Shopping ads under AI Overviews: impressions down, CTR up about 20%

Shopping ads are appearing on fewer searches and getting clicked more often when they do appear, according to benchmark analysis published on 7 September 2026 by Mike Ryan, head of ecommerce at Smarter Ecommerce. Median Shopping ad impressions per advertiser account fell from about 1.85 million to about 1.4 million between mid-2025 and mid-2026, while median click-through rate rose from 1.20% to nearly 1.55%, an increase of roughly 20% year over year.

Ryan reads those two lines as one event seen twice. His argument is that Google is showing Shopping ads on fewer, higher-intent queries and serving AI Overviews on the lower-intent ones, so a rising CTR describes where the ads are being placed rather than how well they are performing. Search Engine Land reported the analysis, which draws on thousands of Shopping and Performance Max campaigns across hundreds of advertiser accounts and roughly 175 billion impressions.

What the Smarter Ecommerce benchmark actually measured

Mike Ryan published the benchmark on 7 September 2026 in his role as head of ecommerce at Smarter Ecommerce. The dataset behind it covers thousands of Shopping and Performance Max campaigns running across hundreds of advertiser accounts, and roughly 175 billion impressions in total. The comparison window runs from mid-2025 to mid-2026, so it is a full twelve months of movement rather than a single-month snapshot that a seasonal swing could explain away.

Two medians carry the whole story. Median Shopping ad impressions per account moved from about 1.85 million to about 1.4 million. Median CTR moved from 1.20% to nearly 1.55%. The word median is doing quiet work there: these are typical accounts in the sample, not the sum of a handful of very large advertisers whose spend would otherwise drown out everyone else. That is the reason the pattern is worth a mid-sized advertiser's attention at all.

Optmyzr separately reported a 17% year-over-year increase in Shopping CTR. It is a different number from a different dataset, and the useful part of it is not the second decimal place but the agreement in direction. Two independent measurements of roughly the same period both show Shopping CTR climbing, which makes it harder to write the move off as one company's sampling quirk.

Why Ryan calls it the inverse of the crocodile effect

Ryan wrote on 7 September 2026: "Ever since AIOs launched, publishers have been gaining impressions and losing clicks: the dreaded SEO 'crocodile effect.' For Shopping ads, it's the opposite." The comparison carries real information. On the organic publishing side, in the shape he describes, the impression count keeps growing while the click count refuses to follow it. On the Shopping side the arrow points the other way: impressions shrink while the click rate climbs.

He also stated the conclusion in plain language: "Shopping ads aren't necessarily generating significantly more clicks. They're being shown less often, which is pushing CTR higher." That is the sentence worth pinning to the wall, because it pulls a rate apart from a volume. A rate can improve at the same moment the volume underneath it is falling, and the two facts sit together without contradiction.

The mechanism: a smaller query set lifts CTR with no ad improving

CTR is clicks divided by impressions. Shrink the denominator, hold the numerator roughly steady, and the ratio rises by itself. No headline was rewritten, no bid was raised, no product title was fixed. That arithmetic is what Ryan's interpretation rests on, and it is why he says the ads are not necessarily generating significantly more clicks than before.

The second half of the mechanism is about which impressions went missing. Ryan's reading is that Google has pulled Shopping placements off the browsing-style queries where a shopper rarely clicks a product tile, while keeping them on queries where someone already knows roughly what they intend to buy, with AI Overviews served on the lower-intent side instead. If the impressions that disappeared were the ones that historically converted to clicks least often, then removing them raises the average even when every remaining impression behaves exactly as it did a year ago.

This is why a CTR figure on a monthly report can quietly mislead. Two very different accounts produce the same headline improvement. In the first, a merchant rewrote product titles, tightened the feed and gained genuine engagement. In the second, Google simply stopped showing that merchant's ads to the least interested slice of the audience. The report line looks identical in both. The business outcome does not, and only the impression and click columns beside the CTR column can tell them apart.

There is a reporting habit underneath this worth naming. CTR is popular in client decks because it is a tidy percentage that moves in a satisfying direction. Impressions and absolute clicks are less flattering and more informative. When the supply of impressions is itself changing, a percentage built on that supply stops being a performance metric and starts being a description of the auction.

The figures side by side

The table below holds only the numbers reported in the analysis: the two medians Ryan published on 7 September 2026 and the corroborating figure from Optmyzr.

What to check in your own Google Ads account

The benchmark describes a sample of accounts, not yours. Before anyone repeats the pattern in a client meeting, the account itself should be asked whether it agrees. Five columns settle it, and all of them are already in the interface.

