AI posts in a quiet community: harmless filler or a broken promise?

Every community goes quiet at some point. The regulars get busy, a bigger platform pulls people away, and the front page shows the same threads for weeks. For anyone who has run a forum or a member site, that silence is uncomfortable.

A newer answer to that silence keeps coming up in the communities we follow, and it has split people sharply.

One long-running forum admitted to using AI to generate posts to counter falling engagement, and some of those posts appeared under the names of real, long-time members. The reaction was close to unanimous on one point: faking activity is bad, but putting words in real people's mouths is far worse. Yet even in that thread, a few people pushed back on the framing. One argued that "fake activity" is still activity, and that if members don't like it they'll simply leave. Another pointed out that plenty of old forums now exist mainly to collect ad revenue, and that the incentive to keep pages fresh doesn't care who wrote them.

Meanwhile, open source communities have been tightening their rules against AI-generated submissions and low-effort project posts, and some are raising the bar for new projects to be shared at all. Elsewhere, people building community digests are experimenting with showing exactly what an AI summarised and what it left out, so members can see where the machine stepped in.

We build community software, so we have a stake in this. But we don't think there's one obvious answer for every community, so let's lay out the camps and the questions.

Camp 1: "Any AI activity breaks the deal"

The argument here is about the promise a community makes. When you join a forum, you assume the people posting are people. The value of a reply is that someone with real experience took the time to write it.

• If members can't tell which posts are real, every post loses value, including the genuine ones.
• Faking activity treats members as an audience to be retained, not as participants.
• Once trust is gone, a quiet community becomes a dead one very quickly.

People in this camp often say the honest options are better: post less often, say the community is quieter, or close it gracefully.

Camp 2: "Labelled AI help is fine"

This camp separates deception from assistance. An AI that writes posts under members' names is clearly over the line. But an AI that drafts a weekly summary, suggests related threads, or answers a basic question under a clearly marked bot account may actually help a small team keep a community alive.

• A clearly labelled helper is no different from an FAQ or a search box.
• Small volunteer teams are already stretched thin; burnout kills more communities than AI does.
• Summaries can bring old, valuable answers back to the surface.

The line for this camp is disclosure: say what the machine did, every time.

Camp 3: "Members will decide"

A smaller group argues that communities are self-correcting. If generated content is dull, people stop reading. If it's useful, they stay. Rules about origin matter less than whether the place is still worth visiting.

Critics reply that this ignores how members find out. Usually it isn't a slow drift away. It's one discovery, and then a very loud exit.

The questions underneath

Whichever camp you lean towards, a few questions keep coming up.

Where exactly is the line?

• Is an AI-written welcome post for new members acceptable? What about an AI-written answer posted by a staff account? A seeded "starter" thread to make a new category look less empty?
• Does it matter whether the admin edits the AI draft before posting?

What do you owe members?

• Should communities publish a clear policy on AI-generated posts, the same way they publish a code of conduct?
• If you used AI in the past without saying so, do you owe members a disclosure now?

How should new projects get a fair hearing?

• When open communities tighten rules to keep out low-effort AI content, do genuine newcomers get shut out too?
• Is "already has a community" a fair bar, or does it just reward whoever got there first?

What's the honest way to handle a quiet season?

• Is a quiet forum a problem to be fixed, or just a phase?
• What has actually worked for you: a regular event, a weekly thread, reaching out personally to old members, or simply waiting?

Our take, briefly

We lean towards Camp 2. A community is a promise that the people are real, so anything posted as a person should come from that person. Assistance that is clearly labelled seems fair game; impersonation never is. But we're genuinely unsure where "seeding" ends and "faking" begins, especially for a brand-new community with three members, and we'd rather hear from people who've faced it.

Your turn

1. Which camp are you in, and has a real experience moved you from one to another?
2. Have you ever discovered that a community you belonged to was faking activity? What did you do?
3. If you run a community, do you use AI anywhere in it, and do your members know?
4. What's the most honest thing you've done to get a quiet community talking again?

We'll read every reply, and we'd especially like to hear from admins who've been tempted and decided against it.