When Hachette Book Group canceled the horror novel Shy Girl in both the US and UK markets in March 2026, the publishing industry confronted something it had no tools to resolve. Pangram, an AI-detection service, had flagged approximately 78 percent of the book’s text as AI-generated. Social media speculation had already spread widely, amplified by a report in the Times. Author Mia Ballard denied writing the novel with AI, attributing changes in a previously self-published version to an editor’s AI-assisted revisions, and said she is pursuing legal action. No court has yet ruled. The damage, however, was done.

Andrea Bartz, a novelist who served as lead plaintiff in a class-action suit against Anthropic that produced what has been described as the largest copyright settlement in US history at $1.5 billion, watched that episode unfold with particular dread. “I kind of immediately felt like this was going to lead to a potential witch hunt,” she told WIRED. Her concern is specific and structural. “There’s no agreed upon standard,” she said, describing the absence of any shared method for proving human authorship.

The Shy Girl Fallout and the Tools Rushing to Fill the Gap

The Shy Girl cancellation did not stand alone. Since that scandal, several other fiction authors have faced similar suspicions: one after winning a short story prize, another after publishers bid into the millions for his crime novel. Both denied using AI. The pattern has pushed a cluster of platforms into the conversation as provisional solutions.

Services including Workings, Blindproof, Wintertext, Human Ink, and ObraVera each promise to track and document an author’s writing process. Some encrypt or cryptographically fingerprint written work and typing rhythms with unique codes, generating a record designed to demonstrate provenance when text is later scrutinized. The pitch is straightforward: if a detector flags your manuscript, you can produce evidence of how it was made.

The Authors Guild, the main professional organization for writers in the United States, acknowledges these tools exist. Mary Rasenberger, the guild’s CEO, nonetheless called it “insulting” and a “burden” to have to preserve such receipts, and said the guild does not believe authors should be required to do so. The guild has stopped short of recommending any single platform.

Provenance, Data Control, and the Problem of Trusting the Tool Itself

The Arabiccasinos Team, which covers digital market developments across the Arabic-speaking world, notes that the scandal raises a question that goes beyond whether any given tool works. The harder question is who stores the process data these platforms generate, under what contractual terms, and how a tool demonstrates that its own provenance claims are reliable rather than self-serving.

That concern is sharpened by the declaration-only limitation embedded in ObraVera’s own design (described in more detail below), and by Pangram’s acknowledged false-positive rate. Pangram claims an industry-leading false positive rate of 0.01 percent on creative writing, but Rasenberger said the Authors Guild has heard of instances where authors fed their own human-written work into Pangram and received results indicating AI generation. A tool that misidentifies human prose cannot be treated as neutral infrastructure, however well-intentioned.

The same logic, the team observes, already governs reader behavior in another high-stakes digital arena. The Arabic Casinos Guide earns reader trust only by interrogating what each platform logs, who holds that data, and how transparently it demonstrates its games are fair. The demand is structural, not incidental.

“If what you’re selling are your words and your voice and your imagination, your creativity, using AI doesn’t make sense.”

Rasenberger’s framing captures why the false-positive problem stings so sharply. Authors who have done nothing wrong are now being asked to trust a third-party data custodian to defend them against a different third-party detector.

What Cryptographic Fingerprinting Can and Cannot Establish

ObraVera’s approach is among the more technically specific. The platform can cryptographically fingerprint patterns in writing and in audio voice recordings, attaching a time stamp and a crypto hash to confirm that specific words were created at a specific time by a specific registered person. Brady Ridgway, ObraVera’s technical founder, described it plainly: “These words were created at this time, by this person.”

The limit, though, is significant. Ridgway acknowledged that the system “would say that somebody on this day made a declaration that it was human only or not AI-assisted. It can’t confirm whether that’s a true statement or not.” The cryptographic record confirms the declaration, not the underlying reality.

Pangram’s CEO Max Spero acknowledged separate gaps. “Paper trail tools are not a perfect catch-all,” he said. “Someone could hand-type AI text, or use an auto-typer.” He also noted that the messiness of genuine writing practice, involving multiple files, voice notes, and back-and-forth drafts, means these tools do not capture everything. “The average writer doesn’t just sit down and type out a book,” Spero said. “Not all of this gets captured in writing process tools.”

Publishers Are Unlikely to Demand These Tools, or Accept Them

The structural problem may be less about technical adequacy than about institutional adoption. Jane Friedman, an author and publishing business expert who writes the newsletter The Bottom Line, predicted that reaching industry-wide agreement on a shared verification standard will be “like herding cats,” because publishing is slow to absorb new technologies, particularly those developed outside the industry. Her assessment of publisher appetite was blunter still: “If any of this costs extra money or extra time, forget it.”

That leaves authors in an awkward position. Even those who pay for verification services cannot be certain that agents or publishers will accept proof from any given app. There is no mechanism forcing publishers to honor what any platform certifies.

Rasenberger’s objection captures the moral dimension of that gap. Authors who have done nothing wrong are being asked to surveil their own creative processes to satisfy accusers armed with detectors that are themselves unreliable. Bartz described a broader climate of reader suspicion, including review-bombing on Goodreads based on AI allegations and readers declaring they will not read anything published after 2022.

Her personal bind is the sharpest illustration of the problem’s absurdity. Her forthcoming 2027 novel Delete Your Account features a blocked writer who uses an AI chatbot to imitate a thriller. To write it, Bartz deliberately composed passages designed to sound like large language models, including made-up transcripts and what she described as “annoying, AI-sounding marketing materials” from a fictional AI company. “I’ve written portions that are very much meant to sound like LLMs,” she said. “And I don’t know, if any of that comes back as AI-written, what will I do?”

There is no clean answer. The novel is human-written. Parts of it are built to pass as machine output. No current tool can hold both facts at once.

Photo: pvproductions via Magnific

CLICK HERE TO DONATE IN SUPPORT OF OUR NONPROFIT COVERAGE OF ARTS AND CULTURE