When a horror novel became the center of a social media firestorm over suspected AI use, the fallout was swift: Hachette Book Group scrapped its planned US release and pulled the book in the UK, where it had already been published. The author denied the allegations, but the damage was done. For novelists watching from the sidelines, the episode crystallized a new fear—that a few suspicious sentences could end a career, and that proving human authorship is nearly impossible. Into that gap has stepped a cluster of startups marketing paper-trail tools that promise to document a writer’s process, though whether that documentation can actually shield anyone remains an open question.
The Accusation Problem

Andrea Bartz, a bestselling thriller writer, was among those alarmed by the reaction to the scandal. She saw the makings of a witch hunt, one in which anyone could be accused of using a large language model and would have no agreed-upon standard for proving otherwise. The tools now being pitched—Workings, Blindproof, Wintertext, Human Ink, and ObraVera—aim to fill that void by tracking and recording the writing process. Some encrypt and fingerprint text or typing rhythms with unique codes, producing a record that could show whether a manuscript was human-written or AI-assisted. The idea is to create evidence authors can point to if questioned, and to help establish provenance when text is scraped into training data without attribution.
The Authors Guild has taken notice without endorsing any particular product. Mary Rasenberger, the organization’s CEO, said the guild wants writers to know these tools exist as a safety net, though many of the companies have approached the guild to pitch their services. But Rasenberger also called the need to preserve receipts insulting and burdensome, and said it should not be required of authors. She argued that writers who sell their words, voice, and imagination have little reason to use AI, and that pride in their work means AI-generated submissions are rarer in publishing than outsiders might assume. Editors of literary fiction and nonfiction, she said, often insist they would recognize AI-generated writing if they saw it.
Detectors and Process Trackers Are Not Foolproof
Reliance on AI detection has its own hazards. Rasenberger worries about a climate of distrust and overconfidence in tools that can be fooled and have produced false positives. While Pangram claims an industry-leading false positive rate for creative writing, the guild has heard from people whose human-authored work was flagged as AI-generated. Max Spero, Pangram’s cofounder and CEO, acknowledged that paper-trail tools are not a perfect catch-all. Someone could hand-type AI text or use an auto-typer, he said, and the messy reality of writing—multiple files, drafts sent back and forth, voice notes on the go—means not everything gets captured.
The tools may also be better suited to copyright concerns than to confirming authorship. Brady Ridgway, technical founder of ObraVera, said his platform is designed to trace reuse of material stored in its database by registered users. A time stamp and cryptographic hash let the company later search for those words and say they were created at a certain time by a certain person. But it is up to the registered individual to declare whether AI was used. In the case of a writer accused of winning a short story prize with an AI-generated submission, Ridgway noted, everyone seemed to agree on who submitted the story. The software could only record that a declaration was made—not whether it was true.
Why a Shared Standard May Never Arrive
Jane Friedman, an author and publishing business expert, has argued that the era of AI scandals may be short-lived as detectors improve and authors see the consequences of passing off AI writing as their own. She told WIRED that agents and publishers are unlikely to adopt any verification method that costs extra money or time, and that getting them to agree on a shared standard would be like herding cats. Publishing is slow to adopt new technologies, especially those developed by outsiders, and even a theoretical benchmark would face logistical hurdles and the problem of faked or manipulated certification. If any of it costs extra money or time, she concluded, forget it.
For now, writers are left to fend for themselves. Bartz said she has heard of authors review-bombed on Goodreads because a stranger felt a book sounded like AI, and of readers swearing off anything published after 2022 to avoid generative AI entirely. Such reactions, she said, are a great way to tank an entire industry at once. Her forthcoming 2027 novel, Delete Your Account, satirizes generative AI head-on, with a blocked writer who asks a chatbot to imitate a favorite thriller and ends up trapped in that story. Bartz has written portions meant to sound like an LLM—invented transcripts and annoying marketing materials from a fictional AI company—and wonders what she will do if any of it comes back flagged as AI-written.
