A community feed runs on one quiet assumption: the posts come from members of the community. That assumption is what makes a local recommendation feel trustworthy and a neighbor’s warning feel real. It is also the assumption platforms have quietly started to break.

Increasingly, the AI-generated post in your feed comes from the platform itself, not from any person. Nextdoor and Facebook have experimented with seeding machine-written content designed to look like it came from an ordinary user. The words are smooth, the tone is neighborly, and nothing labels it as manufactured. The platform has become an author impersonating its own members.

When the venue starts writing the crowd’s lines, it is no longer hosting a conversation. It is staging one.

Shailin Dhar

Why it matters is not that the content is bad. Sometimes it reads fine. It matters because the value of a community feed is entirely borrowed from the belief that people are behind it. Spend that trust to boost engagement metrics and you are burning the only asset the feed had. Once you cannot tell whether a post came from a person or the product, you stop believing any of it, including the real neighbors.

How to read a suspicious post

• Look for an account with a history you can inspect versus a fresh or generic identity.
• Watch for posts that are perfectly on-brand for the platform’s engagement goals and oddly free of local specifics.
• Notice whether the platform discloses machine-generated content at all. Silence is itself a signal.

The fix is not complicated, it is just unprofitable in the short term: label what the machine writes, and never sign it with a human’s implied name. A feed that tells you who is talking can survive AI. A feed that hides it cannot survive being found out.