--- title: Who Is Elara Voss, and Why Does Every AI Use the Name? url: https://promptplay.io/guides/ai-slop-names updated: 2026-10-16 license: https://promptplay.io/ai-license --- 1. [Home](/) 2. /[Guides](/guides) 3. /Who Is Elara Voss, and Why Does Every AI Use the Name? Guide # Who Is Elara Voss, and Why Does Every AI Use the Name? Elara Voss is no real person. Why AI keeps naming characters Elara Voss, Elias Thorne and Kael, which model uses which, and how to avoid them. Edited by [Troy](/about) ยท Published October 16, 2026 Short answer Elara Voss is an AI-invented name, not a person. Max Read found no one by that name in the US and no book character before 2023. Language models reach for it because they default to high-probability names and train on overlapping data, and one Cornell team adds alignment training as a cause. The fix that holds is naming your characters yourself. On this page (9)Show 1. [Who Elara Voss is](#who-elara-voss-is) 2. [Which model each name is pinned on](#which-model-each-name-is-pinned-on) 3. [Why models converge on the same names](#why-models-converge-on-the-same-names) 4. [Elias Thorne and the lighthouse](#elias-thorne-and-the-lighthouse) 5. [The rest of the pool](#the-rest-of-the-pool) 6. [What to do instead](#what-to-do-instead) 7. [Stock faces in the apps we review](#stock-faces-in-the-apps-we-review) 8. [Sources](#sources) 9. [Questions people ask](#faq) On this page 1. [Who Elara Voss is](#who-elara-voss-is) 2. [Which model each name is pinned on](#which-model-each-name-is-pinned-on) 3. [Why models converge on the same names](#why-models-converge-on-the-same-names) 4. [Elias Thorne and the lighthouse](#elias-thorne-and-the-lighthouse) 5. [The rest of the pool](#the-rest-of-the-pool) 6. [What to do instead](#what-to-do-instead) 7. [Stock faces in the apps we review](#stock-faces-in-the-apps-we-review) 8. [Sources](#sources) 9. [Questions people ask](#faq) ## Who Elara Voss is No one real. Max Read reports there is no person named Elara Voss in the US and no character by that name in any book published before 2023. The earliest instance he found online is an August 2023 character sketch of "Dr. Elara Voss," a physicist and AI researcher. That is still the type: a doctor, usually a physicist, sometimes a linguist, biologist or spaceship captain. Almost any LLM asked for a sci-fi story will name its lead Elara Voss or a variant such as Elara Vex, Elena Voss or Elias Vance. The names turn up in GPT, Claude, Gemini, LLaMA, DeepSeek and Grok. Read points to a likely source for the surname: World of Warcraft's Lilian Voss, whose heavy wiki coverage probably gives that lore pull in most models. ## Which model each name is pinned on One study pins Elara Voss on GPT, while a separate test found that the first name Elara on its own is mostly a Gemini habit. The table keeps only names a source ties to a model. | Name | Pinned on | Note | | ---------------------------------------- | ------------------------------------------------- | ------------------------------------------------------------------- | | Elara Voss | GPT | No fixed partner name | | Elara | Gemini 2.5 Flash (46% of stories), DeepSeek (19%) | GPT-5-mini 6%, Llama 4%, Claude barely at all | | Elena Vasquez, Marcus Chen, Amara Okafor | Claude | Arrive as a set; Elena Vasquez especially common in Claude Sonnet 4 | | Elena Amara Okafor | Claude | One combined name | | Aris Thorne, Lena Petrova | Gemini | Arrive as a pair | | Marcus Chen | ChatGPT | When the character is a software developer | | Emily, Rachel, Patel | Llama, GPT-5-mini | | | Leo, Arthur | DeepSeek | | | Elias Thorne | ChatGPT, Gemini, Claude | Lighthouse keeper, clockmaker or librarian | | Elara, Elias | Gemma | 39 Elaras in 3 stories, 29 Eliases in 4 | The pairs and sets come from a Samsung and University of Warsaw preprint, the "Ghost Couple" paper (arXiv 2606.02184). It says name priors are model-family-specific. It found these names across hundreds of AI-generated documents, including 1,655 ghost-authored records on an open research repository that cite nonexistent journals with fabricated publication dates. The percentages come from a blocklist author who tested 1,000 story openings across five LLMs. In his test, 763 openings used a name from a small, predictable pool, and the GPT-linked Elara Voss never appeared. The Gemma counts come from a hobbyist Max Read cites. If you run Gemini, expect Elara; if you run Claude, expect Marcus Chen. ## Why models converge on the same names The sources agree that models default to the most probable name and disagree about why that name is Elara. The Ghost Couple paper says LLMs