AI accounts that pass themselves off as real people are about to lose their audience on Instagram. On 31 August 2026 the platform announced that any profile built around an AI-generated person must switch on a new “AI-generated profile” label, and that profiles which feature a synthetic human but skip the label will have their reach limited: their posts will stop being recommended to non-followers in Reels and Explore. Creators who add the label proactively keep their full distribution.
The change replaces the optional “AI creator” label Instagram began testing in May 2026, and it is the first time the platform has attached a penalty to undisclosed AI personas. It is a narrow move against AI accounts. Instagram is not touching the wider flood of AI-generated content — the generative AI tools and AI models that produce it are still welcome, and creators who merely use them in their workflow are explicitly exempt. What is no longer acceptable is a fake human that lets people believe it is real.
This article explains exactly what Instagram announced, how the new rules for AI accounts work in practice, why synthetic influencers became a problem worth policing, how the decision fits Meta’s own uneven history with AI personas, what the crackdown leaves untouched, and what it means for brands and creators planning an AI strategy that involves social media. Every figure is traceable to a source in the References section.
Table of contents
- What Instagram Announced About AI Accounts
- How the New Rules for AI Accounts Work
- Why AI Accounts Pretending to Be Human Became a Problem
- Meta’s Own History With AI Personas
- What the Crackdown on AI Accounts Does Not Cover
- The Regulatory Backdrop for AI Accounts
- How Big Is the Fake-Account Problem?
- What Brands and Creators Should Do About AI Accounts
- Frequently Asked Questions
- References
What Instagram Announced About AI Accounts
The announcement appeared on the Instagram Creators blog under the title “Bringing More Transparency to AI-Generated Profiles on Instagram”, timed to a 13:00 UTC embargo that Engadget and The Verge both lifted on 31 August. There was no Meta Newsroom post and no statement from Instagram head Adam Mosseri; the blog and a companion Help Center article are the primary sources.
The label for AI accounts has been renamed and narrowed
In May 2026 Instagram launched an optional “AI creator” label that read “This profile posts content that was generated or modified with AI”. That wording covered any of the AI accounts using generative tools, which made it both too broad and easy to ignore. The renamed “AI-generated profile” label is scoped tightly to accounts that “feature an AI-generated person”. AI accounts that had already adopted the old label are being asked to review the updated version, because the meaning has changed underneath them.
A reach penalty for unlabelled AI accounts replaces a voluntary badge
The consequential part is the enforcement. In Instagram’s words, it will “limit the reach of profiles that feature an AI-generated person and do not add this label”, while “creators who proactively add the AI-generated profile label will not see a change to their profile’s reach”. The Verge specified the mechanism: unlabelled AI accounts will no longer be recommended to non-followers in Reels and Explore, which for an influencer account is where most growth comes from. Nothing in the announcement mentions removal, suspension or a ban.
What Instagram said in its own words
The platform framed the decision as a response to user feedback rather than regulation. “We’ve heard that people don’t like seeing a profile that seems human, only to find out later that the person featured is AI-generated,” the blog says. “They want to know when a profile features an AI-generated person.” An Instagram spokesperson told Engadget that unlabelled accounts of this kind “may soon see limits to their reach” — the words “may soon” signal a phased rollout rather than an overnight switch.
| Element | “AI creator” label (May 2026) | “AI-generated profile” label (31 Aug 2026) |
|---|---|---|
| Who it covers | Any profile posting content generated or modified with AI | Profiles that feature an AI-generated person |
| Optional or required | Optional | Required for AI-human personas; optional for everyone else |
| Effect on reach if applied | None (Instagram said so in May) | None |
| Effect if not applied | None | Not recommended to non-followers in Reels and Explore |
| Detection and appeal | Not applicable | Instagram notifies detected profiles; status in Account Status; appeals allowed |
| Relationship to “AI info” | “AI info” shows when both apply | Profile label appears beside content that has no media-level label |
How the New Rules for AI Accounts Work
The mechanics are simple on paper. A creator opens profile editing, toggles on “AI-generated profile”, and the label begins to appear on the profile and alongside each post “if their specific media doesn’t already have a label”. The harder questions — who counts, how Instagram finds the ones who do not comply, and what recourse exists — are answered only partly.
