More than a quarter of exact match keyword impressions inside AI Max-enabled ecommerce Search campaigns are no longer exact, according to an analysis of 383 million impressions published on LinkedIn in late July 2026 by Mike Ryan, Head of Ecommerce Insights at Smarter Ecommerce (smec).

The number carries an uncomfortable implication for anyone who still treats square brackets as a control mechanism. Ryan's dataset covers ecommerce Search campaigns where AI Max has been activated, and it counts how many impressions attributed to exact match keywords were in fact generated through the feature's search term matchingtechnology rather than through literal keyword matching. His summary of what that technology is leaves little room for interpretation: it is, in his words, "basically broad match."

Timing gave the post its charge. Ryan published days after Alphabet reported second-quarter 2026 results on July 22, 2026, when Google Search advertising revenue rose 17 percent to 63.3 billion dollars while the Network segment slipped 1 percent. Chief Business Officer Philipp Schindler attributed part of that growth to automated campaign formats. Ryan set the two things beside each other and asked whether the newly matched queries contain anything worth buying.

What the sample measures

The chart accompanying the post is titled "1 in 4 Exact Match impressions are expanded by AI Max" and carries the subtitle "based on ecommerce Search campaigns with AI Max activated | n=383 million impr."

Two series run across the plot. The black line tracks the share of impressions on exact match keywords that resolved as genuine exact matches. The blue line tracks the share broadened by AI Max. The horizontal axis runs from January 2025 to July 2026.

The shape of the two curves is the finding. Exact match holds at or near 100 percent through the first half of 2025, dipping only marginally in July and August. It then steps down in stages: to roughly 90 percent around November 2025, to the high 80s through the first quarter of 2026, and then into a steeper slide through the second quarter. By July 2026 the black line sits at approximately 71 percent. The blue AI Max line mirrors it, moving from zero through low single digits in mid-2025, to around 10 percent in late 2025, briefly retreating in January 2026, and then climbing sharply from April onward to close near 29 percent.

The steepest movement is recent. Roughly two thirds of the total expansion visible in the eighteen-month series occurred in the final four months of it.

Scope, and what the sample does not cover

Ryan confirmed the geographic boundary in the comment thread. Responding to Robert Grassmann, a senior account strategist who asked whether the data was EMEA-only, Ryan wrote: "yes, EMEA." He added that he holds North American data but would not draw a regional comparison from it, while offering a hypothesis about why performance might differ. English-language markets, he suggested, benefit because Google's language models were trained on a larger English corpus, which affects both query comprehension and ad copy generation.

The sample is also vertical-specific. These are ecommerce campaigns, a segment where broad match has attracted less hostility over recent years than in lead generation. That caveat cuts both ways: it may understate resistance elsewhere, and it may overstate expansion in accounts with tighter keyword hygiene.

The earnings call that framed the argument

Ryan quoted Schindler's remarks from the Alphabet call directly, marking the closing clause for emphasis. The legible portion of that quotation reads: "And then we have our AI-powered campaigns like AI Max that help advertisers actually adapt and find the opportunities beyond keywords. Again, that's an ability for us to go deeper and target better."

PPC Land's coverage of the same call recorded Schindler's claim that adopters of AI-powered campaigns see an average of 15 percent more conversions or value on Search at a similar return on ad spend, with half a million advertisers now running AI Max. Ryan's reading of the revenue picture was blunt: "Google is printing money because, all these years later, they're still finding ad inventory in classic search."

His question followed immediately. "But these new searches... is there any meat on the bone?"

Diminishing returns versus the 15 percent claim

Ryan has criticised AI Max performance claims before, and he restated the objection. The problem, as he framed it, is not that the number is impossible in isolation but that it is being applied universally. "This might be plausible in a poorly optimized account, or one that has been underspending, but for anyone serious about search advertising... the next 15% conversion volume will simply not be as efficient as the previous. Law of diminishing returns."

That scepticism has precedent in measurement published outside Google. Smarter Ecommerce analysis of more than 250 retail campaigns in November 2025 found AI Max delivering conversions at approximately 35 percent lower return on ad spend than traditional match types inside the same campaigns, with higher cost per conversion and lower average order values. Earlier, in August 2025, testing across roughly 30,000 AI Max search terms reported that 99 percent of impressions produced zero conversions, alongside evidence of aggressive expansion onto Search Partner Network placements.

