4 rules for AI-drafted posts that do not read like AI

4 rules for AI-drafted posts that don't read like AI.

This week I fully reviewed an open-source toolkit for writing LinkedIn posts. Its methodology is built on 2026 data. The most usable parts for me:

1. Never open with a question. Question-first openings pull median likes down. Move the question to the close.

2. Make line one number-first. A precise number with a clear referent in the first sentence pulls median likes up.

3. Keep links out of the post body. A link lives in the first comment.

4. Scrub repeated AI tells. Template vocabulary and decorative staccato sentences are exactly what expert readers and LinkedIn's filter react to. The same research says LinkedIn's "AI content" report button costs a flagged post roughly 40% of its views.

Rule zero that I add myself: never enter a number that isn't real. A fabricated number burns trust exactly once.

This post was drafted and reviewed against the same 4 rules.

Which AI tell do you spot first?