• Absolute clicks over twelve months, sitting next to CTR. If clicks are flat or down while CTR is up, the account is showing the same shape the benchmark describes. If clicks are genuinely up, something in the account is working and the CTR gain is earned.
• Shopping ad impressions across the same period. A falling impression count is the first half of the pattern. Pull it monthly rather than as a single year-on-year total, because a step change looks very different from a slow drift and points at different causes.
• Impression share lost to rank versus impression share lost to budget. These two answer different questions. Budget loss means the account chose to stop bidding. Rank loss means the auction decided. Neither of them is the same as Google narrowing where the format is eligible to appear at all, which is why the raw impression count still matters alongside both.
• Conversion volume rather than conversion rate. Conversion rate inherits the same arithmetic problem as CTR when the traffic mix shifts toward higher intent. Orders, leads and revenue do not.
• Cost per click and total spend. If impressions fell while spend held, the money moved somewhere. Knowing which campaigns absorbed it is more useful than knowing the blended CTR improved.

None of this requires new tooling. It requires reporting the volume metrics with the same seriousness as the rate metrics, which is a habit rather than a project. For accounts running Shopping or Performance Max at scale, this kind of diagnostic belongs inside routine Google Ads management rather than in a special project.

What the analysis did not establish

It is a correlation observed across a twelve-month window. Impressions fell and CTR rose during the same period. The analysis as reported does not demonstrate that the first caused the second, and a benchmark of this shape is not built to.

The causal explanation belongs to Ryan. His statement about Google showing ads on fewer, higher-intent queries while serving AI Overviews on lower-intent ones is his reading of what the numbers imply. It was not attributed to Google, and no Google confirmation appears in the reporting.

The analysis also does not isolate AI Overviews from everything else that shifted in the ad auction between 2025 and 2026. A twelve-month window in Google Ads contains many changes at once, and the benchmark as reported does not separate the effect of one from the others.

What it does establish is still worth keeping. Across a large sample of accounts, the direction of Shopping impressions and the direction of Shopping CTR came apart, and a second dataset from Optmyzr moved the same way. That is enough to change how a report is read. It is not enough to rebuild a strategy around.

What this means for Thai marketers

The analysis does not break its results out by country, and nothing in it describes the Thai market specifically. Anyone presenting it locally should say so first, because the honest version of this story is more useful than an inflated one.

The reporting habit still travels. Thai ecommerce advertisers running Shopping and Performance Max through a Google account are looking at the same interface, the same metrics and the same monthly review cycle. If a Bangkok merchant's dashboard shows CTR up and orders flat, the benchmark supplies a plausible reason to look at impressions before congratulating the creative team, and it supplies the vocabulary to explain the gap to a client who saw only the green arrow.

There is a budgeting consequence too. If Shopping impression supply narrows toward higher-intent queries, the top of an ecommerce funnel gets thinner even while the efficiency numbers look better, and the demand that used to arrive from browsing-stage searches has to be found somewhere else. That is a planning question rather than an account-settings question, and it is the sort of thing worth working through when reviewing an ecommerce marketing plan for the coming quarter.

FAQ on AI Overviews and Shopping ad CTR

Does a higher Shopping CTR mean my ads got better?

Not on its own. Ryan's stated position is that Shopping ads are not necessarily generating significantly more clicks and are simply being shown less often, which pushes the percentage up. Check absolute clicks and impressions over the same period before treating a CTR gain as an improvement in the ads themselves.

Is this happening in Thailand?

The analysis did not state country-level results, so there is no reported answer for Thailand. The only way to know whether a Thai account shows the same pattern is to compare its own impressions, clicks and CTR across the same twelve-month window.

Do I need to change anything in my account today?

No change is required by this reporting, because it describes a measurement pattern rather than a product update or a policy change. The reasonable response is a reporting change: show impressions and absolute clicks beside CTR, and read conversion volume rather than conversion rate.

How much of the change is caused by AI Overviews specifically?

The analysis does not say, and it does not isolate AI Overviews from other changes in the auction over the same twelve months. Ryan's framing links the two, but that link is his interpretation of a correlation rather than a measured effect.

Why does Optmyzr report 17% when Smarter Ecommerce reports about 20%?

They are separate datasets measuring separate samples, so a gap between them is expected. The reason both figures are worth citing together is that they point the same way, which strengthens the direction of the finding rather than the exact size of it.

If your Shopping or Performance Max reporting is showing a healthier CTR alongside flat revenue, that gap is worth an hour of somebody's attention before the next budget is set. The team at Relevant Audience is happy to look at the impression and click history with you and say plainly whether the account is following this pattern or a different one.