What an AI visibility report
actually checks
Five checks that decide whether a page can be found, read and quoted by AI answer engines. Run on this site, with every number you can open.
AI visibility
What does an AI visibility report check?
It checks five things that decide whether a page can be found, read and quoted by AI answer engines. They are crawler access, AI readiness, page speed, structured data and freshness. Each one is a card. The number at the top of a card is the average of the bars underneath it. The report keeps these jobs apart. So each failing check points to one kind of fix.
Most SEO reports still answer one question. Can Google crawl this URL and does it load quickly? That question still matters. It is no longer the only one.
When someone asks ChatGPT, Perplexity, Gemini or Google AI Overviews for a recommendation, the system has to do more. It has to fetch a page. It has to tell the main copy apart from the menu. Then it has to find a section that actually answers the question. The AI Readiness Report from LLM Scout puts those checks on one page. I ran the same checks against this site, frankvitetta.com. This essay walks through what they found.
I have worked in search since 2008. The reports I used to send clients were a crawl, a PageSpeed export and a schema spot-check, stapled together in a slide deck. Clients could not tell which finding was about being indexed and which was about being cited. This report is the short version of that pile. It has five cards. Each card is a measurement. The number at the top of a card is the average of the bars underneath it.
The five checks
| Check | What it looks at | Why it belongs in an SEO report |
|---|---|---|
| Crawler access | The robots.txt rules for search and AI crawlers. Whether key pages carry noindex. | A page crawlers cannot fetch cannot be read. A page marked noindex is kept out of search results. |
| AI readiness | HTML depth, the share of text inside <main>, image alt text, headings and paragraphs, language and canonical. | These are the signals that make a page easy to extract, for a featured snippet or for an AI answer. |
| Page speed | The Lighthouse performance score, averaged across the URLs that were crawled. | The card shows performance only. SEO, accessibility and best practices are listed under it and do not move that number. |
| Structured data | JSON-LD matched to the page type. A checklist of title, description, headings, lists and Open Graph tags. | Schema tells a crawler what the page is. The checklist shows whether the HTML agrees. |
| Freshness | A date on the page. That can be datePublished or dateModified in JSON-LD, a <time> element, a date in the copy or a Last-Modified header at least two days old. | Editors can see which URLs have a date and which do not. A product page with no publication date shows as "no date". It is not marked as a fail. |
The crawler card also lists llms.txt when the file returns HTTP 200. A missing file shows as optional. Google's guidance on AI features says you do not need to create new machine readable files, AI text files or markup to appear in AI Overviews or AI Mode. So the report does not treat a missing llms.txt as a broken site.
What frankvitetta.com shows
This site is small on purpose. As of 20 October 2026 the sitemap lists 14 URLs. They run from the home page and the essays index to the hire, about and glossary pages. A small site is a useful test. If a check only works on a shop with 10,000 pages, it is the wrong check for a consultant's own domain.
Crawler access
The robots.txt file returns HTTP 200. It names the crawlers one by one instead of hiding them in a wildcard. Googlebot, Bingbot, Google-Extended, GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot and Amazonbot are all allowed. The sitemap is declared in the same file. No page in the sitemap is marked noindex. The llms.txt file is there too, so that row shows as present. My guide to AI crawlers explains what each of those names does.
Page structure
The home page is in British English (lang="en-GB"). It has one canonical URL and one H1, "How brands get found in AI answers." As of 20 October 2026 there are 6 H2s, 9 H3s and 17 paragraphs under it. The page uses header, nav, main and footer elements. Each of the 6 H2s has an id, so those sections can be linked directly. There are no images on the home page, so alt text does not come into it there.
The essay How AI answer engines pick and cite sources is the clearer example of a page built to be quoted. As of 20 October 2026 it has one H1, 10 H2s and 26 paragraphs. Nine of those H2s have an id. That is the extractability check in practice. A long page is broken into sections, with anchors on the sections worth citing.
Schema
The home page publishes one JSON-LD graph with two types, WebSite and Person. The Person has a job title, a sameAs link to LinkedIn and a knowsAbout list that includes technical SEO, structured data and AI readiness. The answer engines essay adds Article, BreadcrumbList and ImageObject. The report scores that fit against the page type. A home page that carries only Person and WebSite is doing a different job from an essay that should carry Article.
Dates
The answer engines essay shows two visible <time> elements, one for the publication date and one for the last update. Its Article markup carries the same two dates as datePublished and dateModified. Its sitemap entry carries the last update. The home page has no publication date of its own. The <time> elements on it belong to the essays it lists. Its own date signals are the sitemap entry and the Last-Modified header.
That header is the weakest signal of the four. On this site it does not follow the edit history. When I checked on 20 October 2026, three pages that had not changed since 6 October showed Last-Modified dates of 15, 18 and 20 October. So the report prefers a date in the markup. It only falls back to the header when that is the only signal and the date is not a same-day deploy stamp.
Speed
I am not going to print a Lighthouse score from a single run. Google says the performance score can change from run to run even when the page has not changed. The report runs PageSpeed Insights on the crawled URLs and puts the performance average on the card. PageSpeed Insights tests four categories: Performance, Accessibility, Best Practices and SEO. The other three stay visible under the card. So a client can see a fast page that still fails accessibility, without that failure being relabelled as "slow".
Why this is useful for SEO
Search engines and answer engines share the same basic requirements. The URL has to be allowed. The answer has to be in the HTML itself. A script that paints it in later is not enough. The title, the canonical and the language have to agree. The sections have to be real headings, because headings give both a snippet and a citation their edges.
The report is useful because it keeps those jobs apart. A red crawler row is a robots.txt or noindex problem. A weak extractability bar means too little of the copy sits in <main>. A structured data gap means the JSON-LD does not match the template. It can also mean the title and description checklist failed. Mixing all of that into one "AI score" is how audits turn into a slide nobody can act on.
From report to audit
I use the same five cards in a fixed-scope AI visibility audit. It covers how the brand shows up in AI answers and in Google. Then it runs the technical checks. Then it sets out a plan in order of priority. The report is the technical half. It is written so a developer can see the failing check without sitting through the strategy. The audit is described on the hire page. For what can and cannot be measured on the answer side, see measuring AI visibility.
If you want the worked example, start with the home page and the answer engines essay. The robots.txt file, the sitemap and the JSON-LD are all public. That is the standard I wanted the report to hold. Every number should point at something you can open.
Sources
- LLM Scout, AI Readiness Report: llmscout.co/buy-ai-readiness-report
- Google Search Central, AI features and your website: developers.google.com/search/docs/appearance/ai-features
- Google for Developers, About PageSpeed Insights: developers.google.com/speed/docs/insights/v5/about
- Chrome for Developers, Introduction to Lighthouse: developer.chrome.com/docs/lighthouse/overview
- Google Search Central, Block search indexing with noindex: developers.google.com/search/docs/crawling-indexing/block-indexing
- Google Search Central, Introduction to structured data markup: developers.google.com/search/docs/appearance/structured-data/intro-structured-data
- IETF, RFC 9309 Robots Exclusion Protocol: rfc-editor.org/rfc/rfc9309
- The llms.txt proposal: llmstxt.org
- MDN, the main element: developer.mozilla.org/en-US/docs/Web/HTML/Element/main
- MDN, the time element: developer.mozilla.org/en-US/docs/Web/HTML/Element/time
This essay was researched and drafted with AI assistance, then reviewed and edited by me before publication. The editorial policy explains the process and how corrections are handled. If you think something on this page is wrong, please tell me.