How AI Name Generators Actually Work (And Where They Fail)

An AI name generator sends your description to a large language model along with instructions about the kind of name you want, and the better ones then filter what comes back in code before you see it. Some tools sold as "AI-powered" do something older: they splice your keyword onto a fixed list of prefixes and suffixes. The two produce very different output, and deserve very different amounts of trust.

Below: how each approach works, what changed when we ran the same brief at Low and at High creativity, and four things no model can do for you however good it is.

The Two Architectures

Pattern recombination

This is fast and nearly free to run, and even when a tool shuffles its lists, every result is still a splice of the same fragments. Its weakness is that it has no concept of meaning. It cannot tell that Coffeeora sounds like a medication, because it is not modelling sound or connotation at all, only joining strings. Output quality is capped by the fragment list, which is why these tools feel repetitive after two pages.

Language model generation

The newer approach sends your description to a large language model with instructions about the kind of name wanted. A language model is trained on what Google's machine-learning course calls "a phenomenal amount of training data (text)" and learns to predict the tokens that come next; the same course describes LLMs as "essentially autocomplete mechanisms" (Google, Introduction to LLMs). That training is why a model carries a statistical sense of what brand names in a sector look like, what a word root suggests, and how names are built in other languages.

It makes a real difference for one reason: the model responds to context. "Coffee subscription for offices" and "third-wave coffee roaster in Lisbon" produce recognisably different names. A recombination engine produces the same list for both, because it only ever saw the word "coffee".

The weakness is the mirror image. A model samples its output, so the same prompt gives different results each time; OpenAI's documentation says plainly that its chat models are "non-deterministic by default" (OpenAI, reproducible outputs). Left unchecked, a model will also confidently produce names nobody can pronounce.

Why Prompting Alone Is Not Enough

This is the part that surprises people who assume the prompt does all the work. Instructions to a language model are followed partially, not absolutely. Telling a model "do not produce names ending in -ora, -ova or -ia" makes it do so less often; it does not stop it.

Anything that has to be reliably true needs a filter in code, after generation, that discards output failing the rule. In our own case, naming banned endings in the prompt took the rate of Latinate tails from roughly 75% down to 22%. A code gate that rejects them outright took it to zero. The prompt got most of the way; only the gate finished the job.

This is worth probing when you compare tools. Generate thirty names and look for ones you cannot pronounce. If unpronounceable output reaches you at all, the tool is relying on the prompt and showing you whatever the model returned.

Four Things No Model Can Do

1. Tell you whether the domain is free. A language model has no live view of the registry. Availability has to come from zone-file data, DNS or RDAP at the moment you ask. A model asserting a domain is available is guessing, and it will sound certain while doing it.
2. Clear a trademark. Trademark rights are specific to a jurisdiction and to classes of goods, and the registers keep changing as marks are filed and lapse. Clearing a name takes a live query against the actual register, not a model's recollection of one. No generator's output is legal clearance.
3. Know whether the name is already in use locally. The corner shop two streets away trading under your candidate name may appear in no training data and in no register a tool checks.
4. Decide for you. A model can produce two hundred plausible names. It has no access to what you want the business to feel like in five years, and that is the actual decision. Generators shorten the list; they do not choose.

Why Two Tools Disagree on the Same Keyword

Four variables account for most of the difference, and none of them are visible from the outside:

• Which model. In our experience, models have characteristic habits: some drift toward Latinate endings, some toward two-noun compounds, some reuse the same handful of roots.
• Temperature. The randomness setting. OpenAI's API reference puts it this way: "Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic" (OpenAI API reference). Not every model still takes the setting: Google's documentation says Gemini 4 and Gemini 3.6+ models ignore custom temperature values (Gemini prompting guide), so on those a creativity slider can only change the prompt. Tools that generate in several passes at different temperatures cover more ground than a single pass.
• The prompt. Invisible to you and the single largest lever. Wording changes here are also unusually hard to evaluate: of ten confident rewrites we tested head-to-head against the existing wording, two won and eight lost.
• What is filtered after. Whether anything checks pronounceability, dictionary membership, banned endings or domain availability before the name reaches your screen.

One Brief at Low and High Creativity: What Changed

On 15 October 2026 we gave Namilio the same brief twice, "a language-learning app for kids", with only the Brandable style selected: once at Low creativity and once at High. In Namilio that setting moves two levers at once. It sets the temperatures of the three parallel passes (0.25, 0.4 and 0.65 at Low; 0.85, 1.0 and 1.2 at High, a little higher if a pass has to top up), and it swaps one line of the prompt: Low tells the model to build every name from the keywords, High tells it to name the idea rather than the sentence. What follows is the effect of both together, not of temperature alone. We took the first 12 names from each run and checked every .com against Verisign's registry (RDAP), where "Available" means the registry had no record of the name.

Low stayed on the obvious roots: 10 of its 12 names are built on ling, vocab, talk, speak, word or phono. High used those roots in 6 of 12 and went elsewhere for the rest: Spelltide, Mimifly, Storymint, Chattermew, Kidlet, Listenloop. Neither instruction was followed to the letter. Low's line asks for a recognisable fragment of the keywords in every name, yet none contains "learn", "kid" or "lang"; High's asks the model to leave the description's own words out, and it returned Kidlet.

What did not move was the .com: 1 of 12 was free in each run (Glimmerglot at Low, Chattermew at High), and one name, Talksy, came back in both. Higher creativity changed what the names were made of, not how many of them anyone had registered already.

How to Evaluate Any Naming Tool in Two Minutes

1. Give it context, not a keyword. Type a full sentence describing the business. If the output does not change meaningfully compared with a single word, it is recombining, not generating.
2. Run the same input twice. Identical lists mean a fixed pattern engine. Varied lists mean either a model or a pattern engine that shuffles; the first test tells you which.
3. Look for unpronounceable names. Any that reach you suggest there is no code-level filtering behind the prompt.
4. Check whether availability is shown. And whether it covers more than .com: one extension is a narrow view of whether a name is usable.
5. Ask what happens when it is wrong. A tool that quietly returns canned names when its model is unreachable is worse than one that tells you it is down, because you cannot tell the difference from bad output.

Namilio runs 11 language-model styles plus 5 instant local ones, generates in three parallel passes at different temperatures, filters every result for pronounceability in code, and checks the extensions you selected (up to 22) before a name is displayed. Try it free and run the two-minute test above on it.

FAQ

Try It Now

Describe your business in a full sentence rather than a keyword and see how much the output changes; that alone tells you what kind of engine you are talking to. Generate names free: no signup, 16 styles, every name checked on the extensions you pick, up to 22.