Dirty chatbot reviews — each bot scored on what actually matters

A leaderboard copied from a marketing page is not a review. Testing is.

The problem with most dirty chatbot reviews is methodology: they describe features from a landing page instead of measuring them in use. A bot that claims to have memory is not the same as a bot that remembered your name two weeks later — and no screenshot can prove the difference.

Our reviews test five things. Memory recall across sessions. Personality that holds past the opener. Reply latency at every hour. A privacy model with no hidden social graph. And a free tier broad enough that the reader can verify the score before spending anything.

The result is a scored roster, not a popularity contest. Every entry lists what worked, what did not, and the exact conditions under which we tested. The limitations sit next to the strengths because that is what makes the strengths believable.

You open a bot's review page, scan the five-criterion score card, and decide in seconds whether the personality fits. The scores are not opinion dressed as data — each one traces to a specific test we ran and documented.

If the review checks out, one tap opens the bot's thread. No form, no credit card wall, no app download between the review and the conversation. The gap between reading and testing is one click.