The drop that rank tracking cannot see
There is a kind of traffic drop that looks impossible when you first see it. Your rankings hold. Your impressions hold, or even rise. Your clicks fall anyway, month after month, and nothing in a rank tracker explains it.
What has usually happened is that an AI Overview now appears above the results for the queries that used to send you traffic. Google generates a summary, cites a few sources, and a share of searchers get what they came for without clicking anything. Your page is still ranking exactly where it was. Fewer people reach it.
This is a visibility problem, not a ranking problem, and it needs a different kind of diagnosis. Two checks confirm it. In Google Search Console, filter the Performance report to your informational queries and compare the last 28 days with the same period before the drop: impressions steady, click-through rate down, position unchanged is the signature. Then run a handful of those queries yourself and see whether an AI Overview appears, and whether your page is one of the sources it cites. If the Overview is there and you are not in it, you have found the drop. Our rank tracking guide covers why position-only tracking misses this, and if the loss is material to your business, the section on recovery at the end of this guide is where to go next.
What Google AI Overviews are and where they appear
AI Overviews are Google’s generated summaries that appear at the top of some search results pages, above the traditional listings. Google describes them in its documentation on AI features in Search: the summary is produced by Google’s models from information in its index, and it shows links to the pages that support it, so people can read more from the sources.
They appear most often for questions, comparisons, how-to queries, and anything where a synthesised answer is useful. They appear less for navigational searches and for queries where the searcher clearly wants a specific site. The share of queries that trigger them has moved over time and varies by topic, which is why the check above is done on your own queries rather than on averages. See the AI Overviews glossary entry for the short definition.
How AI Overviews pick their sources
Google’s documentation is clear that there is no special markup and no separate submission for AI Overviews. The same systems that rank pages in the results feed the Overview, and the same best practices apply. In practice, four things separate pages that get cited from pages that rank but do not.
First, the page is already in the candidate set. Overviews draw on pages that rank well for the query or closely related queries, so a page on page four is rarely cited however good it is.
Second, the answer is extractable at passage level. A clear, direct answer to the question, near the top of the relevant section, in a form a model can lift cleanly. Long preambles and answers buried in the middle of a paragraph get skipped.
Third, the page is technically clean and machine-readable. Crawlable, indexable, fast, with structured data that describes what the page is and who wrote it. Our technical SEO guide covers the foundation.
Fourth, the entity signals are consistent. The author, the organisation, and the topic are identifiable across the site and across the web, which is the E-E-A-T side of the same coin. Pages from sources that Google can identify and trust are cited more readily than anonymous ones.
How to measure whether you are cited
Search Console does not report AI Overview citations separately; clicks and impressions from an Overview are folded into the normal Web performance data. So measurement is partly manual, and it is worth doing weekly for the queries that matter.
Build a short list of your priority questions, the ones that drove traffic before and the ones you want to own. Run each one, record whether an AI Overview appeared, whether your page was cited, and which sources were. Do it from a consistent location and device, and log it. Over a few weeks that log becomes a citation-presence metric, and the competitor names in it tell you who is being trusted for your topic.
The metric to track is share of model: how often you are cited across your priority questions, relative to competitors. The share of model glossary entry defines it, and our AI visibility tracking guide covers the method across Google, ChatGPT, and Perplexity together, since the same pages tend to get cited across all of them.
How to earn a citation for a query you already rank for
The fastest gains come from pages that already rank on page one but are not cited. They are in the candidate set; they are losing on extractability or trust. The edits, in order of effect:
- Put the answer first. Under the heading that matches the question, open with the direct answer in one or two sentences, then expand. A model looking for a passage to cite finds it immediately.
- Use the question as the heading. If people ask “how do I get cited in AI Overviews”, a heading that says that is easier to match than a clever one.
- Add a definition block for the core term. One clean sentence that defines the thing, near the top, in the form “X is …”. Definitions are cited constantly.
- Add first-hand evidence. A number from your own data, a specific example, a dated observation. Pages that restate what is already available are the ones Overviews summarise without crediting anyone.
- Fix the dates and the author. A visible, true “updated” date, a named author with a profile, and Article structured data that carries both. Stale and anonymous pages are cited less.
