What the AI Search Visibility Audit does
This audit checks a single page for the things that decide whether answer engines can use it at all, and whether its content is easy to quote once they can: snippet controls such as noindex and nosnippet, page structure, answer clarity, evidence and authorship, how consistently the publisher is named, self-contained passages and the canonical URL. It lists every finding with the evidence behind it and ranks the fixes.
Each category is labelled by where its rule comes from. Snippet controls are documented by Google as affecting AI Overviews and AI Mode; sourcing and authorship appear in Google's helpful-content guidance; the rest are A2Z heuristics. The result is a rubric score, never a forecast of rankings, citations or traffic.
How to use it
- Paste the page's full HTML, or switch to Fetch a URL. Fetching also reads the X-Robots-Tag response header, which pasted HTML cannot show.
- When pasting, add the page address if you know it; it is used to check the canonical and to separate internal links from outside sources, and is not fetched.
- Select Run audit. The score is a weighted average of the categories; each category shows its basis, its evidence and a link to the guidance it relies on.
- Start with the prioritised fixes. A documented blocker such as nosnippet always outranks editorial suggestions.
- Check the extractable passages: these are the paragraphs that read well on their own. If an important answer is missing from the list, rewrite it so it names its subject and fits in one paragraph.
Reading the results
Access and snippet controls carries the most weight because it is the one area with documented effects. Google's robots meta tag documentation says nosnippet prevents content being used as direct input for AI Overviews and AI Mode, and max-snippet limits how much can be used. noindex keeps the page out of Search altogether. When one of these holds the page back, the verdict says so whatever the other categories score.
Answer clarity, self-contained passages and entity consistency are A2Z heuristics: they describe writing that is easy to quote correctly. Evidence and authorship follows the self-assessment questions in Google's helpful-content guidance, which Google presents as advice, not as ranking factors.
Things the audit deliberately does not score are listed at the end: robots.txt (use the AI Crawler Access Checker), llms.txt (a convention, not a standard) and anything about rankings or citations.
Worked example: a help-centre page and the same page with nosnippet
The help page example is a password-reset article with a meta description, lang attribute, self-referencing canonical, a visible update date and TechArticle JSON-LD. It scores 83 of 100. Snippet controls, self-contained passages and canonical all pass; structure loses a fifth for a jump from H1 to H4 on a "More" heading, which also counts as a vague heading under answer clarity.
Evidence scores 56%: neither of the two specific claims (the 60-minute link expiry and the 10-minute wait) has a source beside it, the only author is an organisation named in JSON-LD, and the one outside link goes to the company's own main domain. Entity consistency scores 30% because the publisher appears as "Northwind Help", "Northwind Traders Ltd" and "Northwind Support".
The blocked example is the same page with meta robots set to nosnippet. Access drops to 20%, the verdict changes to "Held back by a documented snippet control", and removing nosnippet becomes the first fix - ahead of every editorial point.
Documented, convention and heuristic
Search engines say very little about how their AI features choose sources. Google's page on AI features says there are no additional requirements to appear in AI Overviews or AI Mode, and no need for new machine-readable files or special markup. What it does document is how existing controls apply: noindex, nosnippet, data-nosnippet and max-snippet.
Anything beyond those controls is either guidance (write helpful, people-first content with clear sourcing and authorship), a convention some sites follow (llms.txt), or an editorial rule of thumb. Mixing the three into one confident number would mislead, so the audit keeps them visibly separate and says which is which on every line.
Limitations: what the result does not prove
- It reads one page. Site-wide signals - internal linking, how other sites describe you, your robots.txt - are outside its view.
- A fetched page is read as the server sends it, without running JavaScript. Content added by scripts, and meta tags injected by a tag manager, are not seen.
- Scores are rubric scores. A page can score well and never be quoted, or score poorly and be quoted often; no outside tool can observe the selection.
- Name comparison ignores case, punctuation and legal suffixes such as Ltd or Inc, so it treats "Acme Ltd" and "ACME" as the same but "Acme Help" and "Acme" as different.
Privacy: where your data goes
The analysis runs in your browser. If you choose to load a public URL instead of pasting, a2z.tools fetches that address read-only and returns the text to your tab; private and internal addresses are refused and nothing is stored. Session recording and tag-manager scripts are switched off on this page.
Standards and sources
- Google Search Central - AI features and your website - checked 19 Sep 2026
- Google Search Central - Creating helpful, reliable, people-first content - checked 19 Sep 2026
- Google Search Central - Robots meta tag, data-nosnippet and X-Robots-Tag specifications
- Google Search Central - Google common crawlers (Google-Extended)
Frequently asked questions
Can a page opt out of Google AI Overviews but stay in Search?
Google documents nosnippet, data-nosnippet and max-snippet as limiting what can be shown and used as direct input for AI Overviews and AI Mode, while the page can stay indexed. Google-Extended in robots.txt does not do this; it concerns Gemini training and grounding, not Search features.
Do I need special markup or an AI file for AI Overviews?
Google says no: its AI features documentation states there are no additional requirements or special optimisations, and no need for new machine-readable files, AI text files or markup. Normal indexing and snippet eligibility apply.
Why does the audit give a score if it cannot predict visibility?
A score summarises the rubric so you can compare versions of the same page and see progress. It is labelled as a rubric score everywhere, and each category shows the evidence and the basis behind it, so you can disagree with any part.
What does the X-Robots-Tag header have to do with it?
Robots rules can be sent as an HTTP response header instead of a meta tag, and they have the same effect. Only a fetched page reveals headers, so use Fetch a URL when you need that part checked.
Why is my brand named three different ways?
The audit compares the title suffix, og:site_name and every Organization or WebSite name in the JSON-LD. Help centres, shops and blogs often use a sub-brand in one place and the legal name in another. Choosing one form avoids giving readers and parsers mixed signals.
Which paragraphs count as extractable passages?
Paragraphs of 30 to 150 words that do not open with a pointer back to earlier text, such as This, It, However or Also. They are ranked by how many of the page title's words they use and whether they state a specific fact.
Does the audit check robots.txt for AI crawlers?
No, it reads only the page. The AI Crawler Access Checker reads robots.txt for each documented AI user agent, and the AI Bot robots.txt Policy Builder writes rules for them.
Last reviewed by the A2Z.Tools team against the sources listed above.
