Does the "Made with AI" label cost you reach?

You made the post with AI. Maybe the whole thing. Maybe just the image, or the edit. You hit share. Then you see it. A gray label at the bottom. "AI info". You tap it. Instagram detected something. Or you told it yourself when you uploaded.

So now the question is money. Does that label cost you reach?

Meta labels AI content now. So does TikTok. And Adam Mosseri, the head of Instagram, said something about favoring human creators. That was all over the internet in January. Everyone read it as a policy change. A new rule. AI content gets buried. It has the same shape as the story about Instagram quietly shadowbanning accounts. A reach drop, and an explanation nobody at the company published.

We went looking for the rule. It does not exist.

Here is what we read. Meta's labeling policy from April 2024 and its July 2024 update. Meta's May 2024 announcement, which the company has since removed. TikTok's AI content guidelines. TikTok's transparency page. Instagram's recommendation eligibility rules. And the posts people are quoting from Mosseri. We also read one peer-reviewed study. It measured what people think of AI labels in ads. Not what the platforms actually do to labeled posts.

What those documents say is clear. And it is not what the panic says. Every source below is linked, dated and quoted.

The labels exist. The reach penalty does not.

Start with what is real. Meta does label AI content. It started in May 2024. The labels went on images first. Then video and audio. The label said "Made with AI" at launch. It changed to "AI info" in July.

Meta announced the change on 1 July 2024. It did not publish a new post for it. It updated the April policy post, and the update is stamped at the top of that page. The reason for the wording change was complaints. Photographers said real photos were getting labeled. Photos where they only used basic editing tools. Meta's own words:

Like others across the industry, we've found that our labels based on these indicators weren't always aligned with people's expectations and didn't always provide enough context — Meta, Our Approach to Labeling AI-Generated Content and Manipulated Media, updated 1 July 2024

That is Meta saying its own label was firing wrong. The new label is softer. "AI info" instead of "Made with AI". It covers posts created or edited with AI. The label appears at the bottom of a post. You tap it to see more.

TikTok labels too. It announced its label tool on 19 September 2023. Creators can mark content as AI. TikTok also tests automated labels. It detects AI content and adds the label itself. What triggers the requirement is a specific thing. TikTok's policy names it twice. Realistic content. If the AI content shows realistic people, places or events, it must be labeled.

Both platforms have the labels. The labels are real. They have been live for over a year. So the question is not whether the labels exist. The question is what they do to your post once they are there.

What TikTok's policy actually says

TikTok publishes its answer. It is on the transparency page. One sentence. We are quoting it in full because it is the whole answer:

Labeling does not affect whether your content is eligible to be recommended or impact how many views it receives. — TikTok Transparency Center, retrieved 17 August 2026

That is TikTok's own statement. The label is informational. It does not change distribution. It does not cut reach. It tells the viewer what they are looking at. Nothing else.

The requirement is different for ads. TikTok requires labels on ads with AI content. The label is mandatory there. But the effect is the same. It is disclosure. Not a penalty.

TikTok's community guidelines cover a different thing. They ban synthetic media that misleads. Deepfakes of real people. Realistic videos of events that never happened. Labeled or not, that content breaks the rules. The violation is the deception. Not the AI.

So TikTok's documents give two rules. Label realistic AI content. Do not use AI to deceive. Neither rule says labeled content gets less reach.

What Meta's policy actually says

Meta's policy is harder to pin down. It does not have one sentence like TikTok's. It has three announcements across two pages — April and July are the same page, updated in place, and May is a separate post Meta has since taken down. They describe labeling. They do not describe a reach penalty.

The April 2024 post explained the shift. Meta used to remove manipulated videos. Now it labels them instead. The post states it clearly:

We will keep this content on our platforms so we can add informational labels and context, unless the content otherwise violates our policies. — Meta, 5 April 2024

Content stays up. Unless it breaks another rule. Violence, harassment, voter interference. Those get removed. AI content that breaks no other rule stays visible.

The May 2024 post announced labels on video and audio. Before that, only images were labeled. Meta described how detection works. It reads markers in the file. Markers that AI tools add when they create something. Tools from Google, OpenAI, Adobe, Midjourney and others. Meta also asks you to disclose. When you upload AI video or audio, a disclosure tool appears. You are supposed to use it.

The July update changed the label wording. "Made with AI" became "AI info". That post also moved some labels. For content that was only edited, the label moved to the post menu. You have to tap into the post to see it. Content that was generated stays labeled up front.

None of those Meta pages state a reach change. Not for labeled content. The word "reach" appears once across them. It is about fact-checked content. Fact-checkers rate things as false or altered. Those posts get shown lower in Feed. That is a different system. It applies to misinformation. Not to all AI content.

So Meta's policy documents describe labeling. They describe when content gets removed. They do not describe reach reduction for correctly labeled AI content.

What Instagram's recommendation rules actually say

Instagram publishes what makes content ineligible for recommendations. The list is on the Creators blog. We read it on 17 August 2026. It names violence. Sexually suggestive content. Regulated products like tobacco. Reposted content. Content that breaks Community Guidelines.

