Google this week is under scrutiny from paid search professionals after a quiet update to its official ad group prioritization documentation revealed that search terms shown in Google Ads reporting may no longer correspond to what users actually typed. The change, surfaced through a new section in the Google Ads Help Center page titled "About ad group and asset group prioritization within a Google Ads account," has prompted a wave of concern across the advertising community - particularly among practitioners managing lead generation campaigns where precise query control has historically been central to optimization.

The updated documentation introduces a category called advanced search experiences, covering searches conducted through Google Lens, AI Mode, AI Overviews, and auto-complete. According to the Google Ads Help Center, "in some cases, a user's search may involve more complex search journeys, for example, searches on Lens, AI Mode, AI Overviews, or auto-complete searches. In these instances, the search term shown in your reporting represents the best approximation of the user's intent." The critical qualifier is what follows: "because these searches are not considered technically identical to a keyword, keywords may not automatically be prioritised. Instead, AI-based ad group prioritization ensures that the most relevant ad groups or asset groups are selected to match the user's overall intent."

In plain terms, this means that for a growing subset of search activity - queries originating from AI-mediated surfaces rather than direct text entry - the term that appears in an advertiser's search terms report is Google's interpretation of what the user wanted, not the literal string the user submitted. The actual query that triggered the ad may not be shown at all.

A structural change to how reporting works

The four-tier prioritization system described in the same documentation provides context for how ad selection functions more broadly. Exact match keywords identical to a search term hold the highest priority at level one. Phrase and broad match keywords, including AI Max, that are identical to the search term sit at level two. When no identical match exists, level three activates: AI-based ad group prioritization, where relevance is determined by "the meaning of the search term, all the keywords in an ad group, and landing pages within the ad group." Ad Rank breaks ties at level four among equally prioritized candidates.

The advanced search experiences clause operates as a structural exception to this hierarchy. When a search originates from Lens, AI Mode, AI Overviews, or auto-complete, the query is not treated as technically identical to any keyword - regardless of how precise the keyword list is. This means exact match keywords, which have traditionally offered advertisers the highest degree of control, do not automatically receive priority for these query types. The AI-based ad group selection mechanism steps in instead, choosing among ad groups based on inferred overall intent rather than literal text matching.

Anthony Higman, Founder and CEO at ADSQUIRE, was among the first to draw attention to the documentation update on LinkedIn. According to Higman, "in advanced search experiences the search terms that you are seeing in your search query report might not be the search query at all. In fact they are the closest approximation that Google could make based on the complexities of these searches." He also flagged that keywords "may not actually be used at all, but instead Google will use AI based ad group prioritization in these cases to attempt to get the closest match."

Chris Chambers, Head of Paid Search at Understory, went further in a separate LinkedIn post that attracted significant engagement. "If a click comes through AI Max or an AI Overview, the search term you see in your account might not be what the user actually typed," Chambers wrote. "That's not a mistype, the search terms in your account might not be real now. Not a paraphrase of the search. Not a close match. A different query entirely, because Google is interpreting the intent and connecting it to what they think is the right result."

The update was first reported by Barry Schwartz at Search Engine Roundtable on May 13, 2026. A check of archived versions of the documentation on the Wayback Machine confirmed the section did not exist in earlier versions of the page, though the precise date of the addition has not been disclosed by Google.

Why negative keywords break down

The negative keyword feedback loop is where the practical implications become sharpest. Negative keywords function by blocking ads from appearing when specific terms are detected in a user's query. An advertiser who spots an irrelevant term in the search terms report adds it as a negative, and future traffic from that term is suppressed. This logic assumes a one-to-one relationship between what the user typed and what appears in reporting.

That assumption no longer holds for advanced search experiences. If the term visible in the report is an AI-generated approximation rather than the actual query, adding it as a negative excludes that approximation - but the underlying search intent, expressed through a different surface or a different query phrasing, continues to generate similar traffic under a different reported term the next time. Chambers described this as "playing whack-a-mole with shadows."

