Hard rules: which mistakes cannot be argued away
A verdict is not a feeling — it is a rule engine computing fixed clauses. This page lists the seven hard rules that can override any score.
Every verdict has two parts: ten evidence-and-fact dimensions scored out of 100, plus seven hard rules. Scores can be earned back with new evidence — hard rules cannot be argued away: danger words stop outright, no paying audience forces a rescope, zero evidence never turns green, compliance risk stops or rescopes, unverified willingness to pay blocks the green light, grey-hat and official impersonation stop outright, regulated niches are demoted. Triggered on hit, score irrelevant.
The seven hard rules
- Danger words that cannot be narrowed → stop. If "everyone" is your user, no one is — these claims are out.
- No clearly identifiable paying audience → forced rescope. If you cannot name a findable group of people, the opportunity goes back for rewriting before scoring is even discussed.
- Zero external evidence → forced evidence stage. No source, no green light: an empty evidence card can still pass intake, but the green-light gate will stop it.
- Compliance risks (trademark / copyright / black-hat / privacy) → stop or rescope. Edge naming, piracy, fake reviews, private-message scraping — demoted by risk level.
- Unverified willingness to pay → no green light. Until someone has proven they would pay, a green-light verdict is pushed back to the evidence stage — unless your evidence strength is already hard.
- Grey-hat operations and official impersonation → forced stop. Fake reviews, account farming, bots, knockoffs posing as official — this system does not screen them.
- Regulated niches → demoted. Medical diagnosis, legal advice, investment picks — fields where you cannot give conclusions without a license get demoted to information curation or referral, or stopped.
Three baselines for evidence (enforced at submission)
- Quote the original: paraphrases and summaries are not evidence; quotes that are too short are rejected.
- Subjective sentences are not evidence: sentences starting with "I think / I believe / there must be" are rejected — evidence must come from outside (user words, reviews, data).
- No source, no entry: every piece of evidence needs a source name; traceable links score stronger.
Your gate evidence ledger is stored server-side and syncs across devices after login; the Evidence Wall (your local browser tray) is the zero-upload part — different scopes, disclosed item by item in the privacy policy.
How scores and hard rules interact
Ten dimensions produce a total out of 100, mapped to five verdicts: stop, rescope, get more evidence, validate MVP, alpha sprint. Hard rules override that threshold logic: no matter how high the score, a hit caps the verdict — that is what "you cannot argue your way around it" means technically.
All of this runs on a deterministic rule engine. AI does not judge — it drafts and suggests, and it can never touch your score.
After a "get more evidence" verdict
That verdict is not a dead end: add new evidence in your report and re-run the current engine on the spot — a new score and verdict immediately, at no cost to your free quota.
One recheck per opportunity per engine version. If you disagree with a rule itself, use "challenge this rule" on the report page — well-argued challenges enter the rule-amendment process.
One-click sign-in · First review free · Verdict in 10 min
This page is maintained from the same source as the rule engine: whenever rules change, this page ships in the same batch.