Nobody is penalising your AI content, and three quieter things are costing you reach

Every platform has now said the same thing in public, more than once. Content made with AI is not demoted for being made with AI. Google judges the outcome rather than the production method. Meta labels rather than buries. TikTok labels. Nobody is running a penalty.

All of that is true, and it has been read as an all-clear it does not deserve. The reach that softened for no reason, the post that did nothing, the campaign that just quietly underdelivered: none of it arrives labelled as a penalty. It arrives as your content not working, which is the hardest thing in marketing to diagnose, because it looks the same as everything else that does not work.

Three things are costing reach right now. Two of them cannot be seen from inside your own account.

Your audience is switching you off, by category

In October 2025 Pinterest gave users a control to see less generated content in the categories most prone to it: beauty, art, fashion and home décor. TikTok followed in November with a see-less-AI setting in the For You feed. Neither is an algorithm punishing anybody. They are people removing themselves from an audience, one category at a time, with no notice to you and nothing to appeal.

If you sell in one of those categories, a share of your audience has already opted out of seeing generated imagery. There is no line in your analytics that says so. It reads as a soft decline.

The label costs you most on the content that works hardest

Two experiments published in Electronic Markets in 2026 showed people the same Instagram profiles, with the content labelled human-created, AI-enhanced, or AI-generated.

Read the second half of that again. The cost is not spread evenly. It lands hardest on testimonials, on stories, on anything asking someone to feel something, which is the content you were relying on to do the heavy lifting.

Specs, delivery times and how-to guides are barely touched. The photo of a customer holding the thing is where it costs you.

Deceptive gets buried, and nobody tells you

Meta does not suppress a post for carrying a label. It does cut distribution on synthetic media it judges deceptive, and it does that without removing the post or flagging the account. No notification arrives. The campaign underdelivers and you spend a week testing creative that was never the problem.

The one Google does punish is volume, not AI

If you use automation, including AI generation, to produce content for the primary purpose of manipulating search rankings, that's a violation of our spam policies.

Notice what that sentence turns on. Not the tool. Purpose and value. Google's scaled content abuse policy applies the same way to pages written by a person, by a model, or by both. Forty pages with a named editor behind them are fine. Four hundred that nobody read is the shape being deindexed.

What to do about it

• Keep real photography for anything emotional. The hero, the customer, the face, the testimonial. That content carries no label, so no opt-out filter catches it and no engagement cost applies.
• Use generated imagery where nothing emotional is riding on it. Backgrounds, texture, pattern, scene dressing.
• When you disclose, explain instead of stamping. "Illustration, not a photograph of a customer" reads as honest. A bare AI badge reads as a warning.
• Put a named, qualified human on anything in a category people trust with their health, money or safety.
• If you sell in beauty, fashion, homewares or art, look at your Pinterest numbers by category rather than in total.

The thing to take from this is not that AI content is bad. It is that the cost moved. It used to sit with the platforms, who were going to police it. It sits with your audience now, and they do not send a warning. They just scroll differently.