Absence is the normal state
Most indexed pages never appear in an AI Overview, so absence is not a penalty and rarely has one cause. The causes below are not equally likely, and they depend on each other. Working them in order stops you rewriting content to solve a rendering problem.
Cause 0: The query does not produce an Overview
DOCUMENTED Google shows an AI Overview “when our systems determine that generative AI can be especially helpful”. Many queries do not trigger one, and the set changes. TESTED Ahrefs found that 99.2% of the keywords that triggered an Overview in its 300,000-keyword sample were informational. A commercial or transactional query may produce no Overview at all, in which case there is nothing to appear in.
- Check: Search the query from a Perth location, in a fresh session, three times across a fortnight. No Overview on any run means this page’s absence is not a page problem.
Cause 1: The page is not indexed, not eligible, or not reachable
DOCUMENTED “To be eligible to be shown as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to be shown in Google Search with a snippet.” A noindex, a nosnippet, a crawl block, or a page Google has discovered and not indexed is not a candidate. Two variants are common: Google cites the business but picks a different page than the one intended, usually an article rather than a service page; or the intended page is indexed and simply does not rank well enough on the query to be retrieved. Neither is an AI problem.
- Check: Inspect the URL in Search Console for its index status. Confirm no
noindexmeta orX-Robots-Tag, and nonosnippetor restrictivemax-snippet, since Google says “more restrictive permissions will limit how your content is featured”. View the raw source and confirm the content is present without JavaScript, per the crawler access guide. - Fix time: Days to weeks for a directive; weeks to months for a rendering architecture.
Cause 2: The answer is buried
INFERRED Answers are composed from passages. A page that reaches its point in paragraph nine competes with pages that reach it in paragraph one, and the section is the unit being used.
- Check: Count the words before the first sentence that directly answers the query; more than about a hundred is a probable cause. Read the headings: “Our approach” tells a system nothing, “How long does an entity audit take?” tells it what the section contains.
- Fix time: Hours per page, then weeks for recrawl.
Cause 3: Structured data contradicts the page
DOCUMENTED Google’s guidance is that structured data must match the visible text. Markup asserting a name, address or service the page does not show is a conflict the site created.
- Check: Read the JSON-LD against the visible content. The structured data guide lists the five errors.
- Fix time: Hours, then weeks for recrawl.
Cause 4: Entity signals are unclear
INFERRED A business that resolves to three partial records is a weaker source to name than one that resolves to one. DOCUMENTED Google’s own fundamentals for AI features include keeping Business Profile information current.
- Check: Search the business name and read the descriptions that come back; if they disagree, the disagreement is in the data. Compare the site footer, the ABN record and the Google Business Profile for name and address. The consistency audit is the full inventory.
- Fix time: Weeks to correct; months for third-party sources to update.
Cause 5: Coverage is thin or unoriginal
DOCUMENTED Google documents query fan-out, issuing related searches across subtopics, and its advice to “provide them with unique, satisfying content”. INFERRED A page restating what is already indexed hundreds of times gives nothing a reason to prefer it, and a single page on a topic gives a fan-out little to retrieve.
- Check: List the sub-questions a buyer would have around the query and count how many the site answers. Then ask whether anything on the page could not have been written by someone who had read three competitors.
- Fix time: Months. Content work does not compress.
Cause 6: Corroboration is weak
INFERRED A business that is registered, listed and mentioned elsewhere is a less risky thing to summarise than one that exists only on its own site. This is the slowest layer and the one the business controls least.
- Check: Search the business name in quotes, excluding its own domain, and count genuine independent mentions. Fewer than a handful for an established business is a finding.
- Fix time: Many months.
What is fixable quickly, and what is not
| Cause | Effort | Time to any effect |
|---|---|---|
| Query triggers no Overview | None possible | n/a |
| Crawl or index directive | Low | Days to weeks |
| Rendering architecture | High | Weeks to months |
| Buried answer, weak headings | Low | Weeks |
| Contradictory markup | Low | Weeks |
| Entity inconsistency | Medium | Weeks to months |
| Thin coverage | High | Months |
| Weak corroboration | High | Many months |
What is not fixable at all
DOCUMENTED Google’s caveat applies to everything above: meeting every requirement “doesn’t mean that Google will crawl, index, or serve its content. Indexing and serving isn’t guaranteed.” Generation varies week to week; a source present one week and absent the next has not necessarily done anything wrong. Chasing individual query appearances is a poor use of budget compared with fixing the structural causes, and no technique documented by Google forces inclusion.
What is documented and what is inferred
- Documented: when Overviews show; the eligibility rule; preview controls; fan-out; the no-guarantee caveat; Ahrefs’ informational-intent finding.
- Inferred: passage-level extraction; entity and corroboration effects; the effort table.
- Not established: any Perth trigger rate, pending the planned source study.