Yesterday, we looked at an award-winning French novel that was flagged as AI-written. The author and his defenders disputed that, claiming, among other things, that AI detectors can be unreliable.

At Wired, Lexi Pandell looked at the growing use of AI writing earlier this month, noting that

Book publishing, a notoriously slow industry, has been particularly slow to reckon with AI. Still, the reality is that AI is used by a growing subset of authors, publishers, and agents. Last year, Gotham Ghostwriters polled 1,481 working writers and found that 61 percent use AI tools, with 7 percent saying they’ve published AI-generated text. When Tuhin Chakrabarty, an assistant professor of computer science at Stony Brook University, used Pangram earlier this year to analyze 14,419 self-published novels, he found that nearly 20 percent had substantial AI-detection scores.

“Pangram Has Emerged as the Gold Standard of AI Detection. Should You Trust It?”, September 2, 2026

At Gizmodo, Webb Wright offers some thoughts on detector reliability in general. Pangram is highly rated but its CEO warned us against complacency:

Max Spero, the company’s CEO and one of its cofounders, told me detection tools like Pangram are doomed to forever play catch-up with generative models themselves. “I don’t think we’ll ever get to one hundred percent accuracy,” Spero told me. The best that the company can do, he adds, is “accumulate more certainty using more data.” “How Much Should You Trust AI Detection Tools?,” September 9, 2026

Probabilities, not certainties

Does that mean the tools are unreliable? Jon Gillham, CEO at another detector, Originality, says it’s more like this:

He argued determining whether a piece of text has been generated by AI is analogous to predicting the weather: If an algorithm tells you there’s a hundred percent chance of rain, it’s probably going to rain, and you’d be wise to bring an umbrella if you leave the house. But as we’ve all experienced, such confident weather predictions don’t always pan out. “Should You Trust”

So the detectors are dealing in probabilities, not certainties. Now, if six different detectors all come back with high confidence that written work is AI, reasons for doubting them all would require a deep dive into skepticism about the technology. That’s like doubting six different weather services’ storm warnings…

But we can’t outsource personal judgment to AI. Wright warns, “deepfake detection should be thought of more as a personal skill—to be learned and carefully practiced over time—than as something that can just be completely outsourced to a machine. That includes thinking critically about where information is coming from.”

Incidentally, here is a list of the leading AI detectors, as of today.