AI search fix
How Google AI Overview picks sources
Google AI Overviews do not run a separate mystery index. They draw on the same crawl, indexing, and ranking machinery as Google Search, then select passages that answer the query with enough confidence to cite. Your page must be fetchable by Googlebot, indexed without blocking signals, and written so a short excerpt still makes sense out of context. No add-on file guarantees a citation slot. llms.txt is optional for Google; focus on crawl access, helpful content, and accurate schema.org JSON-LD that matches what visitors see.
Teams often chase one-off prompt tests in ChatGPT while the real bottleneck is still technical: 403 from a WAF, accidental noindex, or thin pages that never enter the retrieval pool. Fix those first, then tighten the opening paragraph on each priority URL so it states the answer before background detail.
The four stages behind an AI Overview citation
| Stage | What Google evaluates |
|---|---|
| Crawl | Googlebot can fetch HTML without edge blocks or robots disallow conflicts. |
| Index | Page is eligible in Search; no unintended noindex or canonical to a weaker URL. |
| Retrieve | Query intent matches visible headings, entities, and answer-first copy. |
| Cite | Passage is specific enough to quote; trust and freshness signals support use. |
Free · 2 minutes · no card
See what AI crawlers hit on your site
Technical blockers, missing context, weak AI-readiness signals — in one HTML report.
You'll get an HTML report on crawl and structure blockers before source-visibility work.
Where sites most often get filtered out before the Cite stage
The most common blocker is not content quality — it is that the page never reaches the Retrieve or Cite stage at all because of a Crawl or Index failure upstream. A WAF challenge that only affects specific paths, an accidental noindex left over from a staging deploy, or a canonical tag pointing to a thinner page can all silently remove a page from consideration before its content is ever evaluated.
Because these four stages are sequential, fixing a Cite-stage issue — like vague, non-quotable copy — does nothing if the page is still failing at Crawl or Index. Diagnose in order: confirm Googlebot gets a 200, confirm the page is actually indexed and not canonicalized away, then move on to retrieval and answer quality.
Writing passages that survive the Retrieve and Cite stages
A passage that quotes well states its claim in a self-contained sentence or two near the top of a section, without depending on a preceding paragraph for context. If a sentence only makes sense after reading the three paragraphs before it, an excerpt-based system is less likely to lift it cleanly.
Keep the stated facts consistent with the rest of the page and with your schema.org markup — a mismatch between visible text and structured data is a trust signal that can work against citation, not for it. None of this changes stage one and two: without crawl and index access, none of it is evaluated in the first place.
This is also why the same page can be quoted for one query and ignored for a closely related one — the Retrieve stage matches passages to specific query intent, not to the page as a whole. A page can have one strong, citable paragraph and several weaker ones; only the strong paragraph needs to clear the bar for that particular query.
Frequently asked questions
Is there a guaranteed way to appear in AI Overviews?
No. Google documents quality and technical requirements, not a citation guarantee. You can improve eligibility by fixing crawl issues and strengthening on-page answers.
Does llms.txt control AI Overviews?
No documented requirement ties AI Overviews to llms.txt. Google has stated helpful content and crawl access matter more for generative features in Search.
Do AI Overview citations match blue-link rankings?
Not always. A page can rank traditionally yet never be quoted, or appear in an overview without holding position one. Intent and passage quality both vary by query.
Related questions
Updated