Skip to content

AI Vision

See your site through an AI engine's eyes

AI Vision scores your pages across the ten signals that decide whether an engine can read and trust you, from bot access and structured signals to entity authority, content clarity and what the models already say about you, and shows you exactly where the breakdowns are.

Free starter scan. No card required.

What it checks

One real page, read twice, and the ten passes in between. Everything below comes from the guinness.com scan embedded further down this page.

AI Vision scan of guinness.com - 25 August 2026 - 50 pages crawled - 10 modules - 2 models, 11 prompts

What a person sees
www.guinness.com/en/our-craft/guinness-story

Home / Our Craft

The Story of Guinness

“It takes a thirst for adventure to do things the Guinness way. From our humble beginnings in 1759 to the present day, we’ve gone to extraordinary lengths to bring you exceptional beer. But we’re not ones to rest on our laurels. As we like to put it: our greatest work is yet to come.”

Verbatim from the scan. The crawler extracted 250 words from this page.

What an engine keeps
ClaudeBot - allowed - 50 pages - no JS rendered
Access
All major AI crawlersGPTBot, ClaudeBot, PerplexityBot, CCBot, Applebot and anthropic-ai are all allowed
Content
250 words extractedEvery crawled page carries an H1, a description and a <main>
Hierarchy
BreadcrumbListDeclared on 49 non-homepage pages
Publisher
UnresolvedNo Organization schema on any of the 50 pages
Author
UnresolvedNo Person schema either
llms.txt
Not foundThe site root returned 404

This site does almost everything right. Every AI crawler is welcome, all 50 pages carry structured data, and the prose extracts cleanly. It still cannot say who publishes it. No Organization or Person schema was found anywhere on the site, and that is the primary signal an engine uses to attach a claim to a brand. It is the one failing finding in the whole scan.

And this is what the models already believed, asked cold

ClaudeGuinness is an Irish stout beer brand owned by Diageo, one of the world’s largest alcoholic beverage companies. It’s famous for its distinctive dark color, creamy head, and rich flavor.
ChatGPTGuinness is a famous Irish brewery known for its iconic stout beer, particularly the Guinness Draught.

Both models recognised the brand, and it appeared unprompted in 4 of 10 category answers. These are single-shot API calls at temperature 0, not consumer app sessions, so they report what the model already holds rather than what it would retrieve live.

Ten passes decide how much of the page survives the crossing

  1. 1

    Bot access

    Whether GPTBot, ClaudeBot, PerplexityBot and the rest are allowed to fetch the page at all, or are named and turned away at the door.

    robots.txtX-Robots-Tag

  2. 2

    AI indexability

    Once fetched, whether the page is allowed to be kept, served over HTTPS, listed in a sitemap that is not years stale, and pointed at one canonical address.

    rel=canonicalsitemap.xmlmeta robots

  3. 3

    llms.txt quality

    Whether there is a map written for AI readers rather than for search crawlers, and whether the links inside it still resolve. Guinness has none, which is the one warning in this scan.

    /llms.txt/llms-full.txt

  4. 4

    Structured coverage

    How many of your crawled pages carry any machine-readable markup at all, and how many distinct kinds. Here it was 100% of pages and 5 types: BreadcrumbList, FAQPage, Product, Recipe and WebSite.

    JSON-LD@type

  5. 5

    Entity authority

    Whether an engine can say who publishes this, and be sure it is the same “who” on every page. This is the pass Guinness fails, and it costs the whole category its score.

    OrganizationPersonauthor

  6. 6

    Rich signals and E-E-A-T

    Whether a page states in markup where it sits in the site and what kind of thing it is, so a citation has evidence behind it rather than just text.

    FAQPageBreadcrumbListRecipe

  7. 7

    Content clarity

    Whether the prose declares its own subject: one real H1, a description worth quoting, a <main> to separate body from chrome.

    <h1><main>meta description

  8. 8

    LLM perception

    What the models say when asked about your domain cold, with no page in front of them. This scan asked two, and both recognised the brand.

    ClaudeChatGPT

  9. 9

    Competitive positioning

    Whether you come up unprompted when a model is asked about your category. Guinness appeared in 4 of 10 category answers across the two engines.

    5 questions2 enginestemperature 0

  10. 10

    Citation quality

    Which sources an engine cites back when it answers about you. This needs a live-retrieval engine, so it only runs when Perplexity or AI Mode is enabled on the template. It was not enabled here, and the scan says so rather than guessing.

    PerplexityAI Mode

Every crawled page runs the page-level passes. Pass Warn Fail This run returned 15 passes, 1 warning and 1 failure, and scored 76 out of 100.

What you'll get

See exactly how AI reads your pages

Run a free scan

AI Vision

Each scan scores your site across 10 AI-readiness signals, from bot access and structured signals to content clarity and model perception. Run any time; results are fixed snapshots so you can track progress.

New scan
George
Online
0%