Google Ads Is Testing AI Mode Ads for Exact and Phrase Match Keywords

If you have spent years building a Google Ads account on exact and phrase match specifically because you wanted control over exactly which searches your ads could show against, Google just quietly moved the boundary of that control without asking. On September 4, 2026, Google's Ads Liaison Ginny Marvin confirmed that exact and phrase match keywords are now eligible to serve text ads inside AI Mode, Google's conversational AI search surface. That has never been true before. Until this week, the only doors into AI Mode ad placements were the automated, broad-leaning products: AI Max for Search, Performance Max, broad match paired with Smart Bidding, Dynamic Search Ads, and Shopping campaigns. Exact and phrase match, the two match types advertisers reach for specifically when they want tight, predictable control over search intent, had never been in that list.

The timing matters here as much as the substance. This is not a hypothetical roadmap item or a Google support-page update that quietly reworded some eligibility language months from now. Anthony Higman, founder and CEO of the agency ADSQUIRE, spotted the behavior in his own live campaigns on September 3 and posted about it before Google had said a word publicly. Google confirmed it the very next day, as first reported by ppc.land. Barry Schwartz at Search Engine Roundtable had the write-up live by 2:50pm that same afternoon. That sequence, an advertiser noticing something odd in their own account, then a named Google spokesperson confirming it within 24 hours, is about as close to real-time proof as this industry gets, and it is exactly the kind of story worth understanding in detail rather than skimming a headline about.

This post walks through what AI Mode actually is for anyone who has not been tracking it closely, the exact language Google used and why the phrase "explicit and direct user intent" is doing a lot of unexplained work, a clear before-and-after picture of who could reach AI Mode ad inventory and who can now, the very real reporting gap this creates, and a practical list of what to actually do about it this week, not in six months once Google writes an official policy page.

TL;DR: What Changed and Why It Matters

Google is running what it calls a small experiment that lets exact and phrase match keywords from standard Search campaigns serve text ads inside AI Mode, something that was structurally impossible before this test. Previously, the only ways to reach AI Mode ad inventory were AI Max for Search, Performance Max with text customization enabled, broad match paired with Smart Bidding, Dynamic Search Ads, and Shopping campaigns, in other words, only Google's more automated and broad-leaning products. Marvin's confirmation adds one real constraint: eligibility is "limited to cases where there is explicit and direct user intent," though Google has not defined what that means in practice. There is no announced end date, no confirmed country list, and critically, no way in current Google Ads reporting to separate how much of a keyword's performance is coming from AI Mode versus standard Search results.

If your account leans on exact and phrase match as a deliberate strategy, and you are already generating enough ad creative to test across expanding placements using tools like Miraflow's Cinematic AI Video Generator, this is worth reading in full rather than skimming past as one more minor Google Ads headline, because it removes a restriction that a lot of advertiser strategy has been built around for years.

What AI Mode Actually Is, Briefly

AI Mode is Google's conversational AI search experience, distinct from the AI Overviews that show above traditional blue-link results. Instead of a static summary sitting on top of a normal search results page, AI Mode is closer to a chat interface built directly into Search: a user types or speaks a question, Google's AI generates a synthesized, conversational response, and the user can follow up with clarifying questions the way they would in a back-and-forth conversation rather than firing off a new, unrelated search each time.

The scale here is no longer a rounding error. AI Mode reached 1 billion monthly users by Google I/O 2026, and per Similarweb data from August 2026, nearly 30% of ad-eligible queries inside AI Mode now show ads. Queries inside AI Mode also run roughly three times longer on average than a traditional search query, which tracks with what a conversational interface naturally produces: instead of typing "running shoes," a user in AI Mode is more likely to type or say something closer to "what running shoes should I get for marathon training on a budget under 150 dollars," a query that carries dramatically more explicit intent packed into a single line than the keyword-style queries most Search campaigns were originally built to catch.

That combination, a billion monthly users, rising ad density, and unusually long, intent-rich queries, is exactly why Google opening a new eligibility path into AI Mode is not a footnote. It is a meaningful shift in where a growing share of search-based ad impressions can now come from, for an account type that previously had zero access to that surface at all.

What Exactly Changed: The Marvin Quote and the Ambiguity Inside It

Ginny Marvin's exact statement, confirmed on September 4, 2026, was: "Exact and phrase match keywords are eligible to serve text ads in AI Mode." She followed that immediately with the qualifier that this is "limited to cases where there is explicit and direct user intent," a phrase Google has not defined further in any public documentation as of this writing.

That second sentence deserves more scrutiny than it has gotten in most of the coverage running past it quickly. "Explicit and direct user intent" sounds precise, but it isn't a term Google Ads has a defined, measurable meaning for anywhere else in its policy documentation, and it isn't a checkbox in the interface or a bid modifier either. It reads more like a phrase describing a judgment Google's own AI Mode ranking system is apparently making internally, about whether a given conversational query is specific enough to warrant showing an exact or phrase match keyword's ad, and advertisers currently have no visibility into how that judgment gets made, no setting to influence it, and no reporting signal that tells them when it has or has not been applied to their account.

