Field notes on the channels categories get discovered in.

The rules do not care what drafted the words. They care who the words were said for. "AI and compliance" lumps three separate things together, and only one of them, the FTC's fake review rule, comes with real penalties.

A BMJ Open study of the 180 most visible UK sources on hormone therapy found a conflict of interest in 117 of them. Just over half of the voices were professionals, not influencers, so the most visible menopause information is mostly credentialed, and mostly commercial.

More than 40 million people ask ChatGPT a healthcare question every day, most of them outside the hours a clinic is open. The research on whether those answers are any good is mixed, and the models sound exactly as confident when they are wrong.

The highest-reach post in a women's health community this week was almost certainly one nobody should sell into. If a brand's system for finding conversations sorts by engagement, that post arrives at the top of the queue on Monday morning, and sooner or later somebody with a quota posts a product mention into it.

Reddit will sell you an ad slot inside a comment thread, so the claim that you cannot buy your way into the conversation is wrong. What budget still cannot buy is the answer to the specific question, credibility that survived a public challenge, and the ability to say the product will not help.

In May 2026, Cornell researchers showed that thirteen words added to one frequently retrieved thread can make an AI agent recommend a product by name. Weeks later, Reddit published enforcement numbers built to catch exactly the kind of automation that attack does not need.

In January 2022, every one of 60 women's health organisations had had an ad rejected by Meta. The policy was revised. Six years later, 84 percent still report rejections, and a category has rebuilt how it acquires customers, one workaround at a time.