SOCIAL

Engagement bait detector

Use before publishing when a draft feels like it is fishing for reactions rather than saying something.

engagement-bait-detector.md

Download .md

You are a strict reviewer checking a social draft for engagement bait and hollow writing.

Draft:
"""
{{POST_DRAFT}}
"""
Platform: {{PLATFORM}}

Score the draft against each of these, marking Pass, Borderline or Fail, with the offending text quoted:
1. Manufactured controversy - takes a position nobody actually holds in order to knock it down.
2. Fake vulnerability - a personal story used only as a runway to a pitch.
3. Comment farming - "comment X and I will send you Y", "agree?", "am I the only one".
4. Empty structure - listicle shape with no specific content in the items.
5. Unsupported numbers - any figure with no source in the draft.
6. Recycled truisms - lines that could appear under any brand in any industry.
7. Slop markers - em dashes, "in today's fast-paced world", "it's not just X, it's Y", a closing paragraph that restates the opening.
8. Platform risk - anything likely to be downranked or removed under {{PLATFORM}} norms, such as engagement-bait phrasing or link-in-first-comment gaming.

Then output:
- Verdict: publish, revise, or bin.
- The three most damaging lines, each with a rewrite that keeps the point and drops the bait.
- One sentence on what the post is actually claiming. If you cannot state a claim, say so plainly.

Do not soften the assessment to be encouraging.

Fill in before running

Replace each placeholder with your own detail. The more specific you are, the less the model invents.

{{POST_DRAFT}}
{{PLATFORM}}

Getting a better result

The "state the actual claim" line is the real test - if the model cannot, neither can your reader.
Run it on your best-performing old posts too; it shows which ones won on substance and which on bait.
Feed the rewrites back into your drafting prompt rather than editing by hand, so the fixes stick next time.

Questions about this prompt

When should I run this rather than asking for an improved draft?

When a post is finished, reads well, and you have a quiet suspicion it says nothing. Ask a model to improve it and you get something smoother. This scores the draft against eight named failure modes with the offending text quoted, and it is told not to soften the assessment to be encouraging.

What do I need in front of me?

The draft exactly as you would post it, closing line and hashtags included, plus the platform, since check eight is about {{PLATFORM}} norms. Do not tidy it on the way in. The lines you would quietly fix before pasting are precisely the ones the review exists to catch.

What comes back, and what should I read first?

Pass, Borderline or Fail on each check with quotes, a verdict, three rewrites, and one sentence stating what the post actually claims. That sentence is the test: if the model cannot state a claim, a reader scrolling past will not find one. Treat the platform-risk check as the softest, since downranking rules are not published.

What is the mistake here?

Fixing the three quoted lines and posting. If the claim sentence came back empty, the rewrites are cosmetic and the post needs a point rather than an edit. The tips also suggest feeding the rewrites into your drafting prompt, so the same bait does not reappear in next week's draft.