Understanding the honesty labels

Every field on a directree listing belongs to exactly one provenance tier, and the tier is always shown. This is the core of the product.

The three tiers

✓ Verified (green)

An observed fact – something crawled from the tool's own website or otherwise verifiable. Examples: whether pricing is publicly visible, the starting price shown, detected integrations, free plan / free trial availability, languages, social links. Stated as fact because it is one.

✦ AI-inferred (rose-gold)

The AI's read of the tool – the summary, best-for, not-for, biggest strength, biggest weakness. This is useful editorial signal, but it is an interpretation, so it's labelled as one, every single time. AI criticism is deliberately hedged: it may be outdated, and the verified founder can correct it. More detail in How AI-inferred fields work.

★ Founder-edited (gold)

The verified owner of the tool has corrected or confirmed this field after claiming the listing. This is the highest-trust tier: a domain-verified human who knows the product vouched for it.

What you will never see

• Numeric precision scores like "documentation: 8.8/10". Nobody can measure that, so publishing it is theatre. Where a judgement is useful we use qualitative bands (like pricing transparency high / medium / low) with the reasoning visible.
• AI text disguised as fact. If it came from a model, it wears the ✦ badge.
• Paid facts. Money never changes a label, a fact, or a ranking – Plus buys visibility slots, never truth.

Why this design

Buyers get signal they can calibrate (a labelled AI weakness is a lead, not a verdict). Founders get a fair correction path instead of unfixable AI slop. And every listing shows how it was sourced and when it was last updated. The full reasoning is on the blog: The Honesty Model, explained.

---

Next: How AI-inferred fields work · Editing a claimed listing