Field note

Ranking #2 with zero clicks: the AI Overview effect

Six queries on one site rank between positions 1.8 and 9 with zero clicks in 28 days. The Search Console numbers, what sits above them, what we did.

SK Semih Kavaklıoğlu·7 Sept 2026·5 min read Share

A how-to guide on one of our own sites ranks at position 1.8 for a Dutch query about Amazon product photo requirements. In 28 days it was shown 35 times and clicked zero times. Five more queries on the same site show the same pattern. This note is what the numbers say, what sits above the result, and what we decided to do.

What did Search Console show?

Answer. Six queries rank on page one and received no clicks in 28 days, while the site as a whole grew.

Six queries, page one, zero clicks (bar = expected clicks, 28 days)
amazon productfoto eisen
~5 → 0position 1.8 · 35 impressions
bol.com productfoto eisen
~4 → 0position 2.9 · 28 impressions
bol productfoto eisen
~3 → 0position 3.6 · 24 impressions
zdjęcia mebli na białym tle
~2 → 0position 5.1 · 34 impressions
bol.com productfoto
~1 → 0position 7.5 · 40 impressions
fotografia aranżacyjna mebli
~1 → 0position 9.0 · 22 impressions
Source: Google Search Console, omniroom.ai, 28 days to 4 September 2026.

"Expected clicks" is impressions multiplied by a typical click-through rate for that position (about 14% at position 2, about 5% at position 5). It is a model, not a measurement; it exists to say how far from normal zero is.

The site-wide numbers for the same window moved the other way.

Clicks
145
▲ +31%, was 111
Impressions
2,632
▲ +338%, was 601
Position
11.1
▲ from 13.2
Non-brand clicks
11
▲ +10, was 1

Brand clicks fell from 91 to 77, and non-brand impressions went from 191 to 895.

Why does a page at position 2 get nothing?

Answer. Something above position 1 answers the question, and for these six queries that something is an AI Overview.

amazon productfoto eisenwhat the searcher sees
AI Overview

White background, at least 1000 pixels on the longest side, the product filling 85% of the frame, no text or logos on the main image.

omniroom.ai › nl › playbook
Productfotografie voor e-commerce: de gids voor 2026
Position 1.8, 35 impressions, 0 clicks. The Overview answers the query in full above the first result.

We checked the live result page for each query. All six carry an AI Overview block above the organic results. A Dutch seller asking what Amazon requires of a product photo gets the answer (white background, 1000 pixels on the longest side, the product filling 85% of the frame) without leaving Google. The organic result at position 1.8 is visible, and skipped.

The CTR drop at site level has the same shape. The 28-day CTR fell from 18% to 6%, but that is impressions growing four times faster than clicks: new generic queries at lower positions, plus the six above. It is not a title problem. We had shortened 32 titles on 4 September, and the first reading suggested that made things worse. The per-query view says otherwise: the pages with shortened titles rank higher and are seen more; the ones with no clicks are the ones with an AI Overview on top.

Is AI traffic bad for this site, then?

Answer. No. The same 28 days show AI assistants sending visits that convert, while AI Overviews absorb clicks on informational queries. The two are different channels.

AI assistants as referrer
sessions
83, was 58
key events
41, was 1
chatgpt.com alone
71 sessions, 39 key events
Organic search, same 28 days
sessions
204
key events per session
lower than the assistant referrals
six how-to queries
0 clicks, Overview on top

Sessions referred by AI assistants (chatgpt.com, claude.ai and similar) went from 58 to 83 in GA4. Those 83 sessions produced 41 key events; the previous period's 58 produced 1. ChatGPT alone sent 71 sessions and 39 key events. Organic search sent 204 sessions in the same window.

So on this site the assistant is a referrer worth more per session than Google, and the Overview is a competitor for the click on how-to queries. Content decisions have to be made per query, not per channel.

What is worth doing about the six queries?

Answer. Become the source the Overview cites, or stop investing in the query. We chose the first, for the Dutch guide only, as a one-hour edit.

Become the cited sourcechosen
  • One plain sentence under the question heading
  • A table with the platform's actual requirements
  • Named source document
Stop investing
  • Six queries, 22 to 40 impressions each
  • Under 200 impressions a month combined
  • 15 to 20 visits even at a normal CTR

In more words:

  1. Become the source the Overview cites. Open the matching section with one plain sentence that answers the query, add a table with the platform's actual requirements (pixel size, background, frame coverage, file format) and name the source document. Overviews quote pages structured like that; they paraphrase pages that bury the answer in a paragraph.
  2. Stop investing in the query. Six queries at 22 to 40 impressions each add up to fewer than 200 impressions a month. Even at a normal click-through rate that is 15 to 20 visits.

We are doing option 1 for the Dutch guide only, as a one-hour edit, and leaving the Polish one alone until a native speaker checks it. One complication: the Dutch page is inside a 30-day measurement window for the title change, so it comes out of that cohort. Changing a page you are measuring and keeping it in the measurement is how you end up trusting a number that means nothing.

How do you find this on your own site?

Answer. Look for queries in the top 10 with real impressions and near-zero clicks, then look at the result page.

In Search Console, open Performance, filter to the last 28 days, sort by impressions, and scan for rows with position under 10 and clicks at or near zero. For each one, search the query in a private window and look at what sits above the first result. If it is an AI Overview, a shopping carousel or a video block, more ranking work will not bring the click back; the page has to become the thing the block cites, or the query is not yours.

Moonleap does this automatically when Search Console is connected: it flags queries in the top 10 with under a quarter of the expected clicks, checks the SERP for the feature above them, and raises the AI Overview ones as findings so the loop with your agent applies to them too.

FAQ

Does an AI Overview always kill the click?

No. For queries where the Overview is incomplete or the reader needs the source (prices, specifications, downloads), cited pages still get visits. For a short factual how-to, the Overview is usually the last stop.

Should I block Google's AI crawler to avoid this?

Blocking Google-Extended affects Gemini training data, not AI Overviews, which use the normal Googlebot index. The only way out of an Overview is to be cited in it or to rank for queries it does not answer.

How long before an edit shows in Search Console?

Search Console data lags by two to three days, and Google needs to recrawl the page. Read the effect at 28 days after the change, not before.

Sources

  1. Google Search Console, property omniroom.ai, 28 days to 4 September 2026 versus the 28 days before (our own export)
  2. SERP feature check for the six queries, 6 September 2026 (DataForSEO SERP API, cached in Moonleap)
  3. Google, "AI features and your website" https://developers.google.com/search/docs/appearance/ai-features
  4. Google Analytics 4, property for omniroom.ai, sessions by source, same 28 days

Share the number

0

clicks from 35 impressions at position 1.8, in 28 days

moonleap.io · 7 Sept 2026Post

Related