Being cited in Google's AI Overview is usually described as the goal of GEO work. We thought so too. Then we measured it on our own site: the 21 queries where we were cited brought 1,252 impressions and 3 clicks in 28 days.
This article shares that measurement with its method, raw numbers and limits. It is one site, and one agency's own queries. Read it as a case, not a general rule.
How we measured
- Queries: 60 queries chosen from those with at least 20 impressions and an average position of 20 or better in the last-28-days Search Console data we pulled on 27 August 2026. 52 land on our service, sector and location pages and 8 on six blog posts. 49 are Turkish and 11 English, across 30 different pages.
- Results page: on 29 August 2026 we captured Google's results page for each query on desktop, to a depth of 10 organic results, and recorded whether an AI Overview appeared and which domains it cited.
- Join: we matched each record to that query's Search Console impressions, clicks and average position.
We had to combine the two sources. According to Google's documentation on AI features, appearances in these features are reported in Search Console inside the "Web" search type, as part of overall search traffic. The document describes no separate filter for them. So Search Console alone cannot tell you whether a query showed an AI Overview.
Result: three groups, almost no clicks in any of them
An AI Overview appeared on 46 of the 60 queries (76.7%). We were among the cited sources on 21 of those and absent on 25. The remaining 14 queries showed no overview.
| Status | Queries | Impressions | Clicks | Average position* |
|---|---|---|---|---|
| AI Overview present, we were cited | 21 | 1,252 | 3 | 6.2 |
| AI Overview present, we were not cited | 25 | 1,182 | 1 | 14.3 |
| No AI Overview | 14 | 827 | 0 | 14.9 |
| Total | 60 | 3,261 | 4 |
* Search Console average position, weighted by impressions.
The click-through rate in the cited group was 0.24%. For comparison, the whole site's rate in the same 28 days was 0.46% (110 clicks, 23,914 impressions). The two rates do not measure the same thing, because the site total includes brand searches. Still, being cited did not lift these queries above the site average.
One more detail: where an overview appeared, it cited 5.3 different domains on average (minimum 2, maximum 6). Being a source means being one of several links in that list.
The clearest example
One English query: meta ads policies for botox and aesthetic treatments. It had 548 impressions in 28 days at a Search Console average position of 3.7. When we captured the results page we ranked third, an AI Overview was present, and it cited us. Clicks: zero.
We had already rewritten that page's title and description more than once. The click-through rate never moved. At the time we recorded the reason as "the overview absorbs the click because we are not cited". The measurement disproved the second half of that sentence: we were cited.
17 September: the same picture
We repeated the measurement on 17 September 2026 with 44 queries. An AI Overview appeared on 31 (70.5%) and cited us on 15. On 11 of those we also ranked in the top 10 of the live results page.
One of the 11, the English query above, recorded an unusual number of impressions in that period (4,964) and 1 click; we do not interpret a jump of that size. The other 10, all Turkish, looked like this:
| Query | Live position | Impressions (28 days) | Clicks |
|---|---|---|---|
| google reklamları izmir | 1 | 22 | 0 |
| google adwords izmir | 2 | 21 | 0 |
| saç ekimi reklam ajansı | 2 | 29 | 0 |
| google ads izmir | 3 | 60 | 0 |
| saç ekimi reklamları | 3 | 20 | 0 |
| google ads ajansı izmir | 4 | 48 | 0 |
| izmir google ads ajansı | 4 | 42 | 0 |
| dental turizm reklam ajansı | 6 | 50 | 0 |
| izmir seo fiyatı | 7 | 55 | 0 |
| medikal turizm reklam ajansı | 8 | 77 | 0 |
| Total | 424 | 0 |
These are purchase-adjacent searches that combine a city or a specialty with the Turkish words for ads, agency and price. We rank on page one, the overview cites us, and nobody clicks.
What this data shows and does not show
Keep in mind while reading
- It shows: on this site and for these queries, being cited did not by itself bring clicks.
- It shows: our position was already much better on the queries where we were cited (6.2 against 14.3). In our data the overview appears to draw its sources largely from pages that already rank well.
- It does not show: causation. We do not know how many clicks we would have received without the overview. The 14 queries with no overview also had zero clicks; there the likely explanation is position (14.9).
- It does not show: value beyond the click. Someone who reads our name in an overview and comes to the site directly later is invisible in this table.
- Limits: one site, an agency's business-buyer queries, between 20 and 548 impressions per query, desktop only, a single-day snapshot. Search Console withholds some low-volume queries for privacy.
Google states in its own documentation that it has seen clicks from results pages with AI Overviews to be higher quality. Our measurement counts clicks; it says nothing about their quality, so the two statements do not contradict each other.
What we changed as a result
These are decisions we took after reading the data, not proven fixes.
- Success metric: we stopped treating "number of queries where we are cited" as a goal on its own. We now track citation status next to impressions and clicks for each query.
- Three separate situations: if an overview exists and we are not cited, the question is whether our page contains a clear answer that can be quoted. If we are cited and get no clicks, the question changes: what does the page offer that a summary cannot? If there is no overview, it is ordinary ranking work.
- For the second situation: we are adding things a summary cannot replace to those pages: a calculator, a lookup tool, downloadable data. That test is running; we will not claim it works before we see the result.
How to repeat this on your own site
- Export the last 28 days of query and page data from Search Console. Keep queries with at least 20 impressions and an average position of 20 or better.
- Capture the results page for each query on the same day and the same device type. Note whether an AI Overview is present and whether your domain is in its source list. This can be done by hand; if you use a SERP API, make sure the overview is fully loaded, because the API we used returned it empty unless we requested it explicitly.
- Split the queries into the three groups and calculate total impressions, clicks and weighted position for each.
- Repeat a few weeks later. A single day can mislead.
For the technical side, which crawler controls what, see our guide to AI crawler controls. If GEO is new to you, start with what GEO is and our AI search optimization service page.
Sources
- Google Search Central, AI features and your website (accessed 11 November 2026).
- Our own measurements: 29 August 2026 (60 queries) and 17 September 2026 (44 queries); Search Console data are the last-28-days pulls made on 27 August and 17 September 2026 respectively.