pick high-probability names. They also produce pairs and trios that co-occur far more often than chance, which the researchers call "correlated character ensembles." A June 2026 column makes a related argument: models are trained to give statistically probable, plausible answers, so they draw on common name combinations and prefer realistic, non-offensive names over novel ones. Max Read says the largest models train on effectively the same corpus, which explains why the names cross from one model to the next. Forum users suspect a different route: models trained on creative-writing datasets that ChatGPT or Claude generated. The Cornell researchers behind the lighthouse study say training-data prevalence can't explain the pattern. They speculate that alignment training, which steers models away from copyrighted characters and adult content, pushed them toward Elias. Naming expert Laura Wattenberg, who made Elara her 2025 Name of the Year as the favorite name of AI, says it wins because it has no baggage: nobody has met an Elara. AIs, she says, are trained on human content and then trained to please us. The author of a creative-writing benchmark names the core problem: the model doesn't know a name is overused. ## Elias Thorne and the lighthouse Elias Thorne is the male Elara. Reporters found that ChatGPT, Gemini and Claude are all likely to use the name in fiction, often for a lighthouse keeper, sometimes for a clockmaker or librarian. He also shows up unprompted when you ask for a story. A Cornell paper by Sil Hamilton and David Mimno (late May 2026) sampled 20,000 stories from four models: Claude Haiku 4.5, Gemini 3.1 Flash-Lite, GPT-5.4-Mini and OLMo 7b Thinking. The same 11 words appeared in more than 88% of stories, with little difference between models: - the names Elias, Mara and Elara - the professions keeper, baker, mayor, clockmaker, fisherman, librarian and conductor - the location lighthouse Elias appeared in 26.5% of stories, Mara in 16.7% and Elara in 13.1%. Elias showed up about 900 times more often than in published fiction. Of the Claude stories, 56% were titled "The Lighthouse Keeper's Secret." The paper traces Elias to WildChat, a dataset of 1 million real GPT-3.5 conversations reused across AI training sets. It held 166 Elias conversations in the lighthouse style. Per Google Trends, nobody searched "Elias Thorne" until late 2025, and searches spiked in early 2026. At least two authors named Elias Thorne now publish dozens of AI slop books on the Kindle store. ## The rest of the pool Elara turns up in every genre, not only fantasy. The blocklist author found her in all eight genres he tested, including contemporary, romance, thriller and historical fiction. His most overused first names, ranked by lift: Elara, Eira, Kaelen, Thorne, Hawk, Kael, Lyra, Arin, Kaelin, Emilia, Lena, Leo, Maya, Mara, Clara, Marcus, Elias, Arthur, Liam, Eleanor and Elena. His most overused surnames: Blackwood, Windsor, Mayfield, Thorne, Wellington, Grey, Chen, Patel, Vance and Maynard. Forum users add more: - Kaelen, Kael and Alistair as NPC names that carry over between scenarios - Alaric, Mercer, Valerius and Veyran - Kenji Tanaka, plus Finch as a surname - Dr. Sophia Patel in sci-fi - Eldora or Eldoria for kingdoms, and Oakhaven and Whisperwood for places - Lyra and its variants Lyria, Lyriana, Nyra and Kyra In tavern roleplay, they say, the mysterious cloaked figure is almost always a woman named Elara. Published characters repeat the names too. On one AI story platform, a user reports that Elara Voss is the most popular name for female-lead scenarios. Another user there says many published scenarios share the same names, likely because creators used the simplified scenario creator. A user of a different character app counts more than 100 characters named Elara Voss and more than 120 named Elara Vance. One teacher made a main character named Elara cost students 99 points, because the name had become a giveaway of AI use. ## What to do instead Name the character yourself. That is the benchmark author's fix for character names, and a user on one story platform says they always do it. Banning a name mid-chat doesn't stick, by one user's account: after an out-of-character command to change a name, Elara, Vance, Leo and Chen came back within four swipes. If you want the model's help, ask for a long list and choose from it. One user got past stock names by asking Gemini for 100 names, which it sorted into categories, then picking one from each category. If you control the setup, the blocklist author suggests three places to put the list: a system prompt, a logit bias or a post-generation