01Diagnosis
Missing from AI Overviews while ranking on page one
Holding a top-ten position no longer guarantees appearing in the AI Overview above it. The link between the two weakened sharply through 2026, and the fix depends on which of three things is actually happening.
Quick answer
Holding a top-ten position no longer guarantees appearing in the AI Overview above it. The link weakened sharply: top-10 share of Overview citations fell from roughly 76% in July 2025 to 38% by March 2026.
Check three things in order: whether an Overview triggers at all, whether the query is one Google treats cautiously, and whether your page answers early enough.
Key takeaways
- The most common false alarm is a query that stopped triggering. Nothing is wrong with your page and there is nothing to fix.
- Top-three pages are still cited 98.9% of the time. Ranking remains the strongest single gateway.
- But 55.2% of citations came from outside the top ten. Both facts are true.
- Answer in the first 100 words. 22.1% against 8.7%: the strongest structural signal measured.
- Health, financial and legal queries lean institutional. Shift effort to questions you can actually win.
- llms.txt, chunking and AI schema do not help here. Google names all three as unnecessary.
Short answer
Check three things in order: whether an Overview triggers for the query at all, whether the query is one where Google prefers institutional sources, and whether your page answers early enough to be usable. Ranking still helps more here than on any other AI surface, but it is no longer sufficient.
The measurable change
The share of AI Overview citations coming from top-ten organic results fell from roughly 76% in July 2025 to 38% by March 2026, according to Ahrefs. In a separate 2026 study, 55.2% of AI top-ten citations came from pages that were not traditional top-ten results.
At the same time, pages in the traditional top three appeared in AI citations 98.9% of the time. Both facts are true because the citation pool is far wider than ten results: ranking gets you in almost automatically, but not ranking no longer keeps you out.
Verified Published third-party research. Sourcing and limits: our measurement standard.
Check one. It does an Overview trigger at all?
Overviews appear on roughly half of queries and the set changes. If your query stopped triggering one, nothing is wrong with your page and there is nothing to fix.
This sounds obvious. It is also the most common false alarm we see, because people check once, see no Overview, and conclude they were removed from something that was never there that day. Track trigger rate as a metric in its own right.
Check two. It is the query one Google treats cautiously?
Health, financial and legal queries lean heavily toward institutional sources. If you are a commercial provider competing with a regulator or a national health service for a clinical question, the realistic move is to shift effort toward questions where you are genuinely the best source — cost, process, eligibility, what happens next.
That pattern is covered in more depth on our healthcare and legal pages, because it changes the whole content strategy in those sectors rather than being a tweak.
Check three. It is the page answer-shaped?
Two structural signals came out clearly in 2026 citation research.
| Signal | Measured effect | What to change |
|---|---|---|
| Topic in first 100 words | 22.1% cited vs 8.7% otherwise | Cut the scene-setting paragraph. Answer, then explain |
| Serves 8+ related queries | 43.3% AI top-ten vs 15.5% for single-query pages | Broaden one page rather than splitting into several |
| Factual specificity | Top quartile outperformed the bottom | Named entities, dated figures, stated sources |
| Contains a table | Present on 50.5% of cited pages | Add one where a real comparison exists. Not decoratively |
| Page length | Almost no relationship with position | Nothing. Length follows coverage; it is not a target |
What does not help
Google's May 2026 guidance names llms.txt, content chunking and AI-specific structured data as unnecessary for its generative features. If a proposal to fix this problem contains those three items, it is not addressing the problem.
If none of the three checks explain it
Then you are probably looking at a coverage or authority gap rather than a structural one, and that is slower to fix. Before assuming so, confirm the pattern holds across several prompts rather than one — the sampling method is published, or the two-week baseline does it across platforms.
Related: generative engine optimisation for the work itself, and platform behaviour notes for how this surface differs from assistants.
02Common questions
Questions people ask next
Why am I not in the AI Overview when I rank third?
Because ranking is no longer sufficient. The share of AI Overview citations coming from top-10 organic results fell from roughly 76% in July 2025 to 38% by March 2026. Top-three pages are still cited 98.9% of the time, so check whether an Overview triggers at all first.
Do AI Overviews appear for every query?
No. They trigger on roughly half of queries, and the set changes over time. If your query stopped triggering one, nothing is wrong with your page. Track trigger rate as a metric in its own right rather than assuming an Overview always exists.
Does structured data help me get into AI Overviews?
Not as a lever. Google states no AI-specific schema is required, and no schema type showed a consistent independent citation advantage in 2026 research. Mark up what is visibly true, validate it, and put the effort elsewhere.
Is AI Mode the same as AI Overviews?
No, and it should be sampled separately. They are different surfaces with different behaviour, and assuming a reading from one describes the other is a common measurement error.
What actually correlates with being cited?
Two things stood out in 2026 research. Pages introducing their topic within the first 100 words were cited at 22.1% against 8.7% otherwise. Pages serving eight or more related queries reached the AI top ten 43.3% of the time against 15.5% for single-query pages.
Should I write shorter content to get into summaries?
No. Length showed almost no relationship with citation position, and the median cited page was 2,681 words. Answer early and cover the question properly. Shortening a page to look summary-shaped does not help.
Sources
Every figure on this page comes from one of the following. Where a claim is ours rather than a source’s, it carries an evidence label instead.
- Ahrefs — AI Overview citation analysis, March 2026. Source for the fall in top-10 organic share of citations from roughly 76% to 38%. Independent study
- Rankability — State of AI Search, 1,645 AI citation observations across 31 topics, 2026. Source for the 24.1% cross-platform overlap, the 73.5% ChatGPT exclusive citation rate, the 22.1% vs 8.7% first-100-words finding and the 43.3% vs 15.5% query-breadth finding. Independent study
- Google Search Central — Optimizing your website for generative AI features (published 15 May 2026, updated 10 July 2026). Source for the llms.txt, chunking and AI-schema positions, and for the statement that generative features run on core ranking systems. Primary
- Google I/O announcements, 19 May 2026. Source for AI Overviews at approximately 2.5 billion monthly active users and AI Mode at approximately 1 billion. Primary
- Searlixa sampling across client engagements. Any figure drawn from this carries an Observed label with its sample size and date. Our own
Ask an assistant
Hand this page to an assistant and ask it to summarise or challenge the argument. If it gets something wrong about us, that is itself a useful finding.