Google is testing a narrower citations panel inside AI Overviews, the second visual de-emphasis of AI Overview links reported in 2026. No click data exists for either test yet — and none will arrive through Google's own tooling, because the Search Console report that covers AI surfaces measures impressions only. The practical response is not to chase a layout you cannot see, but to measure the citation as an impression and stop pricing it as a click.

Key takeaways

  • Search Engine Roundtable reported a thinner AI Overview citations panel on 8 September 2026, following a separate test in March 2026 that swapped the citation cards' blue background for white.
  • Pew Research Center found users clicked a source link inside an AI summary on just 1% of visits, and clicked any result on 8% of visits with an AI summary versus 15% without.
  • Ahrefs measured position-1 clickthrough on AI Overview keywords falling from 0.073 to 0.016 between December 2023 and December 2025 — a 58% drop, up from the 34.5% the same team measured a year earlier.
  • Google's position is that "total organic click volume from Google Search to websites has been relatively stable year-over-year" and that AI Overviews mean "people are seeing more links on the page than before."
  • Search Console's Generative AI performance report went worldwide on 31 August 2026 and reports impressions only — no clicks, no CTR — for AI Overviews and AI Mode.

What is actually changing in the AI Overview citations panel?

Search Engine Roundtable reported on 8 September 2026 that Google is testing a visibly narrower version of the citations panel that sits alongside an AI Overview. The publication replicated the layout itself. Nothing else about the panel — its position, the number of cards, the sources chosen — was reported as changing.

That is the whole factual base. There is no announcement, no rollout percentage, and no measurement of what the narrower panel does to clicks. Treat it as an experiment sighting, not a shipped product.

It is worth logging, though, because it is not isolated. On 30 March 2026 the same publication reported Google testing AI Overview citation cards on a white background instead of the blue shading that had previously separated them from the rest of the page. Two tests, roughly five months apart, both reducing how much visual weight the citation block carries.

Date reported Test Direction
30 March 2026 Citation cards moved from blue background to white Removes the visual boundary marking citations as distinct
8 September 2026 Citations panel rendered noticeably thinner Reduces the panel's share of screen width

Neither test has a published click outcome. What we can say is that the direction of both is consistent, and that direction is toward less prominence.

Do people click AI Overview citations at all?

Rarely — and this is the number that reframes the entire design question.

Pew Research Center analysed the browsing of 900 U.S. adults across 68,879 Google searches during March 2025, of which 12,593 produced an AI summary. Users clicked a link within the AI summary on 1% of visits. Not 1% of clicks — 1% of visits to a page carrying an AI summary.

The same study found the broader pattern:

Behaviour With AI summary Without
Clicked any traditional search result 8% of visits 15% of visits
Clicked a link inside the AI summary 1% of visits —
Ended the browsing session on that page 26% 16%

Pew also found around one in five Google searches in March 2025 produced an AI summary, and 58% of respondents ran at least one search that month that triggered one.

Here is the uncomfortable implication for anyone alarmed by a narrower panel: a citation block that already converts at 1% has very little left to lose. Shrinking a panel from a 1% click rate to something lower is a change measured in basis points. The traffic damage attributed to AI Overviews is overwhelmingly happening in the other row of that table — the drop from 15% to 8% on ordinary organic results pushed down the page — not in the citation panel itself.

That distinction matters because it changes what you should optimise. Fighting for a bigger citation card is fighting for the smallest number on the page.

How much have clickthrough rates actually fallen?

Independently of panel design, the measured decline is steep and it is getting steeper.

Ahrefs compared aggregated Search Console clickthrough rates across 300,000 keywords — 150,000 that trigger AI Overviews and 150,000 informational keywords that do not. In the first pass, comparing March 2024 with March 2025, position-1 CTR on AI Overview keywords was 34.5% lower. In the updated run comparing December 2023 with December 2025, average position-1 CTR on those keywords fell from 0.073 to 0.016 — a 58% decline, against a forecast of 0.037 had AI Overviews not appeared.

The control group matters here. Informational keywords without AI Overviews also declined over the same window, from 0.076 to 0.039. Search clickthrough is falling generally; AI Overviews accelerate it rather than solely causing it. Any analysis that attributes the full drop to AI Overviews is overstating a real effect.

What does Google say, and where does it conflict?

Google's public position is not that clicks are falling. Its stated claims are:

  • "Total organic click volume from Google Search to websites has been relatively stable year-over-year."
  • With AI Overviews, "people are seeing more links on the page than before."
  • "Average click quality has increased and we're actually sending slightly more quality clicks to websites than a year ago (by quality clicks, we mean those where users don't quickly click back)."

These claims and the third-party CTR studies are not strictly contradictory, and the reason is worth understanding. Google is describing aggregate click volume across all of Search, buoyed by query growth — it notes that with AI Overviews "people are searching more and asking new questions that are often longer and more complex." Ahrefs and Pew are describing per-query clickthrough rate. Total clicks can hold flat while every individual query converts worse, provided there are enough more queries. Both can be true; they answer different questions.

For a single site, the per-query rate is the one that governs your traffic, because your impressions are not growing at Google's aggregate rate.

The gap nobody is talking about: you cannot measure this

This is the part the design-test coverage consistently misses.

On 31 August 2026 Google finished rolling out the Generative AI performance report in Search Console to all sites worldwide. It breaks out AI Overviews and AI Mode from classic Search results — genuinely useful, and the first native visibility into AI surfaces publishers have had.

