A document you can act on, not a score.

The report answers where you stand today. Monitoring answers whether you held - and nothing else does: 134 of the 253 names we saw did not survive a re-read.

The ranked plan comes with the report. The next rung up is us doing the work: we make the fixes on your site ourselves.

What the free scan gives you

• One real customer question for your business in your city, asked on the consumer app, with the names it gives back.
• Six mechanical signals decide whether an AI engine can pull your page into an answer. We grade each one.
• A letter grade on those six signals for your website.
• A public page carrying that grade at a fixed address, refreshed while you fix things. Grades can be refreshed once an hour.
• A preview scan across three AI services, sent to your inbox.

It says nothing about the business itself, and it is not a measure of whether the engines recommend it. Being technically readable does not mean the AIs recommend you.

What is in it

• About 60 questions your customers really ask, asked of six AI services – roughly 360 answers in one run
• Three different outcomes per answer: recommended, mentioned, or your own pages cited as a source
• The full leaderboard of your market, with your row in it – including businesses you never listed
• A battle card for every business that beats you: what the answers credit them for, and which services never pick them
• The questions you lose, ranked, with the name given instead of yours and the sources feeding that answer
• The questions nobody owns yet – unclaimed ground, cheaper to take than ground you have to prise away
• Your site read page by page: structured data, llms.txt, crawlability, titles, and whether your content is shaped like an answer
• Every source behind the answers, sorted by what you can actually do about it – and where to publish next
• A confidence range on every rate, with the formula published – counts printed before percentages
• A frozen baseline – the same questions, the same city, the same date, saved so a later reading compares like with like

The baseline is yours once the report is delivered. Monitoring re-reads the same frozen questions later, so the second reading is comparable to the first - and nothing about the report depends on buying it.

Where you stand

The report opens with the number, and with everything the number is made of.

• Being named is not being endorsed. We score every answer three ways: recommended (the engine steers the buyer to you), mentioned (you appear at all), and cited (your own pages are used as a source). The gaps between the three are usually where the actionable work hides.
• Share of recommendations – Your recommendations divided by every recommendation counted across the whole field, times 100. Plain arithmetic on the numbers already in your report. Nothing is weighted, scored or modelled.
• Recommendation rate – Answers in which an engine names you as a pick, divided by answers we successfully analysed. Counted per engine and pooled. We report the count alongside the percentage, always.
• One score out of 100, built from those rates, your share of the market’s recommendations and how you are spoken about.
• How concentrated your market is – a field split between many names, a field fought over by a few, or a field one name already holds – and how many names hold half of all the recommendations.
• A service-by-business matrix: how many recommendations each business took, in each of the six services.

Who wins, and why

• Every name the AI services recommend across roughly 360 live answers (about 60 buyer questions, all six services), ranked, with your row highlighted, including competitors you never listed.
• Where each rival is strong, where they are weak against the whole market, which engines never pick them, and the sources powering their picks.
• Being listed and being the pick are different outcomes. We report how each business is treated, praised, listed flat, or warned against, and why: price, reliability, credentials, availability.
• The attributes the answers hand out credit for – quality, choice, price, credentials, experience, availability – and which of them the market gives you.

The questions you lose, and what it takes to win them.

Every scan ends in a ranked battle plan. Each opening is one real buyer question, the name the engines give instead of yours, and the sources feeding that answer, so the next move is obvious rather than inferred.

• The question – Asked the way your buyers actually phrase it, not keyword-stuffed. Weighted by where it sits in the buying decision.
• Who wins it, and why – The competitor the engines name instead of you, what the answers credit them for, and how many engines agree.
• What it takes – The sources that answer is built from, ranked by whether you can get listed, reviewed, published or nothing at all.
• Every question is tagged by buyer intent: researching, comparing, ready to buy. Many businesses win the researchers and vanish for everyone about to spend. One averaged score hides that.
• Questions are classified by type, best-of, comparison, price, trust, so the finding is “strong on best-of, invisible on price” instead of one number you cannot act on.

Questions no one owns are listed too. Most AI demand sits in categories with no established winner, and an unowned question is cheaper to take than one you have to prise away.

What the answers are built from

• The engines do not search your buyers’ questions. They compose their own, and we record every one. No keyword tool shows these, because no human typed them.
• Connect Search Console and we match the engines’ sub-searches against what your site ranks for, so you see whether you are even in the pool the answer is built from. That connection is optional, yours to make and yours to withdraw.
• Every source behind the answers with an action against it – your own pages, a rival’s site, a community you can take part in, an editor you can pitch, or nothing anyone can do – and how long a result takes: a day, a week, three months, over a year.
• Where to publish, in order: the sites the answers actually quote in your market.
• The exact pages where your competitors appear and you do not – the URLs, not a summary.
• The signals outside the AI services that feed them: Wikipedia, Google autocomplete, and what people search alongside your trade.

Your site, page by page

Your own website is audited in every run – yes, with the evidence quoted.

• Structured data – JSON-LD schema that tells engines what your page is and who you are.
• llms.txt – The emerging file that hands AI crawlers a clean map of your site.
• Crawlable – No noindex or blanket robots block quietly telling engines to leave.
• Title and description – Shaped so engines quote them cleanly instead of improvising.
• Viewport and page quality – The basic signals both crawlers and engines read.
• Answer-shaped content – Question-shaped headings and FAQ structure engines assemble answers from.

Every page of your site gets the same six checks, and every page that gives an assistant nothing to quote is named.

How sure the numbers are

• Every number comes with an honest plus-or-minus, and we publish the formula behind it.
• Every rate carries a 95% confidence range. Answers are not independent, since the same question is put to every engine, so we account for how clustered your results are and widen the range rather than report a falsely narrow one.
• Numbers are reported as plain counts first, not dressed up as percentages, because counts are checkable and percentages hide sample sizes.
• Every answer in your report was obtained through the official API as an unpersonalized request. A signed-in person with chat history can be answered differently. Your report states this on every run.
• From the second run on, the report shows what moved since the last one.

What comes with it

• If our questions missed how your customers really ask, tell us. One free re-run with your own wording comes with every scan.
• One time, no subscription. Full refund if we cannot deliver.
• In your inbox within the hour.
• Generated from your own data: exact domains to get onto, pages to create, fixes to make, each with an honest effort tag.
• Every action in the plan carries what we expect it to change.
• Every answer is stored word for word. Click a claim in your report and read exactly what the engine said.

How it is made

The questions are the ones customers put to an assistant when they look for a business like yours in a city like yours - asked of ChatGPT, Gemini, Perplexity, Claude, DeepSeek and Mistral, one question per fresh conversation, the whole answer stored as the assistant wrote it. Nothing in the report is paraphrased from memory: every name it says an assistant gave is in the raw answer beside it, and every claim about your site points at the page it was read from. Every line above is a thing you can open and check in the document itself, not a number we computed for you. The measurement log lists every run behind this site, with its sample size.

The free scan reads the consumer app the way your customer sees it, signed in, one question at a time. The paid report asks six AI services through their official interfaces, logged out, in one deep pass. The two can differ, and you get both.

Where to start

The free scan asks one of those questions for your business and tells you whether the paid report is worth it - sometimes the answer is no, and then you have spent nothing. The prices are on the pricing page: one exact price each, no quotes. What this is not says where the measurement stops.