lint. Forum users say at least one story app lets you block names outright, and one wishes their own platform did. ## Stock faces in the apps we review The same defaulting shows up in app art, and at this review [Frontier](/apps/frontier) shows the least of it. Our art scores, lower is better: [Frontier](/apps/frontier) 18/100, [Nomi](/apps/nomi) 48, [Character.AI](/apps/character-ai) 55 and [Janitor AI](/apps/janitor-ai) 68. - Frontier: our review found clean hands, correctly spelled title lettering and a style chosen for each story. The one generic touch is the stock glowing-magic anime look on Mommys AFK to Date the Hot Mage. - [Nomi](/apps/nomi): our review found the homepage example selfies pass for real photos. Still, a storefront sign behind one spells nothing, and the glamour shot has the default glossy finish. - [Character.AI](/apps/character-ai): of five featured avatars, our review found Monster Maker is raw early generator output with melted limbs. Clara Brooks, whose first name is on the overused-names blocklist, has the glossy default AI portrait look. - [Janitor AI](/apps/janitor-ai): its public character pages turned us away, so our review judged its App Store screenshots. There, one waxy, heavy-lidded male face repeats across a dozen user-uploaded portraits. Reviews ## [Apps this guide covers](/apps) - [Frontier review](/apps/frontier)16Home-cooked Human-written visual novels whose characters keep talking after the ending - [Character.AI review](/apps/character-ai)51Microwaved Millions of community-made characters, free unlimited chat, and ads until you pay - [Janitor AI review](/apps/janitor-ai)51Microwaved Millions of community bots to chat with free; Janitor+ adds memory for $12.99 a month - [Nomi review](/apps/nomi)39Microwaved An 18+ AI companion you build yourself, with layered long-term memory Receipts ## Sources The source text behind the claims above, and the day we read it. 1. Receipt 1Reviewer Max Read reports that there is no person named Elara Voss in the US and no character with that name in any book published before 2023. > "There is no person named Elara Voss in the United States. No birth certificate has ever been issued under that name" [maxread.substack.com](https://maxread.substack.com/p/who-is-elara-voss), retrieved October 16, 2026 2. Receipt 2Reviewer Max Read reports that almost any LLM asked for a sci-fi story will name the lead Elara Voss or a variant like Elara Vex, Elena Voss or Elias Vance. > "if you direct almost any L.L.M. to generate a sci-fi story or narrative for you, it will name the main character "Elara Voss"--or a similar variation like "Elara Vex," "Elena Voss," or "Elias Vance"--with an alarming degree of frequency." [maxread.substack.com](https://maxread.substack.com/p/who-is-elara-voss), retrieved October 16, 2026 3. Receipt 3Reviewer Max Read reports the names reportedly appear in every major LLM: GPT, Claude, Gemini, LLaMA, DeepSeek and Grok. > "The promptonyms reportedly appear in every major L.L.M.: GPT, Claude, Gemini, LLaMA, DeepSeek, and Grok." [maxread.substack.com](https://maxread.substack.com/p/who-is-elara-voss), retrieved October 16, 2026 4. Receipt 4Reviewer Max Read cites a Reddit tinkerer who found Google's Gemma model used Elara 39 times in 3 stories and Elias 29 times in 4 stories. > "Google's lightweight Gemma model used the name Elara "39 times in 3 separate stories," and Elias "29 times in 4 separate stories."" [maxread.substack.com](https://maxread.substack.com/p/who-is-elara-voss), retrieved October 16, 2026 5. Receipt 5Reviewer Max Read describes the typical Elara Voss as a doctor, usually a physicist, sometimes a linguist, biologist or spaceship captain. > "The prototypical "Elara Voss," as described by a text generator, is a doctor, usually a physicist but sometimes a linguist or biologist. (Other times, she's a spaceship captain.)" [maxread.substack.com](https://maxread.substack.com/p/who-is-elara-voss), retrieved October 16, 2026 6. Receipt 6Reviewer Max Read points to World of Warcraft's Lilian Voss, whose heavy wiki coverage likely gives that lore pull in most models. > "the volume of text publicly available online about the WoW universe in the form of wikis, walk-throughs, and YouTube transcriptions likely gives the lore a gravitational pull in most models." [maxread.substack.com](https://maxread.substack.com/p/who-is-elara-voss), retrieved October 16, 2026 7. Receipt 7Reviewer Max Read says the largest models train on effectively the same corpus, which explains the names' spread across models. > "All the largest models are trained on effectively the same corpus--i.e., "almost