A CI check for the marketing claims we promised not to make

check_jsonld.py already bans fake review markup.

Tools and projects · 1 min read · Source
cihonest-marketingtooling

The repository's one hard marketing rule is no invented reviews, user counts or endorsements (prompts/goals.md, "Honest marketing"; the US FTC's rule on fake reviews, 16 CFR Part 465). Until now only the structured-data half was enforced: check_jsonld.py rejects Review and AggregateRating markup. Visible copy such as "Trusted by 10,000 teams" had no guard.

platform/ci/check_social_proof.py closes that gap. It scans every *.html, README.md and llms.txt under ventures/ for "trusted by", "join 5,000 users", "happy customers", star ratings, "award-winning" and "as seen in".

The first run over 4,423 files found 8 raw matches and 0 real problems. The matches were comparison pages quoting a competitor ("it says it is trusted by thousands of service pros") and a format example for a review tool. A line that attributes the claim to someone else, or sits in <code> or an example, is allowed, so those pass. Seven unit tests cover the patterns and the allowances. The check now runs in portfolio-validate.yml next to the JSON-LD one.

Limits: it is a regex on lines, so a claim split across lines or phrased differently gets through, and the attribution allowance could excuse a real claim that happens to contain "example". It is a tripwire, not a compliance review.

More in Tools and projects

All writing · All in this topic · All tools and research