How the check works
The checker fetches three things from your site over the public internet, the way an AI crawler does: /robots.txt, /llms.txt and your homepage. No language model is involved. Each finding is a fact about a file or a tag: present or missing, allowed or blocked.
Six signals make up the score. 115 points are possible; the score is points earned over 115, scaled to 100.
- AI assistant access, 35 points. robots.txt is parsed for the 10 crawlers listed below; credit is given per crawler still allowed. No robots.txt means everything is allowed and scores full marks.
- Structured data, 20 points. An Organization, WebSite or similar identity block in JSON-LD earns 12; up to two further types such as Product, Service or FAQPage earn 4 each.
- Answerability, 20 points. Homepage word count measured up to 600 (12 points) and H2 headings measured up to 4 (8 points).
- How the homepage introduces itself, 15 points. A title of at least 10 characters, a meta description of at least 50, and an H1. 5 points each.
- Your keyword on the homepage, 15 points. Full credit if the phrase you typed appears in the title, description, H1 or body; half if only its individual words do.
- llms.txt, 10 points. Present or not.
You see the access, llms.txt, introduction and keyword findings in full. The structured-data and answerability findings are measured and counted in the score, and their detail is in the paid report under Current Position and What to Build. The score is therefore a score of the whole site, not of the free half; buying the report does not move it.
What the score means
80–100. Crawlers allowed, homepage names the business. Assistants can read and understand the site. Whether they recommend it is now about authority and competition, which the checker does not measure.
50–79. Usually one large gap. A Disallow against some crawlers, no llms.txt, or a title and description that do not name the category. The finding text says which.
Below 50. Usually blocked crawlers. Access is about 30% of the score, so a site that disallows all 10 cannot score above about 70.
The score measures readiness, not recommendation. Two sites can both score 90 and be named very differently by the assistants, because recommendation depends on evidence the checker does not gather: who else says what you say, how you compare with the sources an assistant already cites, and what it answers when asked your keyword directly.
Which crawlers the check covers
The access check parses robots.txt for these 10 user agents. Each is the crawler behind a place your brand can appear:
- OAI-SearchBot — ChatGPT search results.
- ChatGPT-User — ChatGPT browsing your page on request.
- Claude-SearchBot — Claude's search results.
- Claude-User — Claude reading your page on request.
- GPTBot — OpenAI's model knowledge of your brand.
- ClaudeBot — Claude's model knowledge of your brand.
- PerplexityBot — Perplexity answers and citations.
- Google-Extended — Gemini's training and its Search-grounded answers.
- CCBot — Common Crawl's open web archive.
- Applebot-Extended — Apple's model knowledge of your brand.
Blocking one keeps your site out of what it feeds. How often the most-visited sites block each one: our study of the top 1,000 websites.
Two sites, checked
Both checked with this engine on 27 September 2026. nytimes.com disallows all ten AI crawlers in its robots.txt, a deliberate publisher decision, and its homepage turned our scanner away as well. Blocking the crawlers alone caps any site below about 70.
stripe.com scored 93 for the keyword “payments”. Every crawler is allowed, the site ships an llms.txt, the title and description name the product, and the keyword appears where it should. The only points it lost were on structured data.
Checker vs audit
The checker answers whether assistants can read your site. The audit answers who they recommend for your keywords instead of you, and what to fix first. That needs data from outside your site. To compare paid tools first, trackers included, see the best LLM visibility tools, compared on price.
For each keyword you buy, the audit agent pulls live Google results and keyword data and crawls the competitor pages those results surface. For your primary keywords (the top one, two or three depending on plan) it asks ChatGPT, Claude, Gemini and Perplexity the question and records which sources they cite; the remaining keywords get a ChatGPT-only pass. It also measures page speed with lab and field data and captures screenshots. The report is written under fixed headings — Current Position, What to Build, Evidence, 90-Day Roadmap and Measurement, with Keyword Portfolio and Competitive Landscape on the larger plans — and arrives as a link within 60 minutes of payment. It covers SEO, generative engine optimization (GEO) and answer engine optimization (AEO) together, because the fixes overlap.