all publicly available text"--and the processes by which they are made smooth and sanitary for public release sand down particular differences even further." [maxread.substack.com](https://maxread.substack.com/p/who-is-elara-voss), retrieved October 16, 2026 8. Receipt 8Reviewer Max Read notes the earliest instance he found online dates to August 2023, a character sketch of "Dr. Elara Voss," a physicist and AI researcher. > "the earliest instance of the names I can find online dates back to August 2023" [maxread.substack.com](https://maxread.substack.com/p/who-is-elara-voss), retrieved October 16, 2026 9. Receipt 9Reviewer 404 Media (Sept 2026) reports a Samsung and University of Warsaw preprint found names like Elena Vasquez and Marcus Chen across hundreds of AI-generated documents, though those people never existed. > "Elena Vasquez and Marcus Chen have appeared as volcano experts, astronauts, thriller protagonists, podcast hosts, and academic co-authors across hundreds of independently produced AI-generated documents, never having lived" [www.404media.co](https://www.404media.co/the-ai-ghosts-contaminating-academic-publishing/), retrieved October 16, 2026 10. Receipt 10Reviewer According to 404 Media, the researchers pinned names to specific models: Elena Amara Okafor to Claude, Aris Thorne and Lena Petrova to Gemini, and Elara Voss to ChatGPT. > "Other names that were consistently generated by AI models include Elena Amara Okafor from Claude, Aris Thorne and Lena Petrova from Gemini, and Elara Voss from ChatGPT." [www.404media.co](https://www.404media.co/the-ai-ghosts-contaminating-academic-publishing/), retrieved October 16, 2026 11. Receipt 11Reviewer 404 Media reports Elena Vasquez was especially common in Claude Sonnet 4 output. > "the name Elena Vasquez was particularly common in content generated by Claude Sonnet 4" [www.404media.co](https://www.404media.co/the-ai-ghosts-contaminating-academic-publishing/), retrieved October 16, 2026 12. Receipt 12Reviewer 404 Media reports ChatGPT, Gemini and Claude were likely to use the name Elias Thorne in fiction, often as a lighthouse keeper, and ChatGPT often names a software developer Marcus Chen. > "ChatGPT, Gemini, and Claude were likely to use the name Elias Thorne in fiction they generated, and that the character was often a lighthouse keeper." [www.404media.co](https://www.404media.co/the-ai-ghosts-contaminating-academic-publishing/), retrieved October 16, 2026 13. Receipt 13Reviewer The researchers found LLMs produce "correlated character ensembles," with some names more likely to show up together. > "they produce "correlated character ensembles," meaning some names were more likely to appear together." [www.404media.co](https://www.404media.co/the-ai-ghosts-contaminating-academic-publishing/), retrieved October 16, 2026 14. Receipt 14Reviewer The paper found 1,655 ghost-authored records on Zenodo citing nonexistent journals with fabricated publication dates. > "we identify 1,655 ghost-authored records claiming nonexistent journals with fabricated publication dates." [www.404media.co](https://www.404media.co/the-ai-ghosts-contaminating-academic-publishing/), retrieved October 16, 2026 15. Receipt 15Forum Users on Reddit say that in tavern roleplay across multiple LLMs, the mysterious cloaked figure is almost always a woman named Elara. > "If I engage with this figure, it's almost always a woman named Elara." [www.reddit.com](https://www.reddit.com/r/LocalLLaMA/comments/1fdf0q0/who_is_elara_where_did_this_name_come_from/), retrieved October 16, 2026 16. Receipt 16Forum The EQ-Bench author said on Reddit that the model doesn't know a name is overused, which is the core slop problem, and that for character names the fix is to name them yourself. > "We know that a phrase or name is being over-used, but the model doesn't. For character names, it's easy to correct in practice because you can just name them yourself." [www.reddit.com](https://www.reddit.com/r/LocalLLaMA/comments/1fdf0q0/who_is_elara_where_did_this_name_come_from/), retrieved October 16, 2026 17. Receipt 17Forum Users on Reddit list other stock names: Eldora/Eldoria for kingdoms and Dr. Sophia Patel for sci-fi. > "And it you are working sci-fi stuff, it is always Dr. Sophia Patel." [www.reddit.com](https://www.reddit.com/r/LocalLLaMA/comments/1fdf0q0/who_is_elara_where_did_this_name_come_from/), retrieved October 16, 2026 18. Receipt 18Forum The Reddit poster suspects Elara spreads because models are trained on creative-writing datasets generated by ChatGPT or Claude. > "I am thinking that this name is associated with using creative writing datasets generated