Seven Reasons AI Search Is Not Citing Your Site

Tomohiro Iida · Published June 25, 2026 · Updated July 7, 2026

Requests to figure out why AI search is not citing a company's website have been increasing, but the cause is rarely a single thing. Before producing an answer, an AI system passes through several stages: it retrieves relevant information, selects sources to cite from what it retrieved, and works the cited content into its answer. Not being cited means getting stuck at one of these stages, and the fix differs depending on which stage that is. Writing "more specific" content does nothing for a page that is blocked at the retrieval stage, and fixing indexing is beside the point if a page is already being cited as a source but its content simply is not showing up in the answer text. This article sorts seven causes into three stages — retrieval, citation selection, and answer integration — and for each explains the failure point, the symptom, how to check for it, and how to fix it.

How AI selects its sources

Google's own documentation describes RAG (retrieval-augmented generation) and query fan-out as the mechanisms behind this. RAG means the AI searches indexed content and builds its answer from what it finds, so being indexed and evaluated in ordinary search is the foundation. Query fan-out means a single question generates several related queries at once, so a page can be picked up even without matching the original question word-for-word, as long as it covers a related topic. Google also states that optimizing for generative AI search is optimizing for search generally, in other words that this is SEO rather than a separate discipline with a back door, which is why most of the seven causes below trace back to ordinary SEO and information-architecture issues.

Seven reasons AI is not citing you

• 1. Not indexed (retrieval, entry point) — pages that are not indexed cannot be candidates regardless of quality. Check via Search Console URL Inspection, robots.txt blocks, noindex tags, and whether the content renders without relying on client-side JavaScript.
• 2. Indexed but not surfaced for the query (retrieval) — the page passes the indexing gate but does not get picked up for related searches. Check by testing the target query and its likely fan-out variants, and fix by making each page cover one clear topic plus its related questions, linked internally to related pages.
• 3. Surfaced but not selected as a citation (citation selection) — the page ranks, but a competitor is cited instead. Compare what the cited competitor has that yours does not, such as first-party data, concrete numbers, examples, or judgment criteria, and add verifiable specifics rather than a generic restatement.
• 4. Cited but not reflected in the answer text (citation selection and answer integration) — your URL appears in the source list, but the answer text does not reflect your content. Check whether the key point is buried in a long preamble or vague phrasing, and fix by stating the point clearly and concretely, not by adopting a special "AI-only" format, since Google states no such special formatting is required.
• 5. No evidence, source, or figures behind the claims (answer integration) — assertions such as "highly effective" or "widely adopted" without a source, number, or example are less likely to be treated as reliable. Add verifiable evidence, such as data-collection conditions and dates for first-party data, or sources for external information.
• 6. Facts differ across the site or channels (affects every stage) — inconsistent company name, role, offerings, or track record across the company page, service pages, PDF materials, and external profiles leaves AI unable to tell which version is authoritative. Fix by designating one place as the single source of fact and aligning everything else to it.
• 7. Results vary each time you check (a precondition for diagnosis, not a cause in itself) — AI answers vary by model, search-feature availability, region, and time, so a single observation should not be mistaken for the cause. Observe a fixed set of queries repeatedly across conditions and judge by the trend, not one snapshot.

The order to fix things in

• 1. Confirm retrieval first (causes 1–2) — if a page is not indexed or is not surfaced, citation and answer-integration issues are moot; confirm with Search Console URL Inspection and live searches.
• 2. Then look at citation (causes 3–4) — check first-party specificity and clarity of wording once the page is confirmed to surface in search.
• 3. In parallel, build the evidence base for answer integration (causes 4–5) — sourcing and figures help both citation and integration, so address them early.
• 4. Establish consistency as a foundation (cause 6) — factual inconsistency degrades accuracy at every stage, so align basic facts early.
• 5. Judge everything on continued observation (cause 7) — no single observation decides success or failure.

AI search answers are not stable from one check to the next, and a single result should never be treated as final. The seven causes above are mostly extensions of ordinary SEO and information architecture rather than a separate "AI-only" discipline, so fixing them in this order, from retrieval through to consistency, tends to be the more efficient path.