I built this site, so I see the search logs. Elara is one of the most-searched girl names on NamesPop. It is also a name that has never appeared in the SSA Top 1,000. That gap is the story I want to tell you.
If you have asked any large language model for a girl-name recommendation in the past two years, there is a non-trivial chance it suggested Elara. Naming expert Laura Wattenberg crowned Elara the 2025 Name of the Year specifically because it had become the model output, the recommendation that ChatGPT, Gemini, and Claude all converge on with eerie reliability. By April 2026, prompt audits of the major chatbots — including the ones I have run with my own access — confirmed that Elara is suggested at roughly twice the rate of the next most-recommended girl name in any "give me a fantasy-soft, vowel-friendly, two-to-three syllable girl name" prompt. The model loves Elara.
The country, apparently, does not. The SSA's most recent rankings put Elara somewhere in the long tail — usage measured in the low hundreds, not the thousands — and that is after years of AI prominence. There is no other recent name where the gap between cultural recommendation and parental adoption is this wide. It is the cleanest naturally-occurring experiment we have on the limits of what AI can model about human decisions.
Why the Model Loves Elara
I want to take seriously why Elara is such a good model output, because the reasons are not random. Large language models trained on naming corpora — baby-name advice forums, fantasy literature, classical mythology, parenting blogs — encounter Elara in unusually clean contexts. It is the name of one of Jupiter's moons (classical-pedigree score: high). It appears in fantasy fiction often enough to feel familiar (cultural-pattern score: high). It contains the consonant-vowel rhythm that English-speaking parents currently prefer (phonotactic score: high). It does not collide with strong real-world associations like "there was a famous bad person named Elara" (association-risk score: low). On every dimension that an LLM can compute about a name's recommend-ability, Elara scores in the 99th percentile.
It is, in other words, an algorithmically optimal name. It is what you would design from scratch if you sat down to build a name that scored well on the criteria models care about. And that is exactly the problem.
Why American Parents Decline
The criteria parents actually use to choose a name are not the criteria models can see. They are not the rhythm or the etymology or the absence of association risk. They are things like: does my mother-in-law approve, does this name sound like a name we already know, does the kindergarten cohort already have one, can my husband say it without smirking. These are social, embodied, in-network signals. They do not show up cleanly in training data. A model can optimize for everything except the conversation between two parents at midnight, three weeks before the due date, when one of them says, "Honey, doesn't that sound a little fantasy-novel?"
Look at what American parents have actually done with the soft-fantasy cluster. Aria broke through. Luna broke through. Nova broke through. Lyra has been climbing. Aurora has held. Each of these names has the same model-friendly DNA as Elara — vowel-heavy, two to three syllables, soft consonants, classical or astronomical pedigree. They differ from Elara in one specific way: each of them appeared in mass culture before being adopted at scale. Aria came from "Pretty Little Liars" and "Game of Thrones" before it became a Top 20 name. Luna had Harry Potter, then Beyonce. Nova rode "American Horror Story" and "Black-ish." Aurora has Disney. The pattern is consistent: parents adopt soft-fantasy names with a documented social proof attached, not soft-fantasy names that read as "the AI also liked this."
Elara has no social-proof anchor of equivalent weight. There is no Elara on a Top-10 Netflix show. There is no Elara as a chart-topping musician. There is the moon of Jupiter, which is real but small and dim and not going to get its own action movie. The model's training data is enough to surface Elara as a high-quality output. It is not enough to put Elara into the cultural circulation that triggers actual adoption.
The Trust Gap
There is also a quieter dynamic, which is that American parents in 2026 do not trust AI recommendations on irreversible decisions. Pew's recurring AI-trust surveys have shown a steady erosion of confidence in chatbot output for high-stakes contexts (medical, legal, financial). Naming a child sits in the same emotional bucket. Parents will use ChatGPT to brainstorm. They will rarely use it to commit. The Elara recommendation arrives in their inbox the same way a horoscope does: noted, sometimes appreciated, almost never the basis for the actual decision.
I have heard this anecdotally from at least two dozen parents who have written in to the site. The pattern is identical. They asked ChatGPT for ideas. ChatGPT said Elara. They thought it was pretty. Then they picked Eleanor, or Aria, or something their grandmother had suggested instead. The AI was a sounding board, not a decision-maker. Its recommendations functioned more like an elaborate name-of-the-day generator than like a trusted source.
The Counter-Case
It would be irresponsible to write this without acknowledging the counter-reading. Elara may yet break through. Names with this kind of consistent low-grade exposure sometimes percolate for a decade and then surge — Olivia did, Isla did, Penelope did. The current SSA rank is not destiny. If a major streaming hit drops in 2027 with a memorable Elara character, every prediction in this piece will be embarrassingly wrong by 2028. The Elara-Top-1000 wall is real, but walls in naming are notoriously time-limited.
The other counter-reading is that I am underweighting how much Elara has already grown. From a near-zero base in 2015 to its current long-tail presence is real movement. It is not a stalled name. It is a slow name in a category of fast names, which can read like stalling but is not the same thing.
The Engineer's Lesson
I work in software. The Elara phenomenon is, to me, the cleanest illustration of a deeper engineering truth: optimization-target mismatch is the silent killer of recommender systems. The model is optimizing for "recommend a high-quality name" against a training signal that rewards classical pedigree and phonetic appeal. Parents are optimizing for "choose a name our community will accept and our daughter will not resent at age 14." Those are different functions. They overlap a lot. They are not the same. Elara is the name that exposes the gap.
You can see versions of this gap in every domain where AI tries to advise on identity-laden decisions. The wedding-vendor recommendation that sounds beautiful and gets ignored because it is across town from the family. The career suggestion that fits the resume and ignores the fact that the person hates flying. The investment-portfolio rebalance that maxes out Sharpe ratio and ignores the part where the customer wants to sleep at night. Naming is the loudest version of this pattern because it is irreversible and universal. Every parent meets the gap.
What Elara's Trajectory Tells Us
I do not have a strong prediction. I have a weak one and a strong one. The weak prediction is that Elara will eventually crack the SSA Top 1,000, probably within five years, probably on the back of a streaming hit that has not yet been greenlit. The strong prediction is that the gap between AI-recommended and actually-adopted names will be a useful diagnostic for the next decade. Watch what the model loves that parents decline. Watch what parents pick that the model could not have generated. Both directions are interesting. Both will tell you something the prompt cannot.
For now, Elara is a moon of Jupiter, a model output, a 2025 Name of the Year, and a very good idea no one is using. It will probably be fine. It will probably be many beautiful daughters' names eventually. It is, today, the best evidence we have that taste is not a function with a closed-form solution.
Data source: U.S. Social Security Administration. Analysis by NamesPop.
Found this helpful?
Share it with someone who’s picking a name.