Google's own headline figure has moved. The May 2025 open beta launched with a 14 percent uplift claim. The April 2026 general availability announcement revised that to an average of 7 percent for the full feature suite measured against search term matching alone, a footnoted internal figure that excluded retail advertisers. In May 2026, Google told advertisers that campaigns previously dominated by exact and phrase match were seeing 27 percent more conversions at a similar CPA or ROAS after adopting AI Max. Three different numbers, three different measurement frames, one product.

Lucas Mota, a senior paid media specialist, put the commercial framing plainly in the thread: "15% with the same ROAS is a crazy commercial claim to make." He described AI Max as partly a rebrand. Ryan agreed, calling it "very much a rebrand of basically the campaign-level broad match recommendation," while noting that the reputation of broad match in ecommerce has improved over recent years.

Where Ryan changed position

The post is not a flat attack. Ryan wrote that he has come to accept Schindler's characterisation of the moment as one of adaptation, and grounded that acceptance in query behaviour rather than platform messaging.

"We've been seeing longer, more complex search queries for years. But that trend has dramatically accelerated in the era of AI Search."

The supporting evidence sits outside the post. Google AI Mode passed one billion monthly active users by May 2026, with average query length running roughly three times that of traditional search and more than one in six United States searches arriving as voice, image or video rather than text. Keyword lists built for two-word product lookups were not designed for multi-clause descriptive prompts.

Ryan's conclusion accepts a performance trade-off rather than denying one. "AI Max is a technology built to upgrade beloved campaign types like Search and Shopping, and take those advertisers along for the ride. No, the AI Max segment of your Search campaigns will probably not be the best performing; however, it is exactly the part that helps you meet evolving demand."

The practitioner objections

The thread drew 35 comments and more than 65 reactions. The objections clustered around three themes: signal quality, visibility, and containment.

Signal quality came first. Adrian Canovas, Paid Search EMEA Lead at HP, distinguished between conversion types: the system behaves acceptably when optimising toward a macroconversion such as a real lead or sale, and becomes unmanageable when the signal is a site engagement event. Birgit De Vries pushed the point further, asking at what cost the conversions arrive and listing the mechanisms a buyer cannot observe from the interface, including cheap impressions on low-quality placements and expansion onto competitor search terms.

Visibility was De Vries's second objection, and it is the one with the clearest technical basis. She identified two blind spots: match type turning exact into something broader without disclosure of which queries or prompts the system opened up to, and landing-page induced queries where data exists but interpretation does not. Her summary was that the promised future looks reasonable while the present reporting does not support assessment of either outcomes or impact.

That measurement gap has been documented before. In December 2025, Google clarified that AI Max search term matching relies on inferred intent rather than raw query text, following analysis showing that AI Max traffic is assigned to exact and phrase keywords in reporting when no corresponding broad match keyword exists, making per-match-type evaluation impossible.

Containment produced the most operational responses. Martin Große, Head of SEA and Programmatic Display at Suchmeisterei GmbH, described keeping AI Max away from exact-heavy campaigns entirely and building a separate campaign for the new traffic instead, on the reasoning that budget shifts toward new search terms are difficult to control once enabled. Ryan called that a fair reframing, adding that Google would argue the opportunity is largest for exact-match-heavy advertisers, "but so of course is the risk."

Shaun Elley, founder of BlueOxDigital, reported terrible results and terrible queries across several accounts, and questioned whether prior search history justifies serving a car advertisement against a query for bus routes. Ryan responded with a case from his own work: a health and beauty retailer receiving impressions on the single words "hair" and "skin."

Pedro Talaia said clients wanting bid control on specific terms found that control ignored, with spend flowing to generic queries. Ryan pointed to ad group level settings as a partial answer, noting that exact match can still be protected where necessary.

A Google employee describes a different mechanism

One exchange is worth isolating because it exposes how little consensus exists about the underlying auction logic. Fabian H. Panah, who works in digital growth at Google and prefaced his comment by stating he was not defending the company, described his understanding that AI Max can bid on exact and phrase match keywords only when the cost per click is cheaper than what the keyword itself would have paid to win the auction.

Ryan did not accept the description. He suggested it resembled a cross-campaign dynamic, the sort that governs whether a Search campaign or a Performance Max campaign wins an identical query, and offered his own reading: exact match continues operating unchanged, with AI Max broadening only on top of it, serving queries that exact match would otherwise have skipped. Both parties flagged uncertainty about their own account.

