Reading your AI Visibility results

The headline

• Visibility score, out of 100. Every answer is scored by how prominently you appear: a lead recommendation counts fully, being listed counts 0.6, a passing mention 0.3, and absent 0. The score averages every answer collected. The word under it is its band: Leader (85 and up, AI puts you first in most answers), High (70 to 84), Medium (40 to 69, named in many answers but usually not first) or Low (under 40).
• Lead recommendation in X of Y answers, and how many answers named you at all.
• Margin of error. Each question is asked once per assistant, so the page shows a range, for example "72% of answers, give or take 8 points". Read the range, not the decimal.
• Tiles: assistants asked, live answers, pages the AI read, answers given from memory (with no current sources at all), and your average place when an answer puts brands in order (1 means named first).
• How the answers lean when you are mentioned: positive, neutral, mixed or negative.

The matrix

Chapter 02 shows every question against every assistant:

• 1st, Lead recommendation: named first, or as the top pick.
• ✓, Listed: named as a real option, not the first.
• ·, In passing: mentioned briefly, not recommended.
• A dash, Absent: answered, but never mentioned you. Often the honest result.
• ?, No answer: that assistant returned nothing that day. See Why an assistant shows "No answer".

An assistant that answered fewer than 60% of the questions is flagged as low confidence.

The chapters

1. 1What AI says about you: Insights, three to five findings written from this check's own numbers (every sentence carries a figure you can find on the page); the summary in the assistants' own words; What drives how AI talks about you, a report card per theme (price and value, product quality, delivery, customer support, ease of use, trust), counting how many answers judge you good, mixed or bad on each, with their own words checked word for word; The words AI uses about you, the short phrases the answers use about you, copied word for word, most repeated first; and on later checks what changed in their wording and where you moved.
2. 2Where you stand: the matrix, wins and losses by buyer stage, Questions by topic (your questions grouped by what they are about, with how often you were named and named first), the score per assistant and every answer in full.
3. 3Who wins the answer: share of voice, who AI names instead, Share of voice by assistant (each assistant's share of answers naming you and your main rivals, with your average place in its ranked answers), How AI describes you and your rivals (a chart of how often each brand is named against how positively, then a card per brand with its tone, its average place, the themes it is judged on and the phrases used about it), a question-by-question comparison with your main rival, the citation gap (sites that fed answers where a rival won and you were absent), and Shopping answers: the product cards Google AI Mode and ChatGPT showed in their answers (product, price, shop, stars), with yours and your rivals' marked. Both are asked from your check's market, so prices and shops are local; a worldwide check is asked as from the United States. ChatGPT's cards are read from chatgpt.com as a signed-out visitor sees it, through our data supplier, so they can differ from what a signed-in user sees.
4. 4What the assistants read: What AI searched for, the web searches the assistants typed before answering (ChatGPT, Claude, Gemini, Grok and Mistral report theirs), with the words they added to your questions and the rivals they searched for; who owns the pages behind the answers (your site, reviews, news, rivals); on later checks, Sources gained and lost since your last check, the sites the assistants read for the first time, the ones they no longer read, and those they leaned on more or less, with your own site's share then and now; which of your own pages were read; Where people talk about you, the Reddit threads, forum posts, videos and other public conversations among the pages they read, by platform and Reddit community, with the assistants that cited each; how old the pages are, and the full list.
5. 5What reached your website: your own data, when you add it: the visitors each assistant sent you and the AI crawlers that read your site.
6. 6What to do next: Your to-do list, most impact first, everything the check found that you can act on in one numbered list, ranked by what was counted (assistants that cannot read your site, facts you marked wrong and questions most answers do not name you in come first; inside each level, the most answers at stake), each with its reason and a link to the detail; What to do, and when, the steps sorted into this week (anything that stops AI reading your site or repeating your facts correctly), this month (pages and fixes on your own site) and the next three months (being cited on the sites AI reads); facts the assistants state about you (see Marking claims as right or wrong), pages worth publishing, each with a draft you can ask for, recommendations, your site's AI readiness, Your site, page by page and Check a page.

Chapters with nothing to show are left out, and the numbers close up.