How do you measure something as fuzzy as an AI answer?
We run a fixed set of real customer questions against each engine we have live, on a schedule, from a clean session with your target location set. Every answer is recorded: which businesses were named, in what order, and which sources the engine cited. Same prompts, same conditions, every time. Any single answer is noisy. The trend across hundreds of them is not. Two honest caveats. We query the engines through their APIs, with no memory and no personalisation, so the answer a signed-in person sees on their phone can differ from ours. And we ask the question a customer would type, then ask the engine to name specific businesses, because an answer with no names in it cannot be measured at all; that instruction is on every question we send, identically, which is what keeps the comparison between engines and between scans fair. That is exactly why we keep the conditions fixed: it is the trend and the comparison with your rivals that carry the meaning, not any single reply.
Can you guarantee I will be recommended?
No, and anyone promising that is selling you something they cannot deliver. The engines change their models and their sources constantly. What we can do is measure the answers under fixed conditions, tell you which domains the engines leaned on, and write the profile and the answer pages for you to publish on your own site. There is no channel that submits a page to ChatGPT or to Claude, so nobody can hand an engine your page, and nobody should tell you they can.
Does publishing a page make an engine cite me?
Not on its own, and we will not sell you that chain. Neither ChatGPT nor Claude takes a submission today: OpenAI documents no channel, and Claude searches an index that grows from pages people already open or that already rank elsewhere. A page a week old, on an address shared with every other site on the same host, has neither. That is why the copy we host is a preview and the deliverable is the file you publish on your own site, next to pages a crawler already visits. What we can promise is the measurement: the same questions, the same conditions, every time, and the list of domains the engines cited when they named someone else.
Is this just SEO with a new name?
It overlaps, but the target is different. Classic SEO optimises for a position in a ranked list of blue links. This optimises for being named inside a written answer, where there is no list and usually only two or three businesses get mentioned at all. How an engine picks those two or three is decided inside a model nobody outside it can inspect. What can be seen is which sources it cited, and that is what we record, scan after scan.
What do I actually have to do?
Connect the business once, which takes about a minute. After that the scans run on schedule on their own. The drafting is a button, not a schedule: you press it when you want a draft, and we write the Answer Profile and an answer page for each question you are losing, up to 30 of them on Growth, with every fact we could not verify marked for you to confirm. You read them, fill in the few things only you can decide, like a price or an answer in your own words, and publish with one switch. Nothing goes live until you flip it. The same page downloads as plain HTML with its structured data, to put on your own site, which is where we would put it.
What happens if I cancel?
Your hosted profile and question pages go offline and return a not-found page. Anything you had copied onto your own site stays where you put it: it is your file on your server, and it never depended on us. Every report you have ever run stays yours: export the answers and citations from any run as CSV, and print the latest visibility report to PDF from the Reports page. But the organisation drops back to the free tier, which keeps 30 days of scan history, and that rule is applied by the next clean-up pass rather than after a grace period. Export what you want to keep before you cancel. We would rather tell you that plainly than surprise you with it later.
Does it work outside English?
In English and in French today. Engines answer in whatever language the question was asked in, and they often answer differently depending on the language, so we track separate prompt sets per language: a business in Geneva can watch its position in French and in English side by side. Other languages are not offered rather than answered in English, which would measure a market nobody there is typing into.