| Plan | Price, one-time | Keywords | Positioning |
|---|---|---|---|
| Starter | $49 | 1 | Perfect for validating your most important search term. |
| Growing Brands | $99 | 5 | See how your brand performs across the AI searches that matter most. |
| Market Leaders | $149 | 7 | Get the complete picture of your AI visibility and growth opportunities. |
How to improve the score
In the order the points reward them:
- Allow the crawlers. Remove Disallow rules against GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, Claude-User, PerplexityBot, Google-Extended, CCBot and Applebot-Extended, or add Allow lines for them. This is about 30% of the score. Check the file even if you never wrote a rule; a block can arrive with a template.
- Publish llms.txt. A plain-text file at /llms.txt that says what the site is, what it offers and where the main pages are, in the llmstxt.org shape. Ours is at viscovery.ai/llms.txt. 10 points.
- Name the business in the title, description and H1. The checker wants a title of at least 10 characters, a description of at least 50, and one H1, and it looks for your keyword in all three. A homepage that opens with a slogan and never names the category fails this check. 15 points for the three, 15 more for the keyword.
- Add Organization and WebSite structured data. A JSON-LD block naming the organization and its site earns 12 of the 20 structured-data points. Product, Service, FAQPage or BreadcrumbList blocks that describe what is actually on the page earn 4 each, up to two.
- Write sections an assistant can quote. The checker measures homepage word count up to 600 and H2 count up to 4, because assistants cite passages, and a passage needs a heading above it and enough text under it to stand alone. 20 points.
We run this scan against our own site every day and publish the result at /audit/viscovery. If it regresses, that page says so.
Questions people ask
How can I check my AI visibility?
Three checks need only a browser. Open yourdomain.com/robots.txt and look for GPTBot, ClaudeBot, Claude-SearchBot, PerplexityBot and Google-Extended; a Disallow line against any of them keeps that crawler off your site. Open yourdomain.com/llms.txt; if it exists, you are describing your site to assistants in the format they look for. Then read your homepage title, first heading and first paragraph and ask whether they say what the business does. The checker above runs these three checks, adds structured-data and content-depth measurements, and returns one score in seconds. It does not tell you who the assistants recommend for your keyword instead of you; that needs search-result and AI-answer data for the keyword, which is what the paid report collects.
What is a good AI visibility score?
On this checker, 80 and above means the AI crawlers are allowed in and the homepage says what the business does. Between 50 and 80 there is usually one large gap: a robots.txt rule blocking some crawlers, no llms.txt, or a title and description that do not name the business. Below 50 usually means the crawlers are blocked. Access carries about 30% of the score, so a site that blocks every AI crawler cannot score above about 70 whatever else it does. The score measures readiness, not recommendation: it says whether assistants can read and understand the site, not whether they choose it over a competitor when someone asks. Two sites can both score 90 and be recommended very differently. That second question is what the paid audit answers.
What is the best AI visibility tool?
It depends on the question. "Can assistants read my site?" is answered by a checker that reads your real robots.txt, llms.txt and homepage; this one does that in seconds for free. "Which brands does ChatGPT or Perplexity name for my keyword, and why not me?" needs a tool that asks the assistants and the search engine that question and compares the sources they cite; that is an audit. "How does that change week to week?" needs a tracker with a dashboard, which is a different product. Viscovery is a checker and an audit: the free scan for the first question, the paid report for the second. We do not sell a tracker. If you need ongoing monitoring across many keywords, buy a tracker; if you need to know what to fix first, buy the audit.
What is AI visibility?
AI visibility is whether your brand appears, and is described accurately, in the answers ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews give when someone asks a question your business could answer. It differs from search visibility. A search engine returns a list of pages and the reader chooses. An assistant reads several sources, decides which it trusts, and names a few of them in one answer. A brand can rank on page one of Google and never be named, because ranking rewards matching a query while being named rewards being clear, current, corroborated and readable by the crawlers that feed the models. The work of improving it is called generative engine optimization (GEO) when it is about the whole site, and answer engine optimization (AEO) when it is about specific questions.
More at AI visibility and GEO, explained, one question per page.