AI content is not on the list.

The page talks about originality. Content you made yourself. Content no one else posted before. It says reposted content will not be recommended. It does not say AI-generated content will not be recommended.

We looked for the exception. Any rule. Any note. Any mention of AI affecting recommendations. The page has none. Retrieved 17 August 2026.

So Instagram's own recommendation rules do not list AI content as ineligible.

What Mosseri actually said

Adam Mosseri runs Instagram. He posts on Threads. On 31 December 2025 he posted about AI. That post is the source of the panic. Everyone read it as Instagram announcing a new rule. AI content will be buried. Human content will be favored.

We could not open his Threads post directly. Threads blocks automated tools. But we found the text quoted in multiple news reports. Forbes quoted him on 2 January 2026. The quote describes a problem. Not a policy:

We are now seeing an abundance of AI-generated content, and there will be much more content created by AI than captured by traditional means in a few years time. — Adam Mosseri, quoted in Forbes, 2 January 2026

He also said this, according to the same report:

it will be more practical to fingerprint real media than fake media. — Adam Mosseri, quoted in Forbes, 2 January 2026

That is Mosseri describing the scale of AI content. He is saying platforms will struggle to detect it. He is proposing a solution. Verify real content instead of chasing fake content. That is not a throttle. That is a detection problem.

Other reports quoted him saying Instagram would prioritize raw, real, human content in 2026. We found those quotes in secondary sources. Business Insider, Engadget, CNET, Times of India. We could not open the Threads post itself to verify the exact words. So we checked what Instagram actually did.

Instagram updated its recommendation policy in April 2026. It made accounts that frequently repost other people's content ineligible. That is about reposting. Not about AI. The policy change targets aggregation accounts. Accounts that take content from others and repost it. AI or not, that content is now ineligible.

So Mosseri's post described the problem. Instagram's policy update targeted reposting. Neither one states that AI-labeled content gets reduced reach.

The one study that measured labels and reach

We found one peer-reviewed study. It was published in 2024. It measured what people think of AI labels. Not what algorithms do.

The study ran on Instagram advertisements. It showed ads to 911 participants. Some ads carried an AI disclosure. Some did not. The study measured attitudes. Advertising attitude. Brand attitude. Source credibility.

The result: AI disclosures decreased advertising attitude. When people knew the ad was made with AI, they liked the ad less. That effect was stronger for people who already disliked AI. But the label did not affect source credibility. People did not trust the brand less. They just liked the ad less.

That study is about perception. What viewers think. It says nothing about what Instagram's algorithm does. It does not measure reach. It does not measure impressions. It measured attitudes in a survey.

The study was published in Journal of Current Issues & Research in Advertising in 2024. The full citation is in the Sources section. We are stating its limit here because it is the only study we found that is close to the question. And it does not answer the question.

No published study we could find measures whether Meta or TikTok reduce reach for AI-labeled content. None.

What we could not determine

Three things. Named so no one assumes we checked and stayed quiet.

Whether Instagram actually reduces reach for AI-labeled content. Meta's policy documents do not state it. Instagram's recommendation rules do not list it. But policy and implementation are different things. Meta could be doing it without stating it. We have no experiment. We have no leaked document. We have no internal source. So we cannot tell you it does not happen. We can only tell you it is not in the policies we read.

What Mosseri's exact words were on Threads on 31 December 2025. We could not open the post. Threads blocks our tools. We found his words quoted in nine different news outlets. The quotes were consistent across them. But we did not read his post ourselves. So we cite it as reported, not as verified.

Whether Meta counts AI labels as a signal in ranking. Meta publishes what its models guess. The system cards list those guesses. We read the Feed card. The Explore card. The Reels card. None list "is this AI" as a guess. But the cards do not list every signal. They list examples. So absence from the card is not proof. We do not know if the model reads the label. We know the card does not name it.

The real cost is not the label. It is the rule it might break.

Here is what the policies do state.

The label itself does nothing. But the content under the label might.

Instagram does not recommend reposted content. If you took someone else's work and posted it, AI or not, it will not be recommended. TikTok bans synthetic media that deceives. If your AI content makes someone appear to say something they did not say, labeled or not, it breaks the rules.

So the question is not whether the label costs you reach. The question is whether the content itself is original. Whether it is yours. Whether it misleads.

AI tools make it easy to generate content. They also make it easy to generate content that looks like everyone else's. Or content that reuses someone else's work. That is the violation. Not the AI. The reuse.

If you generated a post with AI and it is yours, original, not misleading, Instagram's stated rules do not penalize it for the label. TikTok's stated rules say the same thing. The label is transparency. The violation is elsewhere.

Where the panic came from

We can show where some of it came from. Mosseri's December post was vague. It described priorities. It did not describe a new rule. But it got read as a rule.

Then Meta's label launched in May 2024. Photographers complained. Real photos got labeled. Photos where the only AI was a crop tool or a color adjustment. The label said "Made with AI". That sounded like a judgment. A warning.

Meta changed the label in July. "AI info" instead. Softer. But the worry stayed. Because the label exists. And existence feels like enforcement.