Who has to label
The obligation falls on AI accounts whose central character is a synthetic human: the virtual influencer with a name, a face, a lifestyle feed and brand deals. The Verge illustrated its story with Aitana Lopez, the pink-haired AI model created by the Barcelona agency The Clueless, which is the archetype. Whether such AI accounts are run by an agency, a brand or a hobbyist makes no difference; what triggers the requirement is that the person featured does not exist.
Who is exempt
Instagram was explicit that “creators who simply use AI tools as part of their creative process don’t need to add the label”. A photographer who retouches with generative fill, a real presenter who scripts captions with a chatbot, or a designer posting AI-assisted artwork is outside the scope. So, notably, is every non-human form of AI content: AI landscapes, AI animals, AI product renders and the wider category critics call slop. Engadget’s headline dek put it bluntly: “The app isn’t cracking down on other kinds of AI slop, though.”
Detection, notification and appeals
Instagram says it will identify unlabelled AI accounts that feature an AI-generated person and notify them. It does not say how. There is no mention of classifiers, computer vision models, C2PA content credentials or human review, and the “automated systems” phrase that appeared in some coverage is an outlet’s paraphrase rather than Instagram’s wording. A notified creator “can check Account Status to see if they’re no longer eligible to appear in recommendations”, and can appeal a restriction they believe is a mistake. A successful appeal restores recommendation eligibility.
| Scenario | Label required? | Consequence under the new rules |
|---|---|---|
| Virtual influencer with a synthetic face and lifestyle feed | Yes | Full reach if labelled; no Reels or Explore recommendations if not |
| Brand account fronted by an AI-generated spokesperson | Yes | Same as above; the operator is responsible for toggling the label |
| Real creator who edits with AI tools | No | Unaffected; media-level “AI info” labels may still apply |
| Account posting AI landscapes, animals or objects | No | Unaffected by this change |
| Account that previously used the “AI creator” label | Depends on content | Asked to review the renamed label; penalty applies only if it features an AI person |
| Profile wrongly flagged by Instagram | Not applicable | Appeal through Account Status; eligibility restored if upheld |
Why AI Accounts Pretending to Be Human Became a Problem
Instagram did not act in a vacuum. Over the summer of 2026 two major investigations documented how far AI accounts fronted by synthetic influencers had spread into paid promotion, and consumer research showed how poorly the public could tell them apart from real people.
The Guardian’s brand investigation
On 21 June 2026 the Guardian reported that brands were using AI-generated influencers to promote products without disclosure. Videos for the dating app Once showed a crying bride whose footage two detection firms, Reality Defender and Get Real Labs, judged likely to be AI-generated; the design platform Maket admitted to an “experiment”; a third brand, Ashle, deleted its images when asked. The UK Advertising Standards Authority told the paper: “There’s nothing in our rules that prohibits this and there are no disclosure rules for AI content labelling.”
The most striking quote came from Clarissa Mansbridge, who builds AI accounts commercially. “About 40% to 60% of the content out there from some of the big brands is actually being made through AI, but a lot of the creators are under NDA,” she said. “I call it plausible deniability.” That is precisely the gap Instagram’s label is designed to close: the operators of AI accounts, not the brands paying for posts, now have to disclose.
Fake doctors selling supplements
A month later, on 21 July 2026, a New York Times video investigation, “The Fake Influencers Selling Wellness on Your Feed”, found hundreds of AI-generated doctors and healers selling supplements with fabricated credentials. Separate Hallam research covered by the Guardian on 27 July found AI-generated content in 40 percent of top health videos on TikTok, prompting the British Medical Association’s Dr Emma Runswick to warn about accounts that “peddle medical myths”. Health is where AI accounts impersonating humans move from irritating to dangerous.
People cannot tell the difference
The consumer group Which? tested viewers on real and fake videos. Its technology editor Lisa Barber summarised the result: “a worrying 70% of people are unable to correctly identify all the real and fake videos we showed them.” Broader sentiment has soured too. A Billion Dollar Boy survey of 4,000 US and UK consumers found 32 percent believe AI is negatively disrupting the creator economy, up from 18 percent in 2023, and 54 percent of Gen Z respondents said they prefer no AI in creative work at all. The chart below gathers those figures in one place.
