Industry & Advocacy News
May 26, 2026
As AI use by authors has expanded from research and brainstorming to generating entire books, publishers, agents, and readers are growing suspicious of AI-generated content—and online markets are already flooded with scammer AI books, faster than anyone predicted.
The New York Times reported that false AI-generated quotes appeared throughout Steven Rosenbaum’s recent nonfiction book, The Future of Truth—a book, notably, about AI’s effects on truth. The Times identified more than half a dozen fabricated or misattributed quotes; Rosenbaum acknowledged the errors in a statement, called them accidental, and said he has launched his own investigation.
Contests are also reporting an abundance of AI submissions, and suspicions of AI use are extended to the winners as well. Last week, the 2026 Commonwealth Short Story Prize was thrown into controversy when readers flagged that one of the five winning stories, “The Serpent in the Grove” by Jamir Nazir, scored 100 percent AI-generated on Pangram, a widely used AI detection tool. The story was published in Granta, which co-presents the prize. Both organizations have stood by it, noting that AI detectors are imperfect, while the Commonwealth Foundation said it is reviewing its selection process.
The stakes are high in both directions: A false positive can cost an author their publishing contract and reputation, and a false negative, when AI-generated work slips through reputable publishing undetected, can cause readers to lose their trust in authors and the industry. This creates an environment reminiscent of witch hunts where authors are worried about being falsely accused of using AI and how they will defend themselves. Some have reported changing their writing style, so they don’t sound at all like AI.
AI detection tools might seem to offer clarity, but most, including those widely adopted in education, are unreliable, prone to false positives, and likely to become obsolete as AI models evolve. We decided to test them ourselves, focusing on whether they flag wholly human-authored work as AI-generated.
We chose five of the top AI detection tools available today and a list of ten Authors Guild articles—all published in 2022 or earlier, before the widespread availability of generative AI—and submitted each piece to all five detection tools. Because the articles predate the technology, any tool performing accurately should return a low AI-detection score across the board.
All five tools work the same way: Paste in text and receive a score, which is typically a percentage indicating how much of it the tool believes was AI-generated. Most stop there before hitting a paywall, with any deeper analysis locked behind a subscription. What those percentages actually measure is unclear.
One methodological note: some tools report a “human-written” percentage, while others report an “AI-written” percentage. For comparison purposes, all scores in our analysis reflect the percentage of text flagged as AI-written, regardless of how each tool frames its output.
The results varied widely, and in some cases, dramatically.
Pangram and Originality.ai were the most reliable performers. Pangram returned 0 percent across all ten articles. Originality.ai returned 0 percent on eight of ten, with 1 percent on the remaining two. Both tools correctly identified every piece as human written.
Grammarly performed nearly as well, returning 0 percent on eight articles and flagging two at 7 percent and 9 percent respectively—low enough that neither would likely trigger concern in practice.
ZeroGPT was inconsistent in ways that should give pause. Scores ranged from a low of 5 percent to a high of 76 percent, with several articles landing above 50 percent. The obituary for Joan Didion—a piece of plainly human writing if there ever was one—scored 66 percent. The Louise Erdrich Pulitzer Prize congratulations scored 76 percent. There is no obvious pattern that explains why some articles scored low, and others scored high, which is itself a problem: Unpredictability in a tool meant to make high-stakes determinations is not a minor flaw.
Sidekicker.ai produced the most alarming results by far. Every single article in our test was flagged as predominantly AI-written, with scores ranging from 71 percent to 100 percent. Two articles scored a perfect 100 percent. These are articles written and published years before the technology existed. A tool that cannot distinguish between human writing from 2020 and AI-generated content should not be used to make any determination about authorship.
Our test confirms that while a couple of AI detection tools accurately identify human-authored text, some commonly used consumer-facing AI detection tools are wildly inaccurate, which presents a major risk for authors. Moreover, these tools change constantly—updated models, shifting benchmarks, evolving AI outputs—and their accuracy at any given moment cannot be assumed.
AI-detection tools are AI models trained to recognize statistical patterns associated with large language model output, such as sentence rhythm, vocabulary distribution, and predictability of word choice. But polished, edited prose written by experienced human writers shares many of those same characteristics, because LLMs were trained on polished, edited prose written by experienced human writers. The more refined and controlled a writer’s style, the more it may resemble the output these tools are designed to flag.
This creates a troubling paradox. A writer who has spent decades honing clarity, economy, and precision is, by definition, writing in a way that overlaps with what AI has learned to produce. Detection tools cannot distinguish between a human writer who has mastered the craft and a machine that has learned to imitate it, because at the level these tools operate, there may be little difference to find.
Publishers should be extremely cautious about using the tools and should never rely exclusively on them. At minimum, publishers should disclose their methodology, keep themselves updated on the ongoing accuracy of the tools they use, and create processes that give authors a fair and full opportunity to defend themselves. Publishers should never cancel a contract or pull a book on the basis of accusations or use of these tools alone; it would be a material contract breach to do so.
The Authors Guild is looking at these issues closely and will issue further recommendations. In the meantime, if you are an Authors Guild member who has been falsely accused of AI use, please reach out to us for legal assistance.