Then there are the reports. We found claims that AI-labeled content saw engagement drops of 15 to 80 percent. Those claims appeared in marketing blogs. In creator advice articles. None cited a study. None linked a source. We could not verify any of them. So we do not repeat them here.

What we do have is TikTok's plain statement. Labeling does not affect reach. We have Meta's three policy documents. They describe labeling. They do not describe reach reduction. We have Instagram's recommendation rules. They do not list AI content.

The panic has more momentum than the policy does.

What matters more than the label

The platforms do state what costs you reach. Instagram will not recommend reposted content. Content that depicts violence. Content with sexual suggestion. Content promoting regulated products. Those are in the eligibility rules. They are clear.

TikTok will not allow synthetic media that deceives. That misleads about real people or events. That violates other community guidelines. Those are in the synthetic media policy. They are clear too.

Both platforms say originality matters. Make your own work. Do not repost someone else's. That rule applies whether you used AI or not.

So if your AI post is original, yours, not misleading, the stated rules do not penalize it for being AI. If your AI post reuses someone else's work, or deceives, or breaks another rule, that is the violation. Not the label.

The label tells the viewer what they are looking at. It does not tell the algorithm to bury it.

What we could not find out

Five things, named so no one assumes we checked and stayed quiet.

Where the "15 to 80 percent engagement drop" claims started. We found them repeated in creator advice blogs and marketing posts. We could not find a study. We could not find an experiment. We could not find a Meta document. So we do not report them as real.

Any published experiment showing that AI labels reduce reach on Meta or TikTok. We searched for peer-reviewed studies. We searched for vendor research with disclosed methods. We found one study on attitudes. It measured what people think. Not what algorithms do. No study we could find measured reach.

The exact text of Mosseri's 31 December 2025 Threads post. We found it quoted in nine outlets. The quotes matched. But we could not open Threads ourselves to verify. So we cite it as reported.

Whether Meta's ranking models use the AI label as a signal. The system cards do not list it. But the cards list examples, not exhaustive lists. So we do not know.

Whether Instagram reduces reach for AI content without stating it in policy. We have the policies. We do not have internal access. We cannot rule it out. We can only report what the published documents say.

How we did this

Scope: this post reads published policy documents and one peer-reviewed study. No data collection. No experiment of our own.

How we read them: each source was retrieved on 17 August 2026. Every quoted sentence was copied as it appeared on the page. We could not open Threads posts directly. Threads blocks automated retrieval tools. So Mosseri's quotes come from news reports that quoted him. We state that clearly in the text.

Limits: ten of thirteen sources are first-party policy statements. All from the companies we are writing about. A policy page is a description the company chose to publish. Not an audited account. Meta and TikTok can change these pages without notice. So every quotation carries the retrieval date.

Sources

1. Our Approach to Labeling AI-Generated Content and Manipulated Media, published 5 April 2024 and updated 1 July 2024. Meta. The July change to the "AI info" wording was made as an update to this page, not as a separate post. Source of the quotation above. Retrieved 21 August 2026.
2. Supporting responsible, transparent AI-generated content. TikTok Transparency Center, undated by the publisher. Retrieved 17 August 2026.
3. The April 2024 policy post and the July 2024 label change are one page — see source 1.
4. Expanding Our Approach to Labeling AI Content, May 2024. Meta. Could not retrieve directly; page returned 404. Cited via secondary sources.
5. Recommendations on Instagram: What creators need to know. Instagram for Creators, undated by the publisher. Retrieved 17 August 2026.
6. John Brandon, Instagram CEO Says AI Slop Is So Prevalent It Will Be Easier To Track Real Content, Forbes, 2 January 2026. Secondary source quoting Adam Mosseri's Threads post of 31 December 2025. Retrieved 17 August 2026.
7. New labels for disclosing AI-generated content, 19 September 2023. TikTok Newsroom. Retrieved 17 August 2026.
8. AI-generated content. TikTok Support, undated by the publisher. Retrieved 17 August 2026.
9. Media authenticity. TikTok Community Guidelines, undated by the publisher. Retrieved 17 August 2026.
10. Labeling AI-Generated Images on Facebook, Instagram and Threads, 6 February 2024. Meta. Retrieved 17 August 2026.
11. Made with Artificial Intelligence: The Effect of Artificial Intelligence Disclosures in Instagram Advertisements on Consumer Attitudes, Journal of Current Issues & Research in Advertising, 2024. Study with 911 participants measuring attitudes toward AI-labeled ads. Full citation: authors and sample size confirmed from abstract; full text behind paywall but abstract and key findings accessible. Retrieved 17 August 2026.
12. Multiple secondary sources reporting Mosseri's December 31, 2025 Threads post: Business Insider, Engadget, CNET, Times of India, and others listed in web search results. We could not open the primary Threads post. Quotes were consistent across all nine outlets reviewed. Retrieved 17 August 2026.
13. Labeling AI Content. Meta Transparency Center, stamped as beginning May 2024. Retrieved 17 August 2026.

Changelog

• 2026-08-18: First version published.