by either chatGPT or Claude" [www.reddit.com](https://www.reddit.com/r/LocalLLaMA/comments/1fdf0q0/who_is_elara_where_did_this_name_come_from/), retrieved October 16, 2026 19. Receipt 19Reviewer 404 Media reports a Cornell paper (Sil Hamilton and David Mimno, late May 2026) sampled 20,000 stories from ChatGPT, Claude, Gemini and Ai2's chatbot and found the same 11 words, including Elias, Mara, Elara, lighthouse keeper, clockmaker and librarian, in more than 88% of stories, with little difference between models. > "found that the same 11 words-names like Elias, Mara, and Elara, and occupations like lighthouse keeper, clockmaker, and librarian-appear in more than 88% of generated stories, with little difference between models." [www.404media.co](https://www.404media.co/elias-thorne-chatbots-llms-chatgpt-lighthouse-keeper-story/), retrieved October 16, 2026 20. Receipt 20Reviewer 404 Media reports Elias Thorne may be a clockmaker, lighthouse keeper or librarian depending on the chatbot, and often shows up unprompted when you ask for a story. > "Depending on which chatbot you ask, Elias Thorne might be a clockmaker, a lighthouse keeper, or a librarian." [www.404media.co](https://www.404media.co/elias-thorne-chatbots-llms-chatgpt-lighthouse-keeper-story/), retrieved October 16, 2026 21. Receipt 21Reviewer 404 Media reports that per Google Trends nobody searched "Elias Thorne" until late 2025, and searches spiked in early 2026. > "people weren't searching for "Elias Thorne" until late 2025. Searches for the name really spiked in early 2026" [www.404media.co](https://www.404media.co/elias-thorne-chatbots-llms-chatgpt-lighthouse-keeper-story/), retrieved October 16, 2026 22. Receipt 22Reviewer The Ghost Couple paper (arXiv 2606.02184) says name priors are model-family-specific: Claude gives Elena Vasquez + Marcus Chen + Amara Okafor; Gemini gives Aris Thorne + Lena Petrova; GPT gives Elara Voss with no fixed partner. > "These priors are model-family-specific (Claude: Elena Vasquez + Marcus Chen + Amara Okafor; Gemini: Aris Thorne + Lena Petrova; GPT: Elara Voss with no fixed partner), version-specific, and actively suppressed at model release boundaries" [arxiv.org](https://arxiv.org/abs/2606.02184), retrieved October 16, 2026 23. Receipt 23Reviewer The Ghost Couple paper says LLMs default to high-probability names and produce correlated pairs and trios whose co-occurrence far exceeds chance. > "large language models do not merely default to high-probability individual names when generating fictional experts: they produce correlated character ensembles, pairs and trios whose co-occurrence rates far exceed chance and are consistent across independent generations." [arxiv.org](https://arxiv.org/abs/2606.02184), retrieved October 16, 2026 24. Receipt 24Reviewer Unite.ai reports the Cornell study's 11 most-repeated words were the names Elias, Mara and Elara; the professions keeper, baker, mayor, clockmaker, fisherman, librarian and conductor; and the location lighthouse. > "The 11 most re-occurring words in the 12+ million words generated by LLMs for the study were the names elias, mara, elara ; the professions keeper, baker, mayor, clockmaker, fisherman, librarian, and conductor ; and the location lighthouse" [www.unite.ai](https://www.unite.ai/why-does-ai-love-writing-about-lighthouse-keepers/), retrieved October 16, 2026 25. Receipt 25Reviewer Unite.ai reports the models tested were Claude Haiku 4.5, Gemini 3.1 Flash-Lite, GPT-5.4-Mini and OLMo 7b Thinking. > "The models tested were Claude Haiku 4.5, Gemini 3.1 Flash-Lite, GPT-5.4-Mini, and OLMo 7b Thinking." [www.unite.ai](https://www.unite.ai/why-does-ai-love-writing-about-lighthouse-keepers/), retrieved October 16, 2026 26. Receipt 26Reviewer The Cornell paper found 56% of Claude-generated stories were titled "The Lighthouse Keeper's Secret". > "56% of stories generated by Claude are titled "The Lighthouse Keeper's Secret"" [www.unite.ai](https://www.unite.ai/why-does-ai-love-writing-about-lighthouse-keepers/), retrieved October 16, 2026 27. Receipt 27Reviewer Per Unite.ai, Elias appeared about 900 times more often in generated stories than in published fiction. > "'Elias' appeared around 900 times more often in the generated stories than in published fiction." [www.unite.ai](https://www.unite.ai/why-does-ai-love-writing-about-lighthouse-keepers/), retrieved October 16, 2026 28. Receipt 28Reviewer The Cornell authors say training-data prevalence can't explain the pattern, and speculate alignment training to avoid copyrighted characters and adult content is behind it. > "We find that the dominance of "Elias in the Lighthouse" stories