01 The precedent
Google told regulators its own answers just got worse
A redesign forced by antitrust law admits, in writing, that better compliance can mean a worse search page.
On Tuesday, Google rolled out a rebuilt set of search results across the European Union, and it did not pretend the change was an improvement. Google told Reuters the redesign produces “the largest reduction in quality of service” in Search’s 29-year history. The company had little choice: it is complying with a Digital Markets Act ruling, backed by a €460 million ($534 million) fine issued in July for favoring its own shopping and travel services over competitors’.
The new layout puts one specialized comparison-shopping service at the top of results for categories like hotels, flights and restaurants, with two more shown in smaller detail below, and a carousel that no longer displays real-time prices. A traveler comparing flights now sees Google’s own summary replaced by a rival price-comparison service, which is the entire point of the antitrust finding.
Google’s own account of the cost is unusually candid for a company being punished. It says an earlier round of DMA changes already cost European businesses a 30% drop in free, direct-booking traffic as travelers routed through third-party comparison sites instead. This week’s changes go further.
None of this changes what a US search shows today. But it is the clearest public admission yet that a “better” answer, by regulatory definition, can look worse to the company forced to build it, and that whoever writes the definition decides which businesses profit from the redesign. The EU is the one place a search engine has had to publish that tradeoff instead of quietly making it.
What to do about it
Nothing changes in your account this week; this is a precedent to watch, not an action to take today. The mechanism matters because the same logic already applies domestically: a legal directory or aggregator site can outrank your own page for reasons that have nothing to do with the quality of your content.
If a client asks why a directory listing sits above your firm’s own website, this is a plain-language answer: platforms sometimes elevate a comparison layer over a specialist's page by design, not by accident, and it can happen to Google itself.
Source
Search Engine Land, “Google says DMA changes in EU resulted in worse degradation of search quality ever,” September 8, 2026 · searchengineland.com · wire reporting: Quartz, via Reuters
02 The evidence
The businesses AI actually recommends look nothing like the rest
A new comparison of thousands of local brands finds the gap between being visible and being chosen is wider than a visibility report usually admits.
A local-marketing platform, SOCi, tracks visibility data across roughly 350,000 U.S. business locations in 42 industries. Search Engine Land published a fresh cut of that data this week answering a question most visibility reports dodge: out of everyone who shows up in a search, how many actually get named by an AI system? Not many, and it depends heavily on what kind of business you already are.
Averaged across every industry SOCi tracks, a business appears in Google’s local 3-pack 35.9% of the time. Ask an AI assistant the same kind of question, and the same business is recommended by ChatGPT only 1.2% of the time, with Gemini and Perplexity landing in the low single-to-double digits. Earning an AI recommendation, in other words, is three to thirty times harder than earning a spot Google already shows you in.
The sharper comparison is between brands that are actually growing and brands that are shrinking. Expanding restaurant chains score an average of 62 out of 100 on SOCi’s visibility index, against 47 for contracting chains, and the AI gap widens further from there: ChatGPT recommends the growing chains in about 20% of test queries and the shrinking chains in about 3%, a six-to-sevenfold difference that repeats on Gemini and Perplexity. Locations that AI systems do recommend also carry noticeably higher star ratings, which suggests a review average below roughly 4 stars functions less like a ranking factor and more like a cutoff.
SOCi sells the software this finding happens to justify, worth naming rather than ignoring. But the comparison group is what survives that conflict of interest: this isn’t “AI likes visible brands,” it’s growing and shrinking versions of the same category of business measured against each other, not against a vendor’s own baseline.
What to do about it
Read your own reviews the way an AI system apparently does: not “do I have enough,” but “am I clearly and consistently above 4 stars.” Below that line, a whole category of recommendation may exclude a firm regardless of how well the website is built.
Don’t treat a strong local-pack ranking as proof of AI visibility. The two correlate weakly at best, and a firm that dominates the map can still be invisible in every AI answer a prospective client actually reads.
Source
Search Engine Land, “Why growing restaurant chains win at local search and AI visibility,” September 2026, data via SOCi · searchengineland.com
03 The blind spot
Your Google Business Profile numbers may not exist right now
Nine days into September, the dashboard millions of local businesses check every week is still blank, and nobody has said why.
If you logged into your Google Business Profile this month expecting to see how many people called, requested directions, or viewed the listing, you may have seen nothing at all. Search Engine Roundtable reported this week that Google Business Profile’s Insights analytics have shown no data for the whole of September, a gap that is now more than a week old.
Insights normally lags by a few days, never longer. Google has not published an explanation or a fix date, and the reports so far are still scattered Reddit complaints rather than a wave, which suggests either that most businesses have not checked closely yet, or that the outage is not hitting every account.