It reports impressions only. Google's own documentation defines the metric as how many times links to your site were shown to a user in a generative AI feature on Search. There is no clicks column, no CTR, and no position for these surfaces.

Sit that next to the citation-panel tests and the problem is exact:

Question Can Search Console answer it?
Was my site cited in an AI Overview? Yes — impressions
On which pages, countries, devices, dates? Yes
Did anyone click that citation? No
Did the thinner panel reduce my citation clicks? No

A citation-panel redesign can only plausibly affect clicks. Clicks are the one metric Google does not report for AI surfaces. So the industry will keep writing about layout tests it cannot evaluate, and site owners will keep guessing, until either third-party panel data catches up or Google adds the column.

Other documented limits on the report are worth knowing before you build a dashboard on it: it excludes Search Labs experiment data, applies the standard 1,000-row table cap, marks the newest data as preliminary and subject to change, and uses Pacific Time for all dates.

Should you change anything because of a narrower panel?

No — and Google's own documentation says as much about AI features generally: "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary." It also states plainly that you do not need to create new machine-readable files, AI text files, or markup.

Two things in that documentation do carry operational weight:

There is no separate opt-out through snippet controls. AI is built into Search, so robots.txt, noindex, nosnippet, data-nosnippet and max-snippet govern AI features and classic Search together. You cannot rank in one and withhold from the other through those controls. Google has since introduced a distinct Search generative AI control that lets a site decline to appear in and ground generative AI features — and a site using it will receive no traffic or impressions from those features, which is a real cost, not a free hedge.

Citation share is concentrated. Ahrefs' September 2026 snapshot across more than 3 million U.S. queries shows the top of the AI Overview citation distribution is dominated by platforms, not publishers:

Domain Share of top-50 citations
youtube.com 22.9%
reddit.com 18.5%
facebook.com 10.1%
google.com 8.8%
instagram.com 5.6%
en.wikipedia.org 4.0%

Roughly two-thirds of the visible citation mass in that top tier goes to user-generated and platform properties. That is a far larger structural constraint on your citation traffic than the width of the panel those citations appear in.

The honest counterweight: if your presence in AI Overviews is meaningful, the leverage is in being the kind of source that gets cited on queries your competitors own — and in making sure the crawlers that build these answers can reach you at all. Both are checkable today, unlike panel width.

What to do instead

  1. Track citation impressions as a distinct channel. The Generative AI report separates AI surfaces from organic. Baseline it now; there is no historical backfill to recover later.
  2. Stop modelling AI citations as a traffic source. At a 1% in-summary click rate, they are a brand-visibility surface. Value them accordingly and set expectations with stakeholders before a quarterly review does it for you.
  3. Confirm the AI crawlers can actually fetch your pages. A citation you never earn is a bigger loss than a citation rendered in a narrower box.
  4. Separate the two declines. Your AI Overview citation rate and your ordinary organic CTR move for different reasons. Reporting them as one number hides which problem you have.
  5. Ignore unreplicated layout tests in planning. Most never ship. Log them; do not restructure around them.

Frequently asked questions

Is Google's thinner AI Overview citations panel live for everyone?

No. It was reported as a test on 8 September 2026 by Search Engine Roundtable and replicated by that publication, but Google has not announced a rollout and no coverage percentage has been published. Treat it as an experiment sighting rather than a shipped change.

Will a narrower citations panel reduce my traffic?

There is no published data either way, and the likely effect is small. Pew Research Center found users click a link inside an AI summary on only 1% of visits, so the citation panel is already a very low-conversion surface. The larger traffic effect sits in ordinary organic results, where clicks fell from 15% to 8% of visits when an AI summary was present.

Can I see how many clicks my AI Overview citations get?

Not in Search Console. The Generative AI performance report, rolled out worldwide on 31 August 2026, reports impressions only for AI Overviews and AI Mode — how often links to your site were shown in a generative AI feature. It provides no clicks, CTR, or position data for those surfaces.

How much have AI Overviews reduced clickthrough rates?

Ahrefs measured average position-1 CTR on AI Overview keywords falling from 0.073 in December 2023 to 0.016 in December 2025, a 58% decline across 300,000 keywords. An earlier run of the same study, comparing March 2024 with March 2025, found 34.5%. Informational keywords without AI Overviews also declined over the period, so AI Overviews accelerate a broader trend rather than solely causing it.

Does Google agree that AI Overviews cut clicks?

No. Google states that total organic click volume from Search to websites has been relatively stable year-over-year, that people see more links on the page than before, and that it sends slightly more quality clicks than a year ago. Those claims describe aggregate volume across a growing number of queries; the CTR studies describe per-query rates. Both can hold at once.

Can I appear in Google Search but opt out of AI Overviews?

Not through standard snippet controls. Google states that AI is built into Search, so nosnippet, data-nosnippet, max-snippet and noindex apply to both. A separate Search generative AI control does let a site decline to appear in generative AI features, but a site using it receives no traffic or impressions from them.

Measure the surface you can actually see

Panel width is not a metric you can move or verify. Crawler access and citation impressions are. Start by confirming the AI crawlers can reach your pages with the free crawler checker, then set up reporting that separates AI visibility from organic clicks — our guide to AI search KPIs covers which numbers deserve a dashboard and which are noise.

For more on how Google keeps reshaping the AI Overview itself, read Google's AI Overviews Now Expand Themselves, and for the one citation lever Google has actually handed publishers, see Google's Preferred Sources Button.

Sources