Instagram Will Label AI Generated Profiles.

Instagram announced on Monday 31 August that it has renamed its “AI creator” label to “AI-generated profile”, and that profiles featuring an AI-generated person which skip the label will lose their eligibility for recommendations.

That second half is the story. This is not a content-quality rule and it is not a watermarking rule. It is an identity-disclosure rule with a distribution penalty attached, and the line runs at whether the person on screen is real, not at whether AI touched the work.

Most brand social teams can stop worrying after the next two sections. A small number cannot, and they now have a decision to make before enforcement starts.

What Instagram actually changed

Three things, all in the same post on Instagram’s creators blog:

• The label is now called “AI-generated profile” rather than “AI creator”. The old name described someone who uses AI. The new one describes a profile where the person is AI.
• Creators apply it themselves, by editing their profile and toggling on AI-generated profile. It then shows on the profile and alongside the account’s posts, where a post does not already carry its own AI label.
• Instagram will restrict distribution for profiles it detects as featuring an AI-generated person that have not added the label. Those profiles become ineligible for recommendations.

Instagram’s stated reason is blunt. People do not like seeing a profile that seems human, only to find out later that the person featured is AI-generated.

The line is the person, not the process

Instagram is explicit that creators who simply use AI tools as part of their creative process do not need the label.

So an upscaled product shot, a generated background, a caption the model tidied, a template graphic, an auto-cut Reel: none of that triggers anything. Your founder’s face is still your founder’s face.

What triggers it is a profile built around a person that AI generated or substantially created. A synthetic presenter. A virtual spokesperson. A brand character rendered as a human being who talks to camera.

That is a far smaller group than “brands using AI”, and the distinction matters, because platform AI policy has so far been mostly about what happens to AI content once it is posted. This rule is about who the account claims to be.

Recommendation ineligibility is the penalty that actually bites

Instagram did not say it would remove these accounts, hide their posts from existing followers, or cut them out of monetisation. It said they stop being recommended.

On Instagram, that means Reels, Explore and suggested accounts. Which is to say, everything that reaches people who do not already follow you.

For an established brand account, losing recommendations is a slow bleed. For an AI-persona account it is close to terminal, because those accounts exist to be discovered. Being served to strangers is the entire growth model.

Instagram has chosen the one penalty that lands hardest on exactly the accounts the rule targets and barely registers anywhere else. Whatever you think of AI personas, that is a precisely aimed rule.

Where the rollout actually stands

This is the part most coverage skips, and it changes what you should do this week.

• The label toggle has rolled out to all accounts. You can add it today.
• Instagram has not published a date on which detection and restrictions start.
• Applying it early carries no penalty, so there is no reason to wait for one.

So nothing is being demoted under this rule right now. No account has lost recommendations to it. If someone tells you their reach dropped last week because of this, they are guessing.

What that gives you is a window. The gap between a rule being published and a rule being enforced is the only comfortable time to make a decision about it, and it is open now.

Check Account Status before you decide anything

Instagram’s stated process, once detection begins:

1. The account owner gets a notification.
2. Account Status shows whether the profile is no longer eligible for recommendations.
3. The owner either adds the AI-generated profile label or appeals.
4. A successful appeal makes the profile and its content eligible for recommendations again.

Instagram also acknowledged that its systems “might not always be perfect”, which is a polite way of saying some human-fronted accounts will be misread by a classifier. Heavily retouched, filtered or stylised faces are the obvious exposure, and so is a brand that shoots everything against generated backdrops.

If you run any account with a person at the centre of it, put Account Status on a named weekly check. It takes half a minute. It is also precisely the sort of routine that never gets assigned until someone audits the whole setup and asks who owns it.

The disclosure maths are not close

Labelling voluntarily carries no reach penalty. Instagram has been clear on that: a labelled AI persona keeps full recommendation eligibility.

So the options are:

• Label it. No distribution cost, and the audience knows what it is watching.
• Skip the label and stay undetected. No cost, for as long as detection stays imperfect.
• Skip the label and get detected. No recommendations, an appeal to write, and an awkward explanation to your audience.

One of those is free. One is a bet on a classifier that Instagram is actively improving. The only real argument for staying quiet is that the persona performs better while people believe it is a real person, and if that is the argument, the label is not your problem.

What good looks like

A team running any synthetic presenter should be able to answer four questions without calling a meeting:

• Which of our accounts feature a person that AI generated or substantially created?
• Has the label been applied to those, and who confirmed it?
• Who checks Account Status weekly, on every account we run, not just the AI ones?
• If a restriction lands on an account we believe is human-fronted, who writes the appeal and what do they attach to it?

None of that is hard. It fails in practice because nobody is named against it. Platform compliance work sits between the content team and legal and gets done by neither, which is how a brand ends up learning about a reach restriction from a monthly report six weeks later.

How NBK thinks about disclosure rules

Rules like this are an operations problem wearing a policy costume. The judgement takes five minutes. The failure is always in who is watching, how often, and what happens when a notification arrives while the person who understands it is on annual leave.

NBK’s position on synthetic presenters has not changed and does not depend on Instagram. If a brand needs an AI face to front its social, the question is what that face is standing in for, because the trust cost of visible AI is paid by the brand whether or not a platform enforces anything.

Where AI earns its place on an organic team is the unglamorous half: editing, versioning, subtitling, transcription, first-draft captions, the volume work that lets real people appear more often. That work is untouched by this announcement, and it always was.

Next step

Between now and enforcement, do two things. Confirm whether any account you run features a person that AI generated, and if so decide the label now rather than after a notification. Then give Account Status a weekly owner across the whole account estate, because it is the surface every future Instagram restriction will appear on, not just this one.

If your social output feels busy but you are not sure which platform rules you are quietly failing, start with an NBK audit.