The likely outcome, based on how Google has handled comparable Insights gaps before, is that the underlying data was recorded correctly and will eventually be backfilled, rather than genuinely lost. But “eventually” is doing a lot of work in that sentence, and a partner who checks this dashboard once a month, sees zero, and concludes the phone stopped ringing would be reacting to a reporting failure as though it were a business failure.
Do this today
Don’t adjust marketing spend based on a Google Business Profile Insights chart reading zero for September. Cross-check against the actual call log, intake form submissions, or CRM before concluding anything changed.
If you track month-over-month visibility for a client or for your own firm, note the outage explicitly in this month’s report rather than plotting a collapse that never happened. A missing month of data is not the same thing as a bad month.
Source
Search Engine Roundtable, “Google Business Profiles Insights Missing Data For September,” September 8, 2026 · seroundtable.com
04 The one with a clock on it
Google is redefining what counts as an ad being seen
Starting in February 2027, an ad that began loading and then vanished will no longer count as an impression, and every number compared against AdSense will shift under it.
On September 1, Google emailed AdSense publishers to announce a change to how it counts display-ad impressions, effective February 17, 2027. Today, an impression is logged the moment an ad starts to download to a visitor’s device, whether or not it ever actually appears on screen. Under the new method, called Begin-to-Render, an impression counts only once the ad has actually started rendering for the visitor.
The practical effect: impressions that download but never render, because a visitor left the page first, will simply stop being counted. Google frames this as aligning with industry-standard measurement and closing the gap between what publishers report and what advertisers’ own tools show. No action is required from publishers; the switch happens automatically.
The reason this belongs in a briefing about AI visibility rather than an ad-ops newsletter is the pattern, not the mechanism. A platform unilaterally redefined what “seen” means, for a metric millions of dashboards report, on five months’ notice, with no input from the people whose numbers will move. If a firm’s advertising reports include AdSense-served inventory anywhere in the funnel, expect a step change in impression counts next February that has nothing to do with actual reach.
What to do about it
Mark February 17, 2027 on next year’s reporting calendar now. When impression totals move, it helps to already know the definition changed rather than discovering it mid-quarter.
If a vendor’s dashboard reports an impression count, ask which counting method it uses and when it plans to move to Begin-to-Render. A vendor who has not thought about this yet has not thought carefully about the number being sold.
Source
Search Engine Land, “Google AdSense to change how it counts impressions: Begin-to-Render,” September 1, 2026 · searchengineland.com · announcement coverage: Search Engine Roundtable
05 The money
Shopping ad clicks are up while fewer people ever see them
A year-long analysis suggests AI Overviews are quietly trimming the audience for Shopping ads down to the buyers already close to purchasing.
Search Engine Land analyzed a year of Google Shopping ad performance across advertiser accounts and found a pattern worth sitting with. Median Shopping ad impressions per account fell from roughly 1.85 million to about 1.4 million between mid-2025 and mid-2026, a real decline in reach. Over the same period, median click-through rate rose from 1.20% to nearly 1.55%, up about 20% year over year.
Fewer people seeing the ads, and a higher share of the ones who do see them clicking, is not the pattern of a simple slowdown in ad demand. The reading offered: AI Overviews are increasingly answering the lower-intent shopping questions directly, the ones where a browsing shopper might once have seen a Shopping ad and ignored it, which narrows the served audience down to searches specific enough that Google still shows a traditional results page. Fewer, more qualified impressions would explain both numbers moving in opposite directions at once.
The honest caveat matters here: this is a twelve-month correlation, not a controlled experiment, and it does not rule out other explanations, like changes to Google’s own ad auction. Google has not confirmed the causal story. Treat the finding as a plausible read of a real pattern, not a proven mechanism.
If the reading holds, it says something bigger than Shopping ads. AI answers may not be destroying ad-supported discovery so much as filtering it, quietly removing casual browsers from the funnel before an ad ever reaches them, leaving advertisers with smaller, sharper audiences and better-looking numbers that measure something different than they used to.
What to do about it
If a campaign’s click-through rate has climbed while total lead volume has not, check whether served impressions dropped first. A rising percentage on a shrinking denominator can look like an improvement while actual reach contracts.
Ask whoever manages paid search to separate “fewer, better-qualified impressions” from “the campaign lost reach.” Right now, both situations produce an identical-looking click-through chart.
Source
Search Engine Land, “AI Overviews may be affecting Shopping ad CTR and impressions,” by Anu Adegbola, September 7, 2026 · searchengineland.com · coverage: Search Engine Roundtable
