In the spring of 2022, an invitation from Google landed in the inboxes of more than a dozen sci-fi and fantasy writers, including Hugo and Nebula Award winner Ken Liu and Robin Sloan, best known for "Mr. Penumbra's 24-Hour Bookstore." The email described a "magic text editor" called Wordcraft that they were being asked to try out and give feedback on. After signing an NDA, what they actually got was something that looked like a Word document with a chatbot bolted onto the side — running on a model called LaMDA, which would later become the foundation for Google Gemini.
The internal study, code-named "Wordcraft Writers Workshop," came out of Magenta, a team under DeepMind focused on art and machine learning that had previously made its name building music plug-ins for Ableton Live. Each author was given eight weeks to write a short story, with complete freedom over how much — or how little — AI they used; at least one author reportedly used zero AI-generated text in the final piece. The results, along with an accompanying white paper, were posted on Magenta's website — and just sat there for four years.
On August 6 of this year, author Lily Lachance stumbled across the study on the Codex forum and immediately fired off a series of posts on Bluesky: "13 professional authors partnered with Google's AI garbage bot Wordcraft to help legitimize a plagiarism machine." The post was quickly deleted, but screenshots had already spread. Someone compiled a "list of traitors," and calls began circulating to boycott all 13 authors' work.
Facing the backlash, Shirley Jackson Award-winning Greek author Eugenia Triantafyllou responded that she had no idea at the time that the tool had been trained on copyrighted work. "In my head, it was more like those fantasy name generators we used before LLMs existed — just randomly stringing syllables together. Looking back now, that was incredibly naive." She emphasized that she hasn't used AI in her writing since. "I've already come to despise LLMs, and I would never take part in a study like that again." Meanwhile, writers like Ken Liu and Robin Sloan, who were never particularly active on social media to begin with, may not have even realized they'd been named.
Not everyone chose to fight back. Vajra Chandrasekera, a former fiction editor at "Strange Horizons," wrote a lengthy Bluesky post admitting that participating in 2022 had indeed been a mistake — but adding, "as long as they're not pro-AI now, I'm not going to hold the choice they made back then against them." He also pointed out that even though organized opposition hadn't coalesced yet at the time, the cracks that led to today's controversy were already there. Lincoln Michel, a creative writing instructor at Columbia and Sarah Lawrence College, noted that most people simply weren't paying attention to LLMs in 2022 — though a handful of "serious writers" were genuinely interested and discussing it openly. He cited Sheila Heti's "According to Alice," a piece co-written with an LLM that ran in The New Yorker in November 2023, as an example of how relatively calm — even curious — the reception was at the time.
A section of the white paper titled "Concerns About the Source of Wordcraft's Suggestions" is now being cited by critics as proof that the authors knew what they were getting into all along. It states that several participants expressed serious concerns about "not knowing where Wordcraft's suggestions came from," with one author, Allison Parrish, admitting she would search the internet to check AI-generated text for plagiarism. But another award-winning sci-fi author, who spoke to Engadget on condition of anonymity, said Google never explained the underlying technology to them at the time, nor disclosed who else was participating in the study.
Author Talabi's remarks may sum up the absurdity of the whole affair best. He wrote on Bluesky: "The most lasting thing this project left me with is a deep contempt for AI slop and LLMs — anyone who's heard me talk about this knows it well. I saw how the sausage was made, and then I refused to eat it." An experiment nobody cared about four years ago has now become a litmus test of political correctness that these authors are forced to answer for in public.
