Only what a bot sees
We read the site from the outside, like an AI crawler: robots.txt, HTML, markup, server responses. No access to your admin or analytics is needed.
The main score is “AI visibility”, from 0 to 100. 60 points are the site’s technical readiness: 25 items in 6 areas, from bot access to how quotable your text is. 40 points are AI answers: do AI assistants name you themselves when customers ask about your field. The formula is public and the same for everyone: the same site always gets the same score, and every item comes with evidence.
How we check
We read the site from the outside, like an AI crawler: robots.txt, HTML, markup, server responses. No access to your admin or analytics is needed.
The readiness score is set by rules, no model touches it: the same site always gets the same score. AI models work only in the extended audit — they build the brand profile and questions, read the assistants' answers and write drafts.
A robots.txt line, a server response, a page address — you can see where each conclusion came from. In the full audit — for every fix.
Every item says what is there now, what to do and how to verify. Fixes are ordered by impact / effort.
Formula and grades
Readiness score = 100 × points earned / maximum of the items checked. An item we could not check drops out of both. If fewer than 60 of the possible 100 points were checked, we don't give a score, so as not to mislead.
Content items are counted across all pages read: the result is the average.
The report's main score is AI visibility, 100 points: 60 for technical readiness on the scale below (access for AI, content and trust), 40 for AI answers — whether AI assistants name you on their own (the extended audit's Presence Score). Until AI answers are checked there is no total and no grade: we show the known part, at most 60.
The grade comes from the total AI visibility and is not shown until AI answers are checked. Technical readiness has its own grade on the same thresholds: A — AI reads the site freely, B — with minor obstacles, C — errors get in the way, D — reads poorly, E — can hardly read it.
Extended audit
We ask the AI assistants (9) 20 questions a customer would ask — with their own search, in your market's language and from your region, which you confirm before the survey. 60% of the questions don't name the brand: will the assistant name you on its own. Rules, not a model, decide what counts.
ChatGPT · Gemini · Perplexity · Claude · Grok · Qwen · Kimi · Mistral Le Chat · DeepSeek
| Item | Points |
|---|---|
| Visibility: in how many answers to questions without your name the assistant named you on its own | 16 |
| Link to the site: in how many answers the assistant cited your site as a source | 10 |
| Position: where you are in the lists — first or at the end | 8 |
| Sentiment: whether you are spoken of positively, neutrally or negatively | 6 |
26 checks · 6 areas
Six areas, 26 items. An area's weight is how much it contributes out of 100.
robots.txt and the server's answer to ChatGPT, Claude, Perplexity, Gemini and Alice bots.
| Item | Points |
|---|---|
| robots.txt is open to AI search and user-request bots (OAI-SearchBot, ChatGPT-User, PerplexityBot…) | 10 |
| robots.txt is open to model-training bots (GPTBot, ClaudeBot, Google-Extended…) — blocking them is legitimate, the penalty is mild | 3 |
| The server returns the home page to AI bots, not a 403, captcha or firewall page | 5 |
| No noindex, noai or noimageai in meta robots or X-Robots-Tag | 2 |
Whether page text is visible without JavaScript: most AI bots don't run it.
| Item | Points |
|---|---|
| Share of the home page text visible without JavaScript | 10 |
| The raw HTML has the title, H1, main text and contacts | 5 |
llms.txt and llms-full.txt — a site map and the full text for AI agents.
| Item | Points |
|---|---|
| /llms.txt opens and is text, not an HTML placeholder | 4 |
| llms.txt structure: heading, description, sections with links | 3 |
| llms.txt links point to your site, work, and are not template placeholders | 3 |
| /llms-full.txt — the full site text for AI agents | advice |
Organization markup, contacts, profiles — how AI understands whose site it is.
| Item | Points |
|---|---|
| Organization or LocalBusiness markup on the home page: name, address, logo, contacts | 6 |
| sameAs links to company profiles (LinkedIn, Instagram, Google Business Profile…) | 3 |
| Page type markup: product, service, article, FAQ, breadcrumbs | 5 |
| canonical and Open Graph (og:title, og:description, og:image) | 3 |
| There are “About” and “Contacts” pages | 3 |
An answer in the first paragraph, numbers, tables, FAQ — what AI quotes.
| Item | Points |
|---|---|
| Answer first: a 20–90-word paragraph with specifics at the top of the page | 7 |
| Headings: one H1, at least three H2, no skipped levels | 4 |
| Tables and lists — facts that are easy to extract | 4 |
| A question-and-answer block | 4 |
| At least 300 words and numbers in the text | 3 |
| Links to sources and attributed quotes | 3 |
Dates, authors, sources and a fresh sitemap.
| Item | Points |
|---|---|
| Publication and update dates on articles | 3 |
| Authors on articles | 2 |
| sitemap.xml exists, is listed in robots.txt, with recent dates | 3 |
| HTTPS with a valid certificate, http redirects to https | 2 |
| SSL certificate: issuer, free or paid, how long it has left | advice |
AI engine coverage
The same formula, but for each engine — its own path to the site: access for its bots (robots.txt and server response) and for the search index it answers from (ChatGPT, DuckAssist and Meta AI — Bing; Gemini — Google; Alice — Yandex). Google, Bing and Yandex run JavaScript, the ChatGPT, Claude and Perplexity bots don't: for them text without JS matters more. If a site is open to all and doesn't depend on JS, the numbers match — there is nothing to tell them apart.
Separate score
Calculated from the same pages, 0 to 100, and not part of AI readiness: good SEO does not guarantee AI will quote you.
| Item | Points |
|---|---|
| title of 10 to 70 characters | 15 |
| description of 50 to 170 characters | 10 |
| One H1 per page | 15 |
| canonical is set | 10 |
| og:title is set | 5 |
| The page is indexable and returns 200 | 20 |
| The language is set in <html lang> | 10 |
| hreflang for language versions | 5 |
| Server response under 1.5 s | 5 |
| Page HTML under 2 MB | 5 |
Full audit only
These items are not part of the score: we show them with evidence and fixes. They will join the next methodology version, based on real checks.
Crawler
Free check: the home page and up to 9 more pages from the sitemap, of different types (catalogue, service, article, contacts). No sitemap — we take links from the home page.
Basic audit: 30 pages, extended — 80; you can buy more or list your own — but not more than the site has.
We open pages with JavaScript in a browser and compare with what is visible without it: the home page in the free check, the home page and one more in the full audit.
AiseenBot does the reading: it respects robots.txt and does not overload the server. How AiseenBot works
Transparency
What the methodology items are based on: standards, AI vendors' documentation and research.
We look not only at robots.txt but at what the server actually returns to AI bots, and what is visible without JavaScript. A site with an open robots.txt but a firewall that blocks bots scores lower with us.
No. Technical readiness 100 means AI can read and understand the site. Whether AI assistants name you is shown by AI answers — 40 points of AI visibility, checked by the extended audit. Getting into answers also depends on demand, competitors and what other sites say about you.
Below the report there is “Score wrong?”: pick the item and describe what is wrong. If we made a mistake, we fix the check for everyone.
Yes, as AI search changes its rules. A new version gets a new number; old reports are shown by their own version, so before/after comparisons stay fair.
Yes, in the extended audit: we ask the assistants customer questions and see whether they name you, in what position, whether they cite your site and whom they recommend instead. That is 40 points of AI visibility. The free check and the basic audit don't check AI answers — they show only the site's readiness.
15 seconds — and you'll know whether ChatGPT, Claude and Alice can see your site.
Check a site