Your buyer asks AI who to buy from. Three names come back.

The baseline is same day. Your first run tells you which questions name you, which name a competitor instead, what each engine states wrongly about you and which sources it leaned on, and that first honest read is usually the most valuable thing you get in week one. Movement is slower, and we will not quote you a number nobody has measured. Engines re-crawl and re-index on their own schedule, so plan on weeks before a published fix reaches answers, and a quarter before the trend is something you can report.

About an hour, if you run it the way most teams do. The tracking looks after itself: every prompt re-runs daily on every engine your plan covers without anyone pressing anything. Your job is to read what moved, review the agent's queue, approve or reject the one or two proposals worth doing, and pass the drafted content to whoever publishes. Nothing reaches your site without your approval, by design. The work that genuinely takes time is the content and outreach behind an approved fix, and that lands on the team you already have.

Three sources seed the list: your domain and category, the questions your named competitors already appear in, and Query Fan-Out, which mines the answers we have captured for the follow-ups buyers ask next. You can add, edit, tag or remove any of them, and you should. Questions lifted from sales calls, support tickets and Search Console are often the highest-signal ones you will ever track. Your own prompts count against the same allowance as suggested ones, 100 on Starter, 300 on Professional, a custom volume on Enterprise, so the choice is about relevance, not cost.

Nobody can size it precisely, us included. No engine publishes how often a question is asked of it, so every volume figure in this market, ours among them, is derived from adjacent search demand and is useful for ranking a shortlist rather than as a measured count. The better test is whether your buyers already open an assistant before they draw one up. In a thin category you will also see no anonymised benchmark, because we withhold one rather than build it from three brands; named competitors and share of voice give you the comparison instead.

All four use it, though what you lean on differs. B2B works off the unbranded and comparison questions where a shortlist gets drawn. E-commerce leans on AI Shopping tracking and the review sources engines cite, both on Professional. Local businesses can track local-intent questions from a named city rather than a whole country. Agencies get unlimited brands and white-label reports on every plan. It suits you least where buyers do not consult assistants about your category yet, or where you need complete click-level attribution, which nobody in this market can honestly offer.

Alongside, not instead, and it plugs into habits they already have. Prompt research feeds the content backlog the way keyword research does, the citation-gap report feeds outreach the way a link-gap report does, and the GEO page auditor scores a URL against what earns a citation rather than a ranking. Seats are unlimited on every plan, with Owner, Admin and Member roles, so an agency can work in the same workspace without a per-login bill or publishing rights you did not intend to hand over. Exports are on every plan; the REST API, CLI and Looker Studio connector are on Professional.

Yes. Brands are unlimited on every plan, so a group or an agency is never billed per logo. The meter is the prompt allowance, and it is shared across brands rather than granted to each: ten brands at ten prompts each and one brand at a hundred consume exactly the same capacity. Markets work the same way, since a question asked in another market is another prompt, and prompts can be tagged by region so you read each market separately instead of averaging them into one flattering number. Every brand gets its own live, white-label report link.

There is no lever that forces an engine to cite you, and anyone selling one is guessing. What you can influence is the evidence: your own pages and schema, which you control outright, and the third-party sources engines actually reach for in your category. llms.txt is an emerging convention rather than a standard any engine must honour, and we do not claim it guarantees anything. Engines also change behaviour without notice and can restrict automated access, so when one is unavailable we record that day as missing rather than estimating over the gap, and we may adjust plan coverage with notice.

Your tracked prompts go to the engine's operator, which is inherent in asking an engine a question. Anything you connect, such as Google Analytics or Search Console, is read only and limited to the scope you grant. We do not sell customer data and we do not let any third party train foundation models on it. We do train our own internal models, on aggregated and de-identified data only, never on identifiable customer content: Starter and Professional contribute on that basis, and Enterprise customers can opt out in their order form. Our sub-processors are listed in the Data Processing Agreement.

You cancel from account settings. It takes effect at the end of the current billing period and you keep access until then. Fees are not refundable, there are no credits for partial periods, and an annual term runs to the end of that term. You can export at any time while the account is active, as CSV, Excel or JSON, including the captured answer text rather than only the scores, and we make your data available for thirty days after termination on request. After that it is deleted or irreversibly de-identified. Ask Support@airankly.io, Monday to Friday, 8:00 AM to 6:00 PM CT.