing quickly and vary between outlets, so treat them as snapshots.

On enforcement: Meta said in July 2025, as TechCrunch reported, that it had removed about 10 million profiles impersonating large creators and taken action against 500,000 accounts for spammy behavior or fake engagement. Those are Meta's own figures, from the period before the Facebook tool launched, and they measure impersonation and spam, not copied work.

What nobody has published is the figure that would settle the argument: how often Content Protection catches a copy, how many creators are eligible, how often creators choose hide over credit, and whether a credit label sends meaningful traffic back. Until Meta releases numbers like those, any claim that the tool "solves" copying, or that it does nothing, runs ahead of the evidence. If you track your own Instagram Insights after enrolling, you will know more about your own case than any public dataset currently shows.

The most important number in this story is the one Instagram has not released.

What we know, and what we don't

We know the tool is live for eligible creators, scans Instagram and Facebook for visual matches, offers four responses, includes an allow list and an ownership dispute route, and does not penalize the account that posted a hidden copy. We know the script test is small, is described by Meta as separate from the main rollout, and has no announced date for wider release.

We do not know the eligibility thresholds, the detection accuracy, or how the script test will treat quotes, shared premises, trending audio or deliberate parody. Coverage also differs slightly on how the test works: some outlets describe it as matching spoken words or a script, and one described it as matching audio. We do not know whether the reach limits for repeat AI copiers will ever ship. And while Meta has offered ad-earnings claims to large rights holders through Rights Manager since 2017, it has not said whether anything similar will come to Content Protection.

One more honest limit. The examples here are mostly comedians, because the AI comedian story is where the reporting is deepest. Dancers, educators, artists and anyone whose value lives in a recognizable script face versions of the same problem, but the public record on them is thinner.

A practical playbook for creators

  1. Enroll as soon as you are eligible, and add your best older videos by hand. New originals are protected automatically, but the back catalog is not. Since another account protecting your work first is the common dispute case, the order of enrollment matters. Start with the videos that already travel, typically your strongest Reels.
  2. Set up the allow list before you need it. Add your second accounts, regular collaborators and brand partners. A sponsor reposting your video should never land in your alerts, and a partner should never find a branded clip hidden by mistake.
  3. Choose between credit and hide on purpose. Hiding protects the original from competition; a credit label keeps the copy's reach and points it back to you. For a large copy account with an audience you would like to reach, credit may be worth more. For a copy that misrepresents you or feeds a scam, hiding is the cleaner choice. This is a judgment call, since nobody has published data on how much traffic a credit sends.
  4. Keep your own dated record. Raw files, earlier drafts, set lists and the date you first performed a bit are your evidence in an ownership dispute and, if it ever comes to that, in court. The platform's record is a useful timestamp, not the only one you should hold.
  5. Know where the law already helps. In the US, recorded material is protected from the moment it is fixed. If a piece of writing is central to your income, registering it with the Copyright Office is what makes a lawsuit possible later. For dance, most social routines are not protectable, so the platform tool is your main remedy.
  6. Treat any copyright demand outside the app as a scam until proven otherwise. Check complaints inside Instagram, never through links, and never pay to have a claim withdrawn. The BBC's reporting showed that payment can simply invite the next claim.
  7. Build what a copy cannot take. A clone can borrow your words, but not your relationship with the people who already follow you. A real audience that knows your name, your captions and where to find you elsewhere is the protection that no matching system can strip away.

The lost-and-found desk

It is tempting to judge Content Protection by whether it stops the thieves. On that measure it will disappoint, because it was never built to stop them. It was built to find copies, label them, and keep Instagram out of the business of deciding who owns what. Meta has every reason to want that role: it keeps the platform flowing, limits the damage from false claims, and avoids turning a creator tool into a payments system with legal weight.

Once you see it as a lost-and-found desk, the right expectations follow. The desk will tell you where your work turned up, more often than before and across more of Meta's apps. It will let you put your name on it. It will not refund the views a copy collected, and it will trust whoever holds the oldest ticket. So the job for creators shifts. Hoping the platform will protect your work is not a strategy; being unmistakably the first and most findable source of it is. That means enrolling early, keeping your own receipts, and building an audience that already knows the original voice when a borrowed face shows up saying the same words.

The AI comedians exposed something the industry had avoided for years: on a platform built on views, the face on screen was always easier to protect than the writing behind it. Instagram has now started listening for the words. Whether it ever pays for them is the question to keep asking, in the same places where it already counts what a view is worth and decides what Your Algorithm shows to whom.

FAQ

What is Instagram Content Protection?

A tool for eligible creators that scans Instagram and Facebook for visual matches of their original videos. When it finds one, the creator can track the copy, hide it on both apps, add an "original by" credit link, or release the claim. New original videos are covered automatically after enrollment, and older ones can be added by hand.

Does it catch AI videos that copy my jokes or script?

Not yet for most creators. The main tool matches visuals, so a re-performance with a new AI face and setting can slip past it. Instagram is testing a separate system that looks for reused spoken words with a small group of creators, with plans to expand it to everyone enrolled. No date or accuracy figures have been announced.

Will hiding a copy get the copier banned or pay me?

Based on what Meta has described, no. Hiding reduces the copy's distribution without directly penalizing the account that posted it, and none of the four options sends the copy's earnings to the original creator. For serious or repeated theft, the copyright reporting route and, in some cases, legal action remain separate options.

Do I need to register my work to own the copyright?

In the US, no. The Copyright Office says protection exists from the moment a work is created and fixed in a tangible form. Registration is required to bring an infringement lawsuit over a US work. Simple social dance routines, however, are generally not protected by copyright at all.