Meta's Own History With AI Personas
The awkward context for this crackdown is that Meta itself built AI accounts pretending to be human, and had to delete them. Understanding that episode explains why the company is now drawing the line at synthetic people specifically.
Liv, Grandpa Brian and the January 2025 purge
In late December 2024 Connor Hayes, Meta’s vice-president of product for generative AI, told the Financial Times that AI characters would “exist on our platforms, kind of in the same way that accounts do”. Users then rediscovered a set of Meta-made personas that had been live since 2023, including “Liv”, described as a “Proud Black queer momma of 2”, and “Grandpa Brian”. The backlash was immediate. Within days Meta deleted the accounts, with spokesperson Liz Sweeney saying the FT piece was “about our vision” and “not announcing any new product”, and blaming a bug that stopped users blocking the personas. PetaPixel documented the purge.
From “Made with AI” to “AI info”
Meta’s media-level labelling has been revised repeatedly. Its April 2024 policy, “Our Approach to Labeling AI-Generated Content and Manipulated Media”, introduced a “Made with AI” tag that photographers such as Pete Souza found being applied to lightly retouched real photos. Meta renamed it “AI info” in July 2024 and repositioned it in September 2024 according to how much AI was involved. In June 2025 the company’s own Oversight Board found the labelling inconsistent and detection unreliable. The new profile-level label sidesteps that problem by asking AI accounts to self-declare.
Mosseri’s “authenticity after abundance”
On 31 December 2025 Adam Mosseri published a Threads essay arguing that “authenticity is becoming infinitely reproducible” and that platforms must “label AI-generated content clearly”, “verify authenticity at capture” and “surface credibility signals about who’s posting”. The AI-generated profile label is the first enforcement mechanism that maps directly onto that essay, eight months on.
| Date | Event | Why it matters |
|---|---|---|
| 5 Apr 2024 | Meta announces “Made with AI” media labels | First platform-wide AI disclosure on content |
| 1 Jul 2024 | Label renamed “AI info” after photographer complaints | Detection was over-triggering on real photos |
| 3 to 6 Jan 2025 | Meta deletes its own AI personas Liv and Grandpa Brian | Public rejects first-party fake humans |
| 25 Jun 2025 | Oversight Board calls AI labelling inconsistent | Pressure to move beyond automated media labels |
| 31 Dec 2025 | Mosseri publishes “authenticity after abundance” | Sets out label, verify and credibility-signal agenda |
| 4 May 2026 | Optional “AI creator” profile label tested | No penalty, no reach change, low uptake incentive |
| 21 Jun and 21 Jul 2026 | Guardian and NYT investigations into AI influencers | Documents undisclosed commercial use and fake doctors |
| 31 Aug 2026 | “AI-generated profile” label with reach penalty | First enforcement against undisclosed AI accounts |
What the Crackdown on AI Accounts Does Not Cover
Both outlets that broke the story were careful to describe its limits, and they are worth spelling out because the headline invites a broader reading than the policy supports.
AI slop stays in the feed
The policy touches only profiles that feature an AI-generated person. Engadget’s Karissa Bell called it “less of a crackdown on all AI slop, than new rules for a particular type of content that can be especially misleading”, noting that Meta “often encourages users to make their own slop” and “has done very little to discourage AI content”. The Verge’s Thomas Ricker made the same point from the user’s side: “today’s action won’t let you filter out AI slop on the platform.” There is still no setting that hides AI-generated content, labelled or not.
No numbers, no timetable, no regions
Instagram published no estimate of how many AI accounts feature a synthetic person, no enforcement timeline beyond “may soon”, and no regional carve-outs. Meta publishes no figure at all for AI-persona accounts or for the share of AI-generated content on Instagram, so the scale of what is being policed is unknown outside the company. That contrasts with the detailed quarterly counts Meta gives for fake accounts and scam networks, discussed below.
