Free AI visibility checkers: what a one-off scan can and cannot tell you

Yes, and it takes an afternoon rather than a tool. Open the model and ask the questions your buyers would ask, without your brand name in the prompt. Note whether you appear and where in the list. Run each prompt a few times, because one run only tells you about that run. Write down the date and the model version you used, so the note is still worth something next quarter. That is the honest baseline before you consider paying anyone.

It is accurate about what it saw, which is a narrower claim than it sounds. The real question is whether the sample represents your category: how many prompts, how many repeats of each, which providers, and which model version. If a checker does not disclose those four things, treat its number as directional rather than measured. Seenby attaches its methodology caveat to every computed score for the same reason, and surfaces the disclosure on every score display.

Two separate mechanisms are at work. Within one model, repeated answers to the same prompt vary, so a small sample moves on its own. Between models, providers ship new versions, and a new version can reorder which brands it reaches for. Neither is a fault in the checker. Together they mean a difference between two one-off scans is not evidence of anything. Separating the two is a measurement design problem: hold the prompt set constant, repeat it on a schedule, and record the model version behind each answer.

No. There is no free tier and no trial period. The product is in waitlist mode, so signing up today puts you in line rather than into the dashboard. The pricing page lists exactly what the plan includes.

Three things. Whether you are named at all in the obvious category prompts, which competitors the model reaches for when you are not, and how it characterises you when it does mention you. That is enough to decide whether AI visibility deserves a budget line this quarter. It is not enough for a board slide, because you have no baseline to compare against and no way to attribute a change to the market rather than the model.

The count matters less than holding the set constant. Five prompts run once tell you about five prompts on one day. The same fifty prompts run every week tell you about direction. Seenby generates 50 to 100 prompts from your brand description and named competitors at setup, freezes them, and runs that same set each week, so the trend reflects the market rather than changing questions. If you are sampling by hand, pick a set you can repeat and resist editing it between runs.

It varies, and coverage decides what the number means, so ask before you trust it. Ask which providers are queried, which model versions, and whether web search was enabled during the query. If a checker will not name those, treat its coverage as unknown rather than broad. Seenby reads two providers, OpenAI and Perplexity, and tags every stored answer with its model version, so a provider change is visible rather than buried in the score.

Do not over-read one reading. Re-run the same prompts over a few days and note which competitors appear in your place. A consistent absence across repeats and providers is a real gap; a single absence is noise. If the gap holds, ask which prompts you are missing from, since that points at content and coverage work rather than a generic optimisation project. Then decide whether the gap is worth measuring continuously, which is a purchasing question. Our guide on how to choose an AI visibility tool lists the questions worth putting to any vendor, including us.