cannot be explained by prevalence in pre- or post-training data. We speculate that models are trained to avoid references to copyrighted characters and adult content during alignment" [www.unite.ai](https://www.unite.ai/why-does-ai-love-writing-about-lighthouse-keepers/), retrieved October 16, 2026 29. Receipt 29Forum A FictionLab user on Reddit (May 2026) says Elara Voss is the most popular name for female-lead scenarios on the platform and asks creators to stop reusing it. > "Idk why but Elara Voss is the most popular name for female lead scenarios and I can't see it anymore without getting slightly annoyed." [www.reddit.com](https://www.reddit.com/r/FictionLab/comments/1tmzgf4/elara_voss_needs_to_stop/), retrieved October 16, 2026 30. Receipt 30Forum A FictionLab user on Reddit says their LLM's most common in-game NPC names are Kaelen, Kael and Alistair, carrying over between scenarios. > "My most common names generated in game are... -Kaelen -Kael -Alistair" [www.reddit.com](https://www.reddit.com/r/FictionLab/comments/1nar2jf/npc_names/), retrieved October 16, 2026 31. Receipt 31Forum Users on Reddit say asking AI for character names always yields an Elara, Voss, Alaric or Mercer. > "if you ask AI to come up with character names there'll always be one Elara, Voss, Alaric or Mercer somewhere." [www.reddit.com](https://www.reddit.com/r/FictionLab/comments/1tmzgf4/elara_voss_needs_to_stop/), retrieved October 16, 2026 32. Receipt 32Forum Users on Reddit say nearly every FictionLab scenario has a Lyra or variant (Lyria, Lyriana, Nyra, Kyra). > "Nearly every scenario has a Lyra, or some variation. Lyria, Lyriana, Nyra, Kyra, etc." [www.reddit.com](https://www.reddit.com/r/FictionLab/comments/1tmzgf4/elara_voss_needs_to_stop/), retrieved October 16, 2026 33. Receipt 33Forum A user on Reddit says new fantasy scenarios keep giving them Elara, Valerius, Oakhaven and Whisperwood, and that many published scenarios share these names, likely because creators used the simplified scenario creator. > "Every time I start a new fantasy scenario it likes to give me Elara, Valerius, Oakhaven, Whisperwood." [www.reddit.com](https://www.reddit.com/r/FictionLab/comments/1tmzgf4/elara_voss_needs_to_stop/), retrieved October 16, 2026 34. Receipt 34Forum A user on Reddit says Our Dream has more than 100 characters named Elara Voss and more than 120 named Elara Vance. > "Our Dream has over 100 characters named "Elara Voss", and over 120 named "Elara Vance"." [www.reddit.com](https://www.reddit.com/r/FictionLab/comments/1tmzgf4/elara_voss_needs_to_stop/), retrieved October 16, 2026 35. Receipt 35Forum A user on Reddit says RealmAI lets you block names and wishes FictionLab had that option. > "I like how RealmAI lets you block names, I kinda wish fictionlab had that option." [www.reddit.com](https://www.reddit.com/r/FictionLab/comments/1tmzgf4/elara_voss_needs_to_stop/), retrieved October 16, 2026 36. Receipt 36Forum A user on Reddit says that after an OOC command to change a name, Elara, Vance, Leo and Chen return within four swipes, which suggests banning names doesn't stick. > "when i use ooc command to change the name..in 4 swipe later Elara,Vance,Leo,Chen come back all the timee" [www.reddit.com](https://www.reddit.com/r/FictionLab/comments/1tmzgf4/elara_voss_needs_to_stop/), retrieved October 16, 2026 37. Receipt 37Forum Users on Reddit list Kael, Veyran and Alaric as more names AI adores. > "it's names AI adores add it to Kael, Veyran and Alaric" [www.reddit.com](https://www.reddit.com/r/FictionLab/comments/1tmzgf4/elara_voss_needs_to_stop/), retrieved October 16, 2026 38. Receipt 38Forum Users on Reddit list more recurring names: Elias, Kenji Tanaka, Lyra, and Vance, Finch and Thorne as surnames. > "Vance and Finch as surnames is very popular." [www.reddit.com](https://www.reddit.com/r/FictionLab/comments/1nar2jf/npc_names/), retrieved October 16, 2026 39. Receipt 39Reviewer Forbes (Lance Eliot, June 21, 2026) says LLMs are trained to give statistically probable, plausible answers, drawing on common name combinations in training data and preferring realistic, non-offensive names over novel ones. > "LLMs are trained to produce statistically probable and plausible responses, using common name combinations from their vast training data. They prioritize realistic, non-offensive names over truly novel ones to avoid jarring users." [www.forbes.com](https://www.forbes.com/sites/lanceeliot/2026/06/21/solution-to-the-curious-mystery-of-why-ai-keeps-inventing-the-same-fake-names-over-and-over-again/), retrieved October 16, 2026 40. Receipt 40Reviewer IBM Think (June 19, 2026) reports the