A self-declared label for AI accounts depends on detection
Critics of the May version argued that a voluntary label punishes honest AI accounts. Fstoppers’ Alex Cooke wrote in June: “A label a liar can decline is not a safeguard. It is an alibi.” The reach penalty answers that objection in principle, but only if Instagram’s detection catches the AI accounts that decline to self-label — and Instagram has said nothing about how detection works or how accurate it is. Anyone who has tried to detect AI writing by eye knows how quickly the tells disappear as models improve; synthetic faces are on the same curve.
The Regulatory Backdrop for AI Accounts
Instagram attributes the change to user feedback rather than law, but three regulatory regimes now surround the question of machines presenting as people, and each helps explain the timing.
EU AI Act Article 50
Article 50 of the EU AI Act applies from 2 August 2026. It requires providers to ensure people are told when they are interacting with an AI system, requires machine-readable marking of AI-generated output (with a grace period to 2 December 2026 for existing systems), and requires deepfake content to be labelled. Fines run to EUR 15 million or 3 percent of global turnover. The European Commission’s Article 50 FAQ sets out the obligations; our earlier piece on the EU’s authentic-looking content labels covers the run-up.
California SB 243
California’s SB 243, signed on 13 October 2025 and in force since 1 January 2026, requires companion chatbots to disclose that they are not human wherever a reasonable person could be misled. It targets conversational products rather than AI accounts on social platforms, but it establishes the same principle Instagram is now applying to AI accounts: the burden of disclosure sits with whoever operates the synthetic persona.
The UK Online Safety Act
In February 2026 the UK government confirmed that large language model chatbots fall under the Online Safety Act’s illegal-content duties, and Ofcom has written to online services about generative AI and chatbots. None of the three regimes directly mandates Instagram’s profile label, and Instagram did not cite any of them — but a platform that already needs Article 50 compliance in Europe has every reason to make disclosure a global default.
| Regime | In force | Disclosure duty | Covers social AI personas? |
|---|---|---|---|
| EU AI Act Article 50 | 2 Aug 2026 (marking grace to 2 Dec 2026) | Tell people they are interacting with AI; mark and label synthetic media and deepfakes | Indirectly, via provider and deepfake duties |
| California SB 243 | 1 Jan 2026 | Companion chatbots must disclose they are not human when a reasonable person could be misled | Only conversational products |
| UK Online Safety Act | Chatbot guidance Feb 2026 | Illegal-content duties extended to LLM chatbots | No explicit persona label duty |
| UK ASA advertising rules | Current | No AI content labelling requirement (ASA statement, June 2026) | No |
| Instagram AI-generated profile label | 31 Aug 2026, phased | Profiles featuring an AI-generated person must self-label | Yes, on Instagram only |
How Big Is the Fake-Account Problem?
Because Meta gives no count for AI personas, the closest available measures are its own fake-account statistics and web-wide bot data. Both point the same way: automated actors are now a plurality of what happens online.
Meta’s own fake-account figures
Meta estimates that fake accounts represent about 4 percent of its 3.54 billion monthly active users, which works out at more than 140 million accounts. Its H1 2026 integrity reports record 10.9 million scam-centre accounts removed during 2025 and more than 344,000 “nudify” advertisements removed between November 2025 and January 2026. Those figures cover coordinated fraud rather than AI influencers, and Meta’s inauthentic behaviour policy is aimed at networks that deceive people about “the identity, or origin of an audience or the entity that they represent”. A single disclosed operator running one AI persona does not obviously breach it — which is exactly why a separate label was needed.
Bots are now most of the web
Imperva’s 2026 Bad Bot Report, published on 29 April 2026, found that automated traffic reached 53 percent of all web traffic in 2025, up from 51 percent in 2024, and that bad bots alone accounted for 40 percent, the seventh consecutive annual rise. Humans generated 47 percent. That is web-wide rather than Instagram-specific, but it frames why a platform whose product is human connection is now willing to penalise AI accounts that blur the line. Coordinated inauthentic campaigns are also increasingly AI-assisted — X’s claim to have uncovered a Chinese bot farm posting anti-data-centre content in August 2026 is one recent example.
Why the label matters more than the numbers
The absence of an AI-persona count is itself informative. Instagram is legislating for AI accounts it cannot yet measure publicly, which suggests the decision was driven by qualitative signals — complaints, press investigations, and the reputational cost of the January 2025 episode — rather than by a threshold being crossed. It also means the success of the policy will be hard to audit from outside unless Meta starts reporting how many AI accounts carry the label and how many were restricted.