Michael Hulsmann, founder at SearchSavior, identified a structural asymmetry in how Google has evolved its targeting and exclusion tools. According to Hulsmann, "Google's matching keeps getting looser (broad expansion, AI Max keywordless, intent inference) while negatives still use strict rules from a decade ago. Block a junk query and the same intent reappears under a different reported term next week. You're blocking symptoms while the cause keeps producing new ones."

This asymmetry is not incidental. Google has expanded the reach side of its matching systems substantially over the past several years - through broad match, through AI Max's keywordless targeting, and now through AI-mediated surface interpretation - while the exclusion toolset has evolved more slowly. Google raised the negative keyword limit for Performance Max campaigns from 100 to 10,000 in March 2025, and campaign-level negative keyword lists for Performance Max arrived in Google Ads Editor 2.11 in November 2025 - but both of those developments address list capacity, not the fundamental mechanics of what gets excluded.

The search terms report has been eroding for years

This is not the first time the utility of the Google Ads search terms report has narrowed. Google introduced limits on search term data in 2020, restricting the report to terms "that a significant number of users searched for, even if a term received a click." That change removed visibility into low-volume queries, which disproportionately affected advertisers in niche verticals and those optimizing long-tail keyword strategies.

Research cited in LinkedIn discussions estimated that between 27% and 73% of search term data was hidden behind privacy aggregation thresholds as of 2025. The advanced search experiences update adds a second layer of opacity: among the terms that do appear, some now represent Google's inference about intent rather than the literal query string. Alex Lin, a Google Ads Manager commenting on the Chambers post, noted that long-tail queries were already being approximated to what he described as a "synthetic keyword" - often reduced to a maximum of three words - and that additional terms were disappearing into the "Other" category. "The problem now seems to be that a lot of terms still go into 'Other' where you can't see search terms that aren't common... and now the search terms that it does show us are not always going to be real," Lin wrote.

The documentation change also connects to a longer-running pattern of AI-mediated changes in how Google handles attribution and reporting. AI Max attribution discrepancies surfaced in December 2025, when independent analysis by Brad Geddes, co-founder of Adalysis, documented instances where AI Max claimed credit for conversions that exact and phrase match keywords had already delivered. Independent tests across more than 250 campaigns published in November 2025 found AI Max delivering 35% lower ROAS than traditional match types, with less than half of search terms showing keyword-level matching.

AI surfaces are reshaping what a search term means

The root cause of the documentation change is structural. AI Mode, AI Overviews, and Google Lens each process user input in ways that do not map cleanly to the string-based model that Google Ads was built on. A user asking a multi-part question through AI Mode submits a query that may trigger a synthesized response drawing from multiple sources. The ad that appears alongside that response is matched not to the full query text but to the system's interpretation of the commercial intent embedded in the conversation. Queries in AI Mode run two to three times longer than traditional searches, providing richer intent signals to Google's systems but also making keyword-level attribution increasingly difficult to verify.

Google unveiled shopping ads in AI Mode in February 2026, and sponsored stores and quick web results have since been spotted inside AI Mode as commercial inventory within the conversational interface expands. Each new surface introduces another layer where the relationship between a user's input and the ad that appears is mediated by Google's AI rather than by direct keyword matching.

Lens introduces a different dimension. A user searching by image submits no text at all. Google's systems interpret what is in the photograph and construct a semantic understanding of what the user likely wants. The search term that appears in an advertiser's report for a Lens-initiated query is therefore necessarily an inference, not a transcription. Auto-complete presents a subtler version of the same problem: the user selects from a dropdown rather than completing their intended query, and the completed string may not reflect what they would have typed independently.

B2B and lead generation face the sharpest exposure

Chambers argued that the impact falls unevenly across account types. E-commerce campaigns operating at scale generate sufficient conversion volume for Smart Bidding to calibrate effectively, and the incremental loss of query-level clarity may be tolerable when the bidding system has thousands of signals to work from. Lead generation accounts, particularly in B2B, operate at lower conversion volumes and rely more heavily on manual optimization signals - including the search terms report - to identify and exclude irrelevant traffic.