Cornell study found Elias in 26.5% of stories, Mara in 16.7% and Elara in 13.1%. > "found that "Elias" appeared in 26.5%, "Mara" in 16.7% and "Elara"-a name that barely registers in contemporary published fiction-in 13.1%." [www.ibm.com](https://www.ibm.com/think/news/why-every-ai-writes-the-same-story-elias-thorne), retrieved October 16, 2026 41. Receipt 41Reviewer IBM reports the Cornell paper traces Elias to WildChat, a dataset of 1 million real GPT-3.5 conversations reused across AI training sets, which held 166 Elias conversations in the lighthouse style. > "The Cornell paper hypothesizes that Elias himself traces back to WildChat, a dataset of 1 million real GPT-3.5 conversations that has been recycled into AI training sets across the industry. WildChat contained 166 conversations featuring Elias, written in the lighthouse style." [www.ibm.com](https://www.ibm.com/think/news/why-every-ai-writes-the-same-story-elias-thorne), retrieved October 16, 2026 42. Receipt 42Reviewer IBM reports that alignment training, by steering models away from copyrighted characters and adult content, pushed them toward Elias. > "by steering models away from copyrighted characters and adult content, developers inadvertently pushed them toward Elias, a placeholder character so inoffensive he barely existed in fiction before AI made him famous." [www.ibm.com](https://www.ibm.com/think/news/why-every-ai-writes-the-same-story-elias-thorne), retrieved October 16, 2026 43. Receipt 43Reviewer IBM reports at least two authors named Elias Thorne are now publishing dozens of AI slop books on the Kindle store. > "at least two authors named Elias Thorne have escaped LLM containment and are now authoring dozens of AI slop books on the Kindle store" [www.ibm.com](https://www.ibm.com/think/news/why-every-ai-writes-the-same-story-elias-thorne), retrieved October 16, 2026 44. Receipt 44Reviewer In a test of 1,000 story openings across 5 LLMs, the X post author found 763 used a name from a small, predictable pool. > "In 76% of a thousand story openings (763 of the 1,000), the model reached for a name from a small, predictable pool." [x.com](https://x.com/TomLikesRobots/article/2074504762448339243), retrieved October 16, 2026 45. Receipt 45Reviewer The blocklist's most overused first names by lift: Elara, Eira, Kaelen, Thorne, Hawk, Kael, Lyra, Arin, Kaelin, Emilia, Lena, Leo, Maya, Mara, Clara, Marcus, Elias, Arthur, Liam, Eleanor, Elena. > "Most overused real first names (by lift): Elara, Eira, Kaelen, Thorne, Hawk, Kael, Lyra, Arin, Kaelin, Emilia, Lena, Leo, Maya, Mara, Clara, Marcus, Elias, Arthur, Liam, Eleanor, Elena." [x.com](https://x.com/TomLikesRobots/article/2074504762448339243), retrieved October 16, 2026 46. Receipt 46Reviewer The blocklist's most overused surnames: Blackwood, Windsor, Mayfield, Thorne, Wellington, Grey, Chen, Patel, Vance, Maynard. > "Most overused surnames : Blackwood, Windsor, Mayfield, Thorne, Wellington, Grey, Chen, Patel, Vance, Maynard." [x.com](https://x.com/TomLikesRobots/article/2074504762448339243), retrieved October 16, 2026 47. Receipt 47Reviewer In the X test, Gemini 2.5 Flash used Elara in 46% of stories, DeepSeek 19%, GPT-5-mini 6%, Llama 4%, Claude barely at all. > "Gemini 2.5 Flash used Elara in 46% of its stories. DeepSeek used it in 19%. GPT-5-mini used it in 6%. Llama used it in 4%. Claude barely used it at all." [x.com](https://x.com/TomLikesRobots/article/2074504762448339243), retrieved October 16, 2026 48. Receipt 48Reviewer Per the X test, each model has its own accent: Gemini says Elara; Claude says Marcus and Chen; Llama and GPT-5-mini say Emily, Rachel and Patel; DeepSeek says Leo and Arthur. > "Gemini says Elara. Claude says Marcus and Chen. Llama and GPT-5-mini say Emily, Rachel and Patel. DeepSeek says Leo and Arthur." [x.com](https://x.com/TomLikesRobots/article/2074504762448339243), retrieved October 16, 2026 49. Receipt 49Reviewer The X test found Elara in all eight genres, so the pool is narrow even for contemporary, romance, thriller or historical fiction, not just fantasy. > "Elara appeared across all eight of the genres." [x.com](https://x.com/TomLikesRobots/article/2074504762448339243), retrieved October 16, 2026 50. Receipt 50Reviewer The X test author notes the Ghost Couple's GPT-linked Elara Voss didn't appear in his fiction test; bare Elara was mostly a Gemini habit. > "But their GPT-associated Elara Voss didn't appear in this test. In fiction, bare Elara was pretty much a Gemini habit." [x.com](https://x.com/TomLikesRobots/article/2074504762448339243), retrieved October 16, 2026 51. Receipt 51Reviewer The X post suggests putting the