- Tighten the technical layer. Confirm the page is indexed, loads quickly, and has no crawl or render issues; a page Google cannot read cleanly cannot be summarised.
Then re-check the query the following week, after Google has re-crawled the page. Citation changes follow crawls, not edits.
What you cannot control
Some of this is outside your hands and it is better to know which parts.
Google decides which queries trigger an Overview, and that changes. A query that shows one today may not next month, and the reverse.
Google’s models change, and with them the way passages are selected. A citation earned in one month can be lost in the next without anything changing on your page.
Google’s controls for limiting snippets, such as the nosnippet and max-snippet directives, limit what can be shown from your page in Overviews as well as in regular results. There is no separate switch that keeps a page in normal results while removing it from Overviews, so the choice is between being a source and limiting your visibility more broadly.
The Overview’s own behaviour decides how many people click through. For some queries a well-cited source still receives clicks from people who want the full detail; for others the summary satisfies the searcher. You can measure which you have; you cannot change it.
When the click loss is a recovery job
If a material share of your organic clicks has moved into Overviews you are not cited in, treat it as a recovery: establish the baseline, list the queries and pages carrying the loss, do the citation work above on those pages, and measure citation presence weekly until it moves. It is the same discipline as recovering from an update, which is why our algorithm updates guide and penalty recovery guide use the same dated, page-level method.
Two routes if you want it done with you. Our Google penalty and core update recovery service covers the case where rankings held and clicks fell: it diagnoses the loss from your own data, including the AI Overview displacement most agencies cannot see, and does the work to get the affected pages cited. Our AI visibility monitoring service runs the measurement side as an ongoing programme, and our GEO service covers the optimisation work across every AI engine, not only Google.
Common AI Overview mistakes
- Blocking Google’s crawlers to “opt out”. That removes the page from Search entirely, not just from Overviews.
- Rewriting pages for keywords instead of answers. Overviews cite passages that answer; density does nothing.
- Measuring rankings only. A page can hold its position while its clicks disappear, and a rank tracker will call that stable.
- Judging citation changes the day after an edit. Citations follow crawls; give it a week and check the crawl date.
- Chasing every query. Work the priority questions that carried traffic; the long tail follows the pages that earn trust.
Frequently asked questions
Why did my traffic drop but my rankings did not?
Most often because an AI Overview now answers the query above your result, and some searchers no longer click through. Check Search Console for steady impressions with falling click-through rate on informational queries, then run the queries yourself to see whether an Overview appears and whether you are cited in it.
How do I get my content cited in Google AI Overviews?
Rank for the query, then make the answer extractable: direct answer first under a question-shaped heading, a clean definition, first-hand evidence, a named author with a true date, and a technically clean page. Re-check after Google has re-crawled the page.
How do I know if AI Overviews are taking my clicks?
Search Console does not separate Overview clicks out, so use two signals together: impressions and position steady while click-through rate falls on your informational queries, and a manual check of those queries showing an Overview you are not cited in.
Do AI Overviews reduce my clicks?
They can, particularly on informational queries where the summary satisfies the searcher. Being one of the cited sources protects your visibility and can still send qualified clicks from people who want the full detail, which is why citation presence is the metric to manage.
Can I opt out of AI Overviews?
Not while staying in normal results. Google’s snippet controls limit what can be shown from your page in Overviews and in regular results alike, and blocking crawlers removes the page from Search. The practical strategy is to be the source that gets cited so the feature works in your favour.
Do AI Overviews use the same ranking as regular results?
They draw on the same systems and index, and Google says the same best practices apply, but citation is decided at passage level. A page can rank without being cited if its answer is not extractable, and a lower-ranking page with a clearer answer can be cited instead.
How is an AI Overview different from a featured snippet?
A featured snippet lifts text from one page. An AI Overview generates a summary and can cite several sources at once, so the goal is to be one of those sources rather than to win a single box.
Part of our complete GEO guide.
Rankings holding but clicks falling? Our SEO recovery service diagnoses AI Overview displacement from your own data and gets your pages cited. Want to appear in AI answers for your key topics from the start? Explore our AI Search SEO service or start with a free SEO audit.