Fintan O'Toole, a senior paid search specialist, noted that expansion beyond exact match has been visible in search term reports for some time without generating comparable controversy. His explanation was that the underlying mechanism is now better concealed, and that Google's tactics become legible when read across a five-year product horizon.

Why exact match stopped being exact

The erosion Ryan measured did not begin with AI Max. Google altered close variant behaviour in 2018 to match searcher intent rather than literal keyword text, a change that ended single keyword ad group strategies. By November 2025, analysis showed exact match keywords triggering ads for semantically unrelated terms, including a keyword built on the word hypoallergenic serving against allergy queries that never contained it.

What AI Max adds is scale and a different input set. Search term matching combines broad match expansion with keywordless targeting drawn from landing page content, existing ad copy, ad group keywords and real-time intent signals. The result is a query pool that no keyword list defines and no search term report fully reconstructs.

The September 1 deadline gives the number weight

Ryan's dataset covers advertisers who switched AI Max on. That distinction has a shelf life.

The April 15, 2026 general availability announcement set September 2026 as the point at which legacy settings would be upgraded automatically. Campaigns running the campaign-level broad match setting were scheduled to receive search term matching; campaigns using automatically created assets were scheduled to receive both search term matching and text customization. Dynamic Search Ads received a reprieve in June 2026, when the automigration moved to February 2027 after advertiser feedback about fourth-quarter planning risk. The other two settings did not.

For accounts that inherit search term matching without electing it, the smec curve stops being an observation about early adopters and becomes a baseline expectation.

Why this matters for marketers

The practical consequence is a reporting problem before it is a performance problem. When 29 percent of the impressions credited to an exact match keyword were not exact, historical comparisons across that keyword measure two different products stacked in one row. Year-over-year cost per click movements, quality score narratives and incrementality assumptions built on the pre-2025 behaviour of exact match no longer describe the same mechanism.

The second consequence concerns budget attribution rather than budget size. If Große's containment logic holds, the volume arriving through search term matching competes for the same daily budget as the queries an advertiser explicitly targeted, without appearing as a separate line item. Google's own documentation acknowledges the interaction, warning that AI Max will not be effective in budget-limited campaigns.

The third is structural. Google Network revenue, the segment that pays open-web publishers, has declined across consecutive quarters while Search revenue compounds, and independent measurement puts the click-through reduction for top-ranked pages where AI Overviews appear at 58 percent. Search inventory is expanding on the query side at the same moment it contracts on the referral side. Ryan's chart quantifies one mechanism through which that expansion reaches an advertiser's account: not as a new campaign type requiring a decision, but as a rising share of impressions inside a line item that already existed.

The question he closed with was addressed to practitioners rather than to Google. "How do you feel about this? Honestly."

Timeline

Summary

Who: Mike Ryan, Head of Ecommerce Insights at Smarter Ecommerce (smec), published the analysis. Named participants in the discussion include Adrian Canovas of HP, Birgit De Vries, Shaun Elley of BlueOxDigital, Martin Große of Suchmeisterei GmbH, Robert Jakobson, Robert Grassmann, Lucas Mota, Fintan O'Toole, Pedro Talaia, and Fabian H. Panah of Google. Philipp Schindler, Chief Business Officer at Google, is quoted from Alphabet's earnings call.

What: An analysis of 383 million impressions in ecommerce Search campaigns with AI Max activated found that more than 25 percent of exact match keyword impressions were broadened by AI Max search term matching rather than matching literally. The accompanying chart shows the expanded share rising from zero in early 2025 to approximately 29 percent by July 2026, with exact match falling to roughly 71 percent.

When: The analysis was published on LinkedIn in late July 2026, days after Alphabet reported second-quarter results on July 22, 2026. The measurement window runs from January 2025 to July 2026.

Where: The dataset covers EMEA, confirmed by Ryan in the comment thread. Campaigns are ecommerce Search campaigns running in Google Ads.

Why: The finding matters because exact match has functioned as the primary control mechanism in keyword-based search advertising, and the measurement shows that control weakening at an accelerating rate inside accounts that adopted AI Max voluntarily. Campaigns using automatically created assets or the campaign-level broad match setting are scheduled for automatic upgrade on September 1, 2026, which extends the same dynamic to advertisers who did not elect it. Google reports a 15 percent average conversion uplift for AI-powered campaigns; independent testing has repeatedly measured lower return on ad spend for AI Max traffic than for traditional match types, and the reporting surface does not currently allow advertisers to separate the two inside a single keyword row.