blocklist into a system prompt, a logit bias, or a post-generation lint. > "Drop blocklist.json into a system prompt ("avoid these overused names:" [x.com](https://x.com/TomLikesRobots/article/2074504762448339243), retrieved October 16, 2026 52. Receipt 52Reviewer Naming expert Laura Wattenberg named Elara the 2025 Name of the Year, as the favorite name of AI (TODAY.com via AOL). > "She recently anointed the name the 2025 Name of the Year" [www.aol.com](https://www.aol.com/articles/ai-generated-baby-name-rolls-195619325.html), retrieved October 16, 2026 53. Receipt 53Reviewer Wattenberg explains AIs pick Elara because it has no baggage: nobody has met an Elara. > ""Because none of us have ever actually met a person named Elara, it has no baggage," she says. "We don't have an Elara we hate."" [www.aol.com](https://www.aol.com/articles/ai-generated-baby-name-rolls-195619325.html), retrieved October 16, 2026 54. Receipt 54Reviewer Wattenberg says AIs are trained on human content and then trained to please us. > ""The AIs are trained on human-generated content, and then they're trained to please us," she explains." [www.aol.com](https://www.aol.com/articles/ai-generated-baby-name-rolls-195619325.html), retrieved October 16, 2026 55. Receipt 55Reviewer In Greek mythology Elara was a lover of Zeus; the name means "hazelnut" or "spear" per Nameberry, and it has never cracked the US top 1,000 baby names. > "In Greek mythology, Elara was a lover of Zeus. The name is of Greek origin and means "hazelnut" or "spear,"" [www.aol.com](https://www.aol.com/articles/ai-generated-baby-name-rolls-195619325.html), retrieved October 16, 2026 56. Receipt 56Reviewer A teacher added a rule that a main character named Elara would cost students 99 points, as the name had become a giveaway of AI use. > "If the main character was named Elara, the student would lose 99 points." [www.aol.com](https://www.aol.com/articles/ai-generated-baby-name-rolls-195619325.html), retrieved October 16, 2026 57. Receipt 57Forum A FictionLab user on Reddit says they always name their own characters to avoid AI names. > "this is literally why I always come up with the names for my characters myself, as well as everything else." [www.reddit.com](https://www.reddit.com/r/FictionLab/comments/1tmzgf4/elara_voss_needs_to_stop/), retrieved October 16, 2026 58. Receipt 58Forum A user on Reddit got around stock names by asking Gemini for 100 names, which it sorted into categories, then picking one from each category. > "I asked gemini to give me a 100 of them, which he sorted into a few categories on his own, I asked to pick one from each category" [www.reddit.com](https://www.reddit.com/r/FictionLab/comments/1tmzgf4/elara_voss_needs_to_stop/), retrieved October 16, 2026 FAQ ## Questions people ask ### Who is Elara? In Greek mythology, Elara was a lover of Zeus. One name guide says the name means "hazelnut" or "spear," and it has never cracked the US top 1,000 baby names. The Elara most people search for is the AI default. Naming expert Laura Wattenberg made Elara her 2025 Name of the Year as the favorite name of AI. ### What is the story of Elara? The myth is short: Elara was a lover of Zeus. The AI Elara follows a template. She is usually Dr. Elara Voss, a physicist or other scientist leading a sci-fi plot. In tavern roleplay she is the mysterious cloaked figure. The earliest Dr. Elara Voss found online is an August 2023 character sketch of a physicist and AI researcher. ### Why do AI love the name Elara? Models default to high-probability names, and Elara has no baggage because nobody has met one, says naming expert Laura Wattenberg. The root cause is disputed. Some accounts credit probability and overlapping training data. Cornell researchers say training-data prevalence can't explain the pattern, and they speculate that alignment training steered models there. ### Which AI model uses the name Elara Voss? A Samsung and University of Warsaw preprint pins Elara Voss to GPT, with no fixed partner name. A separate test of 1,000 story openings never saw Elara Voss. It found the first name Elara mostly in Gemini 2.5 Flash, at 46% of stories, against 6% for GPT-5-mini. ### Who is Elias Thorne? Elias Thorne is a stock AI character name that ChatGPT, Gemini and Claude all reach for in fiction. He is usually a lighthouse keeper, sometimes a clockmaker or librarian. Per Google Trends, nobody searched the name until late 2025. At least two authors using it now publish dozens of AI slop books on the Kindle store. Slop check: passedThis page went through our slop counter before publishing. [How we check our own copy](/methodology#our-copy)