The 12-Point AI Visibility Audit

Most "AI visibility" advice is unfalsifiable. This checklist is the opposite: twelve checks, each with a pass condition and a fix, ordered so that the fatal and cheap items come first. Run the top four and you have already eliminated the causes behind most zero-citation sites.

1. Can each AI crawler reach you at all?

Pass: robots.txt does not block the crawlers whose answers you want to appear in. Fix: replace blanket bans with a per-bot policy — the robots.txt guide has the exact syntax. Check with: the robots.txt Auditor, which tests all 61 tracked crawlers at once.

2. Does the server actually return 200 to those bots?

Pass: a bot-identified request returns the same content as a browser request. Fail: a WAF, bot-protection rule or challenge page returns 403/429 while robots.txt says "allowed" — a silent block that leaves no trace in the file. Check with: the live scanner, which fetches your URL as each bot and compares it to the browser fetch.

3. Is the content in the initial HTML?

Pass: headings and body text are present in the raw response. Fail: the HTML is an empty shell that hydrates in the browser; several crawlers never execute JavaScript. Fix: server-side render or pre-render the pages that matter.

4. Are key pages accidentally noindexed?

Pass: noindex appears only on pages you want excluded — staging copies, duplicate parameter URLs, internal search results. Fail: a CMS default or a leftover tag has deindexed commercial pages. Check with: a site-wide sweep; this is the single most common self-inflicted wound in an audit.

5. Does every section open with the answer?

Pass: the first sentence under each heading answers the heading on its own. Fix: rewrite so the passage is quotable in isolation — see the content structure guide.

6. Do your comparisons exist as tables?

Pass: "X vs Y" and "best Z" content is a real HTML <table> with explicit column headers and units. Fail: the comparison is spread across paragraphs, which is far harder to summarize.

7. Is identity consistent in structured data?

Pass: Organization, Article with a true dateModified, breadcrumbs and — where genuine — FAQPage. Fail: three different company names across pages, or FAQ markup for questions not visible on the page. Check with: the schema guide examples.

8. Is there a visible date and a real author?

Pass: a rendered updated date and a named author or team behind a link. Fail: undated, unattributed pages — these are the ones engines treat as interchangeable.

9. Does llms.txt point at your best material?

Pass: an llms.txt at the root that lists your strongest pages with one-line descriptions, in addition to a sitemap. It is not a ranking mechanism, but it costs nothing and helps agents find depth without crawling everything. Generate one from your sitemap on the homepage in seconds; the llms.txt guide covers the format.

10. Is your content reachable in a few clicks?

Pass: money pages sit within three clicks of the homepage and are linked from at least one hub. Fail: orphan pages that are crawled rarely and cited never.

11. Are broken pages and redirects cleaned up?

Pass: no 404s or redirect chains in the internal link graph. Why it matters: every dead internal link is a passage the summarizer cannot reach and a crawl budget leak.

12. Are you measuring citations, not just ranks?

Pass: a weekly loop that asks five answer engines your ten most important questions and records which URLs get cited. Fail: checking rankings and calling it GEO. The measurement guide has the full loop.

Run it automatically instead of by hand

Points 1, 2, 3, 4, 5, 7, 9, 11 are mechanical. The free scanner covers points 1, 2, 7 and 9 for a single URL in about thirty seconds. To run 4, 5 and 11 across a whole site — up to 50 pages, with per-page bot access, noindex detection and a prioritized P0–P3 fix list — the Pro Deep Report does it for $49 one-time and hands you a private report URL.