"In ecom nowadays you don't need negatives as much and smart bidding for mid-large brands works great because it has thousands of conversions per month to learn from. That's not true for lead gen usually," Chambers wrote. The structural issue is that lead gen campaigns often target narrow intent signals - procurement-related queries, specification searches, vendor comparison terms - where a single word distinguishing a relevant from an irrelevant query can represent the difference between a qualified lead and wasted budget. If that distinguishing word is lost in an intent approximation, exclusion strategies built around it may fail silently.

Stan Oppenheimer, a Search PPC Specialist and Consultant, characterized the issue in terms of reporting transparency: "this proxy keyword issue in PMAX/AI Max SQR reports is frustrating. Long-tail queries getting buried under 'strongest match' keywords kills real optimization. Google needs to clearly label these proxy terms in reports so we know what's actual user search vs. algorithm guesswork."

The direction of travel

Several practitioners commenting on the LinkedIn discussions framed the documentation update as a clarification of a shift that had already been underway for some time rather than a genuinely new policy. The synthetic keyword behavior that Alex Lin described - where long-tail queries collapse into abbreviated approximations - predates the help center language. The formal acknowledgment that reported terms represent "best approximations" of intent codifies a reality that advertisers in AI Max and Performance Max campaigns had already been encountering in practice.

Google ended Dynamic Search Ads as a standalone format in April 2026, announcing that legacy DSA campaigns will automatically upgrade to AI Max for Search starting in September 2026. That migration will bring more accounts into the AI-mediated matching framework covered by the advanced search experiences documentation, expanding the population of advertisers affected by approximated search term reporting.

Text guidelines went global for AI Max and Performance Max in February 2026, giving advertisers up to 25 term exclusions and 40 messaging restrictions to constrain AI-generated ad copy. That feature governs what ads say, not what queries trigger them - a distinction that underscores the separation between creative controls, which have expanded, and targeting transparency, which has narrowed.

Nico Fonce, a digital growth specialist commenting in the Chambers thread, drew the trajectory plainly: "For as long as I can remember (+15 years), Google is working towards an 'auto platform': set your budget, target, etc - and Google does the rest. The advantage for Google is clear: anyone can start a campaign = more cash. For the pro's = less control."

Chambers offered a view of where the platform may be heading. "I see the near future as Google just wanting us to give them the business app name/url (if urls still exist), then fund the budget and they will do everything in a black box. Gemini search ads. Fully turnkey."

Timeline

Summary

Who: Google, the operator of Google Ads, updated its official ad group prioritization documentation. The update was surfaced and discussed publicly by paid search practitioners including Anthony Higman (Founder and CEO, ADSQUIRE), Chris Chambers (Head of Paid Search, Understory), Michael Hulsmann (Founder, SearchSavior), and others. Barry Schwartz at Search Engine Roundtable reported on the change on May 13, 2026.

What: A new section titled "Advanced search experiences" was added to Google's Help Center page on ad group and asset group prioritization. The section states that for searches conducted via Lens, AI Mode, AI Overviews, or auto-complete, the search term shown in reporting "represents the best approximation of the user's intent" rather than the literal query entered. It also states that keywords may not automatically be prioritized for these searches, with AI-based ad group prioritization selecting ad groups instead.

When: The documentation update was confirmed by Barry Schwartz on May 13, 2026. The precise date of the addition to Google's Help Center was not disclosed by the company. Discussion among advertising professionals on LinkedIn intensified in the days following the report.

Where: The change appears in Google's official Ads Help Center documentation and affects all advertisers using Google Search campaigns, AI Max, and Performance Max globally, wherever searches originate from Lens, AI Mode, AI Overviews, or auto-complete surfaces.

Why: The change reflects a structural shift in how Google processes queries that originate from AI-mediated surfaces. These searches do not produce a standard text string that maps cleanly to a keyword. Google's systems instead infer user intent and select the most relevant ad group accordingly. For advertisers, the consequence is that the search terms report - already limited by privacy-based aggregation since 2020 - now may display AI-generated intent approximations rather than literal user queries for a growing share of search activity, undermining negative keyword strategies that depend on knowing what users actually searched for.

Share this article
The link has been copied!