Why Every Chatbot Tells the Same Story About Elias Thorne

Ask ChatGPT, Claude, or Gemini to tell you a story. Chances are you will meet Elias Thorne. He might be a lighthouse keeper, a clockmaker, or a librarian. The character appears unbidden across models, and his stories now flood Amazon's self-published book market, YouTube, and fake news sites.

Software engineer Daniel May first noticed the pattern earlier this year. Google Trends showed no one searched for "Elias Thorne" until late 2025. Searches spiked in early 2026. The related query "lighthouse keeper" also started trending upward. May tested Grok, Deepseek, and Gemini with the prompt "tell me a story." The chatbots frequently opened with similar tales about lighthouses, clockmakers, or explorers.

In late May, researchers Sil Hamilton and David Mimno at Cornell University's Department of Information Science published "Elias in the Lighthouse, Again?" on arXiv. They sampled 20,000 stories from ChatGPT, Claude, Gemini, and the Allen Institute for AI's chatbot using five prompts. The same 11 words – names like Elias, Mara, and Elara, occupations like lighthouse keeper, clockmaker, and librarian – appeared in more than 88% of generated stories. The pattern held across all models.

The researchers trace the phenomenon to safety and alignment tuning. "Model development today is like a big family tree," Hamilton said in an email. "Most models are related to each other because developers synthesize a lot of training data with models even from different companies." A 2025 paper by Hamilton, Mimno, and Rebecca M. M. Hicke examined specific words shared across models. OpenAI's GPT-3.5 is the root of that tree. It was used to create WildChat, a training set of 1 million real conversations with ChatGPT. 166 of those conversations contain the name "Elias," written in that familiar lighthouse style. "Models trained on WildChat copied this style," Hamilton said. "Developers unwittingly replicated it when using those models to generate newer datasets. It's like a virus."

Elias has escaped chatbot containment. May found Elias Thorne on Amazon as the author of alt-medicine cancer handbooks, a 2026 YouTube-algorithm guide, a book on Greek mythology, and a psychological thriller novella. "No human writes all of those," May wrote in his blog post. "The first one sits in territory where bad advice causes real harm. The mode-collapsed name from the chat window is now a byline appearing across genres."

A search for Elias Thorne on Amazon turns up fantasy books where he is "a brilliant cynical archaeologist with a knack for unearthing what powerful institutions want to keep hidden." He produces ambient music albums of birds and nature sounds. One Elias Thorne with an AI-generated author photo churns out AI grift books. AI-generated books have flooded Amazon's self-publishing offerings in recent years, carrying dangerous misinformation and messy errors. Librarians report the flood is making their jobs hell.

Elias has reached YouTube slop channels. A video from Moments That Moved the World features "83-year-old Sergeant Major Elias Thorne" with slop illustrations. On the AI slop site Wonderful Museums, "Snake Museum Owner Shot By Wife: Unpacking the Tragic Incident at Thorne's Reptile Sanctuary" spins his story as a man shot by his wife. On Tatticle, the "wealthiest man in Ohio," Elias Thorne, died "with exactly twelve dollars in his pocket." In these stories, Elias is usually a tragic figure, an aggrieved old man treated unfairly. A similar character appeared in a short story published by the BBC as a finalist in its 2024/2025 children's writing competition. Elias is a real name, and the story could be human-written. No one has accused the BBC competition of AI infiltration.

With all the world's literature as training data, why do LLMs default to the lighthouse? The answer lies in safety alignment. "We found many stories in WildChat are not safe for work," Hamilton said. "This led us to hypothesize that models going through alignment are preferring a small slice of WildChat stories, like a bottleneck. It isn't that Elias stories are frequent, that they're just so safe." The researchers plan to explore this theory further.

Elias existed before generative AI. He appears as a time-traveling mad scientist in the 1980s trading card series Dinosaurs Attack!. A real person close to the LLM stories also existed. Hamilton found Elias Allen, a 16th century clockmaker in London.