Domain availability is a verification problem, not a suggestion problem

Ask any AI for startup names and you will get fifty. Perhaps five are good. Perhaps one has a domain you can actually register, and you will find that out by copying each into a registrar, one at a time, until your enthusiasm runs out.

The generating was never the hard part. The verifying is.

What we check, and how

Every candidate domain is checked against RDAP — the registries' own structured record service, the successor to WHOIS. A 404 means no record exists, which is the only positive signal that a domain is free. Anything else means it is taken, or that we could not tell.

Names whose domains cannot be positively confirmed as available are discarded before you ever see them. You are looking at a shorter list than the model produced, and that is the point.

The failure mode that matters

Registries rate limit. Early on, our checker fired every lookup at once — up to seventy for a single naming round — and the registry answered with 429s. Those were recorded as 'unknown', unknown names were filtered out, and the user saw an empty list.

An empty list reads as 'nothing is available'. It actually meant 'we were throttled'. Those are completely different pieces of information and the interface treated them identically.

Now throttling is retried with backoff, lookups are capped at six at a time, results are cached for a day, and if too many checks still fail we tell you the registry did not answer rather than implying your ideas are all taken.

Why this is worth the effort

A name you cannot register is not a name. Verifying first turns a brainstorming toy into something you can make a decision with — and it is the difference between a list of words and a shortlist of options.