Exactly how we measure — and what every number means.

AI answers change from one run to the next, so we never give you a fake “rank.” We measure how often AI recommends you across real customer questions, and we show our work. Here's the whole method, in plain English.

What a “check” is

A check is one real customer question asked to one AI assistant. We ask the kinds of questions your customers actually type — “best service company near me,” “who should I hire near me” — across the AI assistants people use: ChatGPT, Perplexity, Gemini, Google AI Overviews, Copilot, and Google AI Mode. Higher plans check more of them, and checks run every day, so each plan has its own honest count:

• Watch: up to 19 questions × 4 AI assistants, checked every day — about 2,280 checks a month when your rows are full and we hold a question set for your line of work.
• Guided Fix: up to 30 questions × 5 AI assistants, checked every day — about 4,500 checks a month when your rows are full and we hold a question set for your line of work (adds Copilot).
• Done-With-You: up to 42 questions × 6 AI assistants, checked every day — about 7,560 checks a month when your rows are full and we hold a question set for your line of work (adds Google AI Mode).
• Done-For-You: up to 50 questions × 6 AI assistants, checked every day — about 9,000 checks a month when your rows are full and we hold a question set for your line of work, with our team running everything for you.
• Your free scan is a smaller check — enough to show the problem, not the full picture.

Every paid plan asks the same 14 questions about your trade, and they never change, because a number is only comparable to last week if the question was the same. What a bigger plan adds is room for your own services: you type what you sell, and your plan holds more of them. So Watch is 14 + 5 of yours, Guided Fix is 14 + 16, Done-With-You is 14 + 28, and Done-For-You is 14 + 36. Those rows are yours to fill: a row you leave empty is simply not asked — we never put a question of our own in it. The 14 about your line of work run once we can place your trade; where we cannot, your own rows are the whole of your week and we say so rather than counting a question nobody asked.

Asked every day — and counted inside your number. Each service you type becomes a question of its own, and we ask the AI assistants about it on every check, same as the rest. What comes back is inside your “appeared in N of M” and inside the week-to-week arrow, because your number is out of everything your plan asks — the 14 about your trade, where we hold a set for your line of work, plus the rows you wrote. Your own rows also keep their own breakdown, row by row, so you can see how each service did on its own. That breakdown carries no arrow of its own: a list you edit is not comparable to last week's list.

Coming soon: Meta AI and Amazon Rufus — not checked yet. We'll add them to your count the moment they're live.

The words we use — four different things

We never blend these into one vanity score. Each means something specific:

• Recommended — AI actively told the customer to pick you. This is the number that matters most.
• Mentioned — AI named you, but didn't actually recommend you.
• Cited — AI used your website as a source for its answer.
• Fact accuracy — whether AI got your details right (hours, services, area) or stated something wrong.

How we count

• Planned — your number is out of the full count of every question your plan asks — the 14 about your line of work plus the ones you add yourself — times the AI assistants on your plan, times the checks in the week (19 questions across 4 AI assistants, checked every day, is 532 in a full Watch week). We don't quietly shrink it. If you haven't filled in all your own questions, we count only the ones we really asked — never the ones you could have added. Your own services also keep their own breakdown, further up this page.
• Unavailable — if an AI doesn't answer (an outage or timeout), we count that check as “did not appear” rather than dropping it — a conservative choice that never inflates your result. We always show how many actually completed.
• Named — the AI mentioned your business in the answer at all, even just in a list.
• Recommended — the AI actually told the customer to pick you: recommended you, recommended you with a reason, or pointed at your website. Every recommendation is also a mention; most mentions are not recommendations.
• So “recommended in 18 of 532 checks” means: across the 532 checks that make up your number for the week, an AI told a customer to pick you in 18. If a week was measured before we counted recommendations separately, we say “named” instead of “recommended” — we don't restate an old number as a new one.

We never call this a “rank.” Research shows AI ordering is close to random between runs — a number like “you're #3 in AI” would be made up. Frequency is the honest signal.

Confidence

More checks, repeated over more days, mean more confidence. Every number is shown with the count of checks it rests on, and a week that rests on too few of them is marked “few checks” right beside it — and where a week is that thin we hold back the up-or-down comparison rather than call a wobble progress. We never show a fake-precise percentage like “73.4% confidence.” A single week is a snapshot; confidence grows as we re-check over several weeks.

What we don't do

• No guarantees — no honest tool can promise what an AI will say, because AI answers change over time.
• No invented numbers, no fake badges, no fake reviews.
• We never touch your site. On Watch and Guided Fix you (or your web person) apply the change yourself; where our team does it for you, nothing goes live until you approve it.