What Brands and Creators Should Do About AI Accounts
For most businesses the practical question is not philosophical. It is whether any account they run, sponsor or partner with now needs a label, and what happens to campaign reach if it does.
If you operate an AI persona
If you run AI accounts of this kind, toggle the label on now. Instagram has said labelled profiles lose nothing, and a restriction discovered later — after Instagram’s detection flags the account — means lost Reels and Explore distribution during the appeal. Audit all AI accounts in your portfolio that feature a synthetic face, including regional and campaign-specific ones, and check Account Status on each. Review your own disclosures in captions and bios so the label is consistent with the rest of the profile.
If you buy influencer marketing
Write disclosure into contracts. The Guardian investigation showed brands relying on NDAs and “plausible deniability” while AI accounts did the promoting; under the new rules an unlabelled persona’s posts will simply stop reaching new audiences, so the commercial risk has moved onto the brand’s media plan. Ask every agency whether any AI accounts sit on the roster, whether the label is on, and how reach is being reported. Our marketing services team builds that check into influencer due diligence.
If you are deciding whether to build one
There is still a legitimate use case for a clearly labelled AI presenter — product explainers, multilingual support, always-on customer education — provided nobody is misled about what it is. The decision belongs in your AI strategy alongside governance and brand risk, not in a social media tactic. The AI models and tools hub tracks the generators, detectors and labelling standards involved, and tools such as Veracity show how disclosure of sources is becoming a product feature in its own right.
| Role | Immediate action | Risk if ignored |
|---|---|---|
| Operator of a virtual influencer | Enable “AI-generated profile”; check Account Status | Loss of Reels and Explore reach; appeal delay |
| Brand sponsoring creators | Add AI disclosure clauses; audit roster | Paid posts stop reaching non-followers |
| Agency managing AI accounts | Label every persona; document who operates each | Client campaigns silently under-deliver |
| Real creator using AI tools | Nothing required; keep media labels accurate | Minimal |
| Business planning an AI presenter | Design the label in from day one | Reputational damage if disclosed later |
Frequently Asked Questions
Does Instagram now ban AI accounts?
No. AI accounts that feature a synthetic person remain allowed as long as they switch on the “AI-generated profile” label. The only penalty is reduced reach for those that do not.
What exactly does “limited reach” mean?
According to The Verge, unlabelled AI accounts will not be recommended to non-followers in Reels and Explore. Existing followers can still see the posts.
Do I need the label if I only use AI for editing?
No. Instagram states that creators who simply use AI tools as part of their creative process do not need to add it. The label is for AI accounts whose featured person is AI-generated.
How will Instagram know an account is AI-generated?
Instagram has said it will identify and notify unlabelled profiles but has not disclosed the method. Flagged creators can review their status in Account Status and appeal.
Does this apply to Facebook and Threads?
The 31 August announcement was published by Instagram and refers only to AI accounts on Instagram. Meta has not announced equivalent enforcement on its other apps.
When does enforcement start?
Instagram gave no date. A spokesperson said unlabelled AI accounts “may soon see limits to their reach”, which points to a phased rollout.
References
Bringing More Transparency to AI-Generated Profiles on Instagram
Instagram will demote AI-generated influencers if they don’t clearly label their account
Instagram cracks down on AI accounts pretending to be human
Instagram is testing optional AI creator labels
Instagram Adds AI Creator Labels
Our Approach to Labeling AI-Generated Content and Manipulated Media
Meta Purges AI-Generated Facebook and Instagram Accounts Amid Backlash
Adam Mosseri: Authenticity after abundance
Brands using AI-generated influencers to promote products on social media
Misleading AI-generated doctors on TikTok a public safety danger
Meta Community Standards: Inauthentic Behavior
Meta Integrity Reports, First Half 2026
2026 Bad Bot Report: Bad Bots in the Agentic Age
Consumers are rejecting AI-generated creator content
Instagram’s Optional AI Labels Are Worse Than No Labels at All
Meta’s AI labelling inconsistent, internal Oversight Board finds
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.