If you run Google Ads, you probably trust the search terms report. It tells you what people actually typed before they clicked your ad. It is the foundation of keyword optimization, negative keyword strategy, and budget allocation. You build campaigns on it. You report ROAS to your boss with it.

Google just admitted that, for a growing share of impressions, the data in that report is not real.

The quiet documentation update

Sometime in early May 2026, Google updated the help documentation for its search terms report. The change was not announced in a blog post, a press release, or a Marketing Live keynote. It was buried in a support page that most advertisers never read.

The new language is remarkably candid for a company that usually wraps product changes in euphemism. For searches that come through Google Lens, AI Mode, AI Overviews, and autocomplete, the search terms report now shows what Google calls "the best approximation of the user's intent."

Not the actual query. An approximation.

In some cases, Google goes further. For these AI-powered surfaces, keywords may not be used at all. Instead, ads are matched through what Google calls "AI-based ad group prioritization." That phrase appears in the documentation without elaboration, without a whitepaper, without a support article explaining what it means.

What exactly changed

Let's be precise about the scope. This is not a blanket change to all of Google Ads. Traditional text searches on google.com still report actual search terms, at least for now. The change applies specifically to Google's AI-powered search surfaces:

Google Lens. When someone searches by pointing their camera at something, there is no text query. Google interprets the image, generates an understanding of intent, and serves ads against that interpretation. The search terms report shows... something. Google won't say exactly what, but it is not a keyword a human typed. AI Mode. Google's conversational search interface, rolled out broadly in 2025, processes multi-turn queries, follow-ups, and natural language that doesn't map neatly to keyword structures. The "best approximation" language applies here too. AI Overviews. Even in standard search results, when Google's AI generates an overview at the top of the page, ad matching for those impressions falls under the new framework. Autocomplete. Predictive suggestions influence which ads appear before a user finishes typing. The actual eventual query may differ from what triggered the ad.

The common thread: none of these surfaces produce a clean, text-based query that maps neatly to a keyword in your account. Google's solution is to show you a proxy and label it as such, if you read the fine print.

Why this matters more than it sounds

A casual reading of this change might suggest it only affects edge cases. Camera searches, chat-style queries, autocomplete previews. Who cares about those, right?

Wrong. These surfaces are growing faster than traditional search. Google Lens now handles over 20 billion visual searches per month. AI Mode is being positioned as the future of Google Search itself. AI Overviews appear on roughly 40% of Google searches in markets where they are active, and that percentage is climbing.

The share of ad impressions affected by "best approximation" reporting is not 5% or 10%. It is on a trajectory to become the majority of Google's search ad inventory within two years. If you think this is a minor documentation tweak, you are misunderstanding the direction of travel.

The practical impact is immediate and severe. Every optimization decision you make based on the search terms report, negative keywords, bid adjustments, keyword pruning, budget reallocation, is now being made on data that is, by Google's own admission, an approximation for a significant and growing portion of your traffic.

Your negative keyword lists might be blocking approximations, not actual queries. Your bid strategies might be optimizing toward approximated intent, not real intent. Your ROAS calculations might be attributing conversions to keywords that no human ever typed.

This is not a small data quality issue. This is a structural shift in what advertisers are allowed to know about their own campaigns.

The historical pattern: death by a thousand cuts

This did not happen overnight. Google has been systematically reducing search term transparency for years, and the "best approximation" language is just the latest step in a well-documented pattern.

In September 2020, Google dropped a bomb: the search terms report would only show queries that met an unspecified "significance threshold." Overnight, the share of search term data visible to advertisers collapsed. Google claimed the change was about user privacy. Critics pointed out that it also made it harder for advertisers to identify waste, which conveniently meant they spent more on broad match and automated bidding.

Industry estimates suggested that after the 2020 change, advertisers could see the actual search terms for only about 30-40% of their spend. The rest was opaque, hidden behind Google's threshold. Google never disclosed what the threshold was, how it was calculated, or whether it changed over time.

That 2020 change itself was a escalation of a 2019 update that first began restricting search term data. Before 2019, the search terms report was nearly comprehensive. You could see almost every query that triggered your ads. By late 2020, most of that visibility was gone.

The "best approximation" update is qualitatively different from the 2020 threshold change, but the direction is the same. The threshold change made less data visible. The approximation change makes the data that is visible less reliable. Together, they create a world where advertisers have incomplete, potentially inaccurate information about what queries are driving their ad performance.

