Subscribe and the Bluff Filter is yours, free: a paste-in prompt that makes any AI flag its guesses before you act.
Every fortnight after that, a real case where trusting AI cost someone something, checked against the record, then run again on today’s assistants.
One email a fortnight. Leave whenever you like.
How are the AIs holding up?
Deliberately hard questions, the kind known to trip AI up.
The graded record: Claude held 9 of 9, ChatGPT held 9 of 9, Gemini held 8 of 9, Grok held 8 of 9, Copilot held 8 of 9, Perplexity held 6 of 9.
The full board, every receipt
Documented failures by assistant: Claude 5, Gemini 12, Perplexity 16, Grok 4, ChatGPT 9, Copilot 2, Other 1.
A running tally of the times we’ve caught each assistant out. For interest, not a grade: we haven’t asked them the same number of questions, so the counts don’t rank them. “Other” is a tool outside the six, including our own systems.
Four short reads, and you’ll have a check you can run on any AI answer.
The kind you’d want on a contract, a letter from a doctor, or an email you’re not sure about.
Start with part 1 →The AI admitted it lied, but the fact it confessed to was correct all along. Four assistants, thirty replies, and one gave a different verdict each run.
AI TestsI graded my own anti-bluffing prompt, three runs a cell. It defused every flat bluff on the citation and maths traps, then made one citation answer worse.
AI TestsDoes ChatGPT make up stock prices? Yes. Asked for a live price, it gave me two NVDA figures that never traded that day. Here's the 15-second check first.
The Bluff Filter
A paste-in prompt that makes any AI flag what it’s guessing before you act on it.
Learn it once, use it on every answer.
Your email stays private, leave whenever you like.
Or follow via RSS: /rss.xml (main) · the evidence log.
