As the internet becomes increasingly saturated with low-effort, machine-generated content, distinct patterns and digital fingerprints have long been caught by keen eyes, as when large language models are tasked with churning out endless blogs, sci-fi stories, and expert personas from thin, underspecified prompts, they do not, sadly, produce infinite variety. Instead, they appear to constructively default to repetition, and the most glaring and human-readable symptom of this homogenization is the creation of the promptonym, a fictional proper name that models repeatedly spit out when asked to invent a character.

A promptonym is a structural artifact of how generative AI behaves at scale. The word itself is a fairly newly coined portmanteau combining “prompt” and the suffix “-onym” (derived from the Greek term for “name,” as seen in pseudonym or synonym), that we ourselves used in our previous articles, and while, on the micro-scale, individual users might think the model is being creative, the aggregate reality of AI content has painfully shown, now supported by analysis from researcher Michał Brzozowski, to be repeatedly settling on the exact same names, or small families of closely related variants, to populate their synthetic worlds. Most notably, the researchers identified that the promptonyms are model-family-specific, with Claude being enamoured with Elena Vasquez, Marcus Chen, and Amara Okafor; Gemini, on the other hand, was caught using none other than Aris Thorne, plus one Lena Petrova. ChatGPT was identified to have an affinity to Elara Voss.

The undisputed poster child of the promptonym phenomenon is Elara Voss. In August 2025, technology writer Max Read coined the term “promptonyms” specifically to describe the unavoidable recurrence of Elara Voss (and variants like Elena Voss, Elara Vex, and Elias Vance) across AI-generated science-fiction and fantasy slop. Read noted that simple web searches for the name now pull up massive quantities of AI-associated material, with examples dating back to 2023. Another prominent promptonym is Aris Thorne. As we ourselves reported, many, many, people have reported seeing this name constantly when prompting LLMs to create personas, and the issue of discovering vast seas of unrelated AI-generated content heavily populated by characters named things like Aris Thorne only seems to have gotten worse recently. Not just Aris Thorne the polar researcher! In the last six months since we published the original article, now observe Aris Thorne the panel engineering specialist, CEO of the Delaware-based solvent producing company ‘Tidegate Solutions’, and a slightly deranged foray into gaming. One simply cannot wait what will come next.

God help us all.

LLMs generate text based on learned statistical relationships, and when a prompt asks for a scientist but provides no name, the model must guess based on its training distribution. A promptonym thus represents a massive, conspicuous local concentration in that distribution, for example – Max Read has shown that datasets on the platform Kaggle may have influenced several of these promptonyms. Instead of names varying evenly across a vast space of possibilities, a few specific token combinations end up being disproportionately favored by the model’s training data, etc., and end up invading the internet with a bajillion Elara Vosses, Aris and Marcus Thornes, and Alistair Finches.

The repetitive nature of AI slop isn’t limited to creative writing. Research by Padmakumar and He demonstrated that participants writing with InstructGPT produced essays with significantly lower lexical and content diversity than both control groups and those using older base models. An isolated article featuring a researcher named “Elara Voss” might look entirely passable, but when viewed collectively across the internet, the promptonym exposes the limitations of AI creativity. (With, to admit, a lot of comedy.) Naming, particularly, does really feel like a highly restricted deck of cards and the most clear example of the phenomenon.

Ultimately, “promptonym” is an informal, descriptive label born from internet observation, not an established concept with an agreed technical definition. You will not find it in the taxonomies of prompt engineering surveys, though it does have a listing on the AI-generated Wikipedia, Grokipedia. However, promptonyms put a recognizable and overarching face, (usually Elara’s or Aris’s), on the creeping AI-ificitaion of the internet, serving as the ultimate telltale sign that you are reading AI content, or at the very least, content whose characters were named by AI.