Google Merchant Center followed a similar trajectory. In 2024, it quietly dropped the word "Feeds" from its interface and documentation, reflecting a shift toward AI-curated product listings that bypass traditional feed management. The message was the same: stop trying to control the inputs. Trust the algorithm.

AdExchanger noted an industry-wide shift in Google's language, from "analytics" to "insights." Analytics implies raw data you can verify. Insights imply interpreted summaries you must trust. That linguistic shift is not accidental. It reflects a deliberate move away from advertiser empowerment toward advertiser dependency.

The post-keyword era

The deepest implication of this change is not about data quality or reporting accuracy. It is about the end of keyword-based search advertising as a conceptual model.

For twenty years, Google Ads has been built on a simple premise: advertisers choose keywords, users type queries, Google matches them. The search terms report was the feedback loop that kept this system honest. You could see what matched, adjust your keywords, and iteratively improve your targeting.

That model is now officially broken for Google's fastest-growing surfaces. When Google says "AI-based ad group prioritization" replaces keyword matching, it is saying that the fundamental unit of search advertising, the keyword, is no longer the primary matching mechanism for a growing share of impressions.

What replaces it? Google isn't saying, not in detail. "AI-based ad group prioritization" could mean contextual understanding, audience signals, behavioral modeling, or some combination. Advertisers have no way to know, no way to influence it directly, and no way to audit it.

This is the post-keyword era. Ads are matched to AI-interpreted intent, not to text strings. The advertiser's role shifts from targeting specific queries to creating ad groups that Google's AI finds relevant to vague, unobservable intent signals. You are no longer bidding on keywords. You are bidding on Google's interpretation of user intent, an interpretation you cannot see, verify, or directly shape.

If that sounds like Facebook advertising, it should. Facebook never let advertisers target specific search queries. You targeted audiences, interests, and behaviors, and trusted the algorithm to find the right people. Google is converging on the same model, but without acknowledging the shift or adjusting its pricing accordingly.

The irony is that Google built its entire advertising empire on the premise that search ads are better than social ads because you know what people want. "Intent-based advertising," they called it. That pitch is getting harder to make when you are showing advertisers approximations of intent instead of intent itself.

What advertisers should do now

First, acknowledge reality. The search terms report is no longer a reliable source of truth for a growing share of your Google Ads traffic. Treat it as directional data, not definitive data. This means changing how you build negative keyword lists, how you evaluate keyword performance, and how you report results to stakeholders.

Second, shift from keyword-level optimization to theme-level optimization. Instead of obsessing over individual query matches, focus on whether your ad groups, creatives, and landing pages align with broad intent themes. Google's AI is going to match based on holistic signals anyway. Your job is to make sure your ad content is thematically coherent, not to chase individual keywords.

Third, aggregate your ROAS measurement. Stop relying on keyword-level ROAS, which is now unreliable for AI-powered impressions. Move to campaign-level and ad-group-level measurement. Use incrementality testing and holdout experiments to understand what is actually working, rather than trusting Google's attributed data.

Fourth, diversify. If Google is going to make it harder to understand and optimize your search ad spend, reduce your dependency on it. This is not a radical suggestion. It is basic risk management. Explore channels where you have more transparency and control. The rise of AI-native advertising platforms, like ChatGPT Ads, offers alternatives where the matching logic is different and potentially more transparent. Our comparison of ChatGPT Ads versus Google Ads lays out the competitive landscape in detail.

Fifth, pay attention to Google's AI search optimization guidance. Google recently published its own recommendations for how brands should optimize for AI-powered search results. We reviewed that guide and found it both telling and incomplete. Understanding Google's stated position helps you read between the lines of what they are not saying about ad transparency.

Sixth, push your Google rep for answers. "AI-based ad group prioritization" is a phrase that appeared in documentation without explanation. Demand details. How does it work? What signals does it use? How should advertisers optimize for it? The silence is not acceptable for a platform that takes billions in ad spend.

Finally, watch the data carefully over the next quarter. Track the share of your impressions coming from Lens, AI Mode, and AI Overviews. You can do this through Google's audience and surface segmentation reports, though Google's own reporting on these surfaces is also evolving. Google's recent GA4 AI Assistant update introduced new channel groupings that may help you track this, though the implementation is still rough.

Is your ad strategy ready for the post-keyword era?

Find out where your brand stands in AI-powered search results. Check your AI visibility with Searchless and get a clear picture of how AI search surfaces are treating your business.

Sources

---

Ready to understand how AI search is reshaping your visibility? Explore Searchless pricing and plans.