This is worth sitting with for a second, because it inverts the usual relationship between match type and control. Exact and phrase match exist specifically so an advertiser can say, in effect, "only show my ad for something close to this exact phrase," and have that promise enforced mechanically and predictably by the matching system. What Marvin's statement describes instead is a second, less transparent layer of judgment sitting on top of that promise, specifically inside AI Mode, deciding whether a query counts as having enough "explicit and direct user intent" to trigger the ad at all. The match type restriction has not gone away. A new, undocumented filter has been added on top of it, for one specific surface, and advertisers cannot currently see it operate.

Marvin also noted, in the same exchange reported by ppc.land, that AI Max and Performance Max retain exclusive access to newer AI Mode ad formats like Highlighted Answers, meaning this test opens standard text ad eligibility to exact and phrase match specifically, not full parity with what the more automated products can already do inside AI Mode. That distinction matters for expectations: this is a meaningful crack in a previously solid wall, not a full opening of the same door AI Max already walks through.

Before and After: Who Could Reach AI Mode Ad Inventory

The cleanest way to understand the scale of this change is to look directly at the eligibility list before and after. Before this test, reaching AI Mode ad placements required using one of a specific, short list of automated or broad-leaning campaign setups. Exact and phrase match keywords were never on that list, no matter how well an advertiser's account was structured or how much budget they were willing to spend.

Look at what that previous list actually had in common. AI Max for Search leans heavily on Google's automated broadening of a campaign's targeting. Performance Max, even with text customization enabled, hands most targeting decisions to Google's automated bidding and asset-matching systems. Broad match paired with Smart Bidding is, by definition, the least advertiser-controlled standard match type paired with the most automated bidding strategy available. Dynamic Search Ads generate targeting directly from a site's own content and Google's crawl of it, not from advertiser-chosen keywords at all. Shopping campaigns work off a product feed rather than keyword targeting in the traditional sense.

Every single route into AI Mode, before this test, ran through automation. Exact and phrase match, the two match types built specifically to give an advertiser tight, deliberate control over which queries can trigger an ad, were structurally locked out. That was not an accident or an oversight Google forgot to fix. It reflected a real product decision: AI Mode's conversational, generative surface was treated as automation-only territory, the same way AI Max and Performance Max already dominate that inventory. This test is the first crack in that wall, and it is worth being precise about what kind of crack it is: a small, undefined-scope experiment, not a policy change with a documented rollout plan.

The Reporting Gap: What You Currently Cannot See

Here is the part of this story that should worry a performance-focused advertiser more than the eligibility change itself. Google Ads reporting currently has no way to separate AI Mode ad delivery from standard Search delivery within the same keyword-level reporting rows. If your exact match keyword "emergency plumber near me" starts serving ads inside AI Mode under this test, that activity gets folded into the same impressions, clicks, and conversions column as every other Search placement that keyword already serves against. There is no AI Mode segment, no separate campaign type flag, no placement report breakout, nothing in the standard reporting UI that would let you isolate what share of that keyword's recent performance is coming from a genuinely different surface with a genuinely different user behavior pattern behind it.

Think through what that actually means in practice. AI Mode queries run roughly three times longer than traditional search queries and arrive with more conversational, specific phrasing. A user asking a fuller, more specific question inside a chat-like interface plausibly behaves differently than someone typing three words into a traditional search box, different click-through rate, different time to convert, possibly a different conversion rate entirely. If your keyword-level performance shifts over the coming weeks, and you cannot separate how much of that shift is normal seasonal or competitive movement versus a chunk of your traffic quietly rerouting through an entirely different surface with different behavioral characteristics, you are optimizing blind. A bid adjustment, a budget reallocation, a decision to pause an underperforming keyword, all of that analysis currently happens without the one piece of context that would explain whether the numbers moved because of AI Mode traffic mixing in.

This is not a hypothetical inconvenience. It is the kind of blind spot that has, historically, taken Google months to close after opening a new ad surface. Shopping ads inside AI Mode launched with roughly 75 million daily active users in February 2026 and it still took time before advertisers got any meaningful segment-level reporting on that inventory specifically. There is no reason to assume this rollout will be faster, and every reason for an advertiser managing meaningful spend on exact and phrase match to build a workaround into their own monitoring in the meantime, covered in the practical section below.

What Advertisers Should Actually Do About This Right Now

Panic is not useful here, and neither is ignoring it. A grounded, practical response looks like this.

Audit which exact and phrase match campaigns you would actually want visibility into. Not every keyword needs a special monitoring plan. Pull the list of your historically highest-performing exact and phrase match keywords, the ones carrying real budget and real conversion volume, and treat those as the priority watch list. A low-spend, low-conversion keyword picking up a few AI Mode impressions is not where your attention should go first.

Watch for impression and click pattern shifts that do not correlate with your normal search volume changes. Since there is no clean segment to check directly, the practical workaround is comparative: track your priority keywords' week-over-week impression and click trends against what you would expect from normal seasonality, competitor activity, or your own budget changes. A keyword suddenly gaining impression volume with no obvious external explanation is a reasonable signal that AI Mode eligibility might be a factor, even without direct confirmation.

Do not panic-opt-out of something you cannot currently opt out of. Google has not described any opt-out mechanism for this test, at the keyword level, campaign level, or account level. Setting negative match modifiers or restructuring campaigns in a rushed attempt to block AI Mode delivery is likely to do more damage to your normal Search performance than it does to solve a problem you cannot currently isolate anyway. Wait for Google to actually document a control before building your account around one that does not exist yet.

Keep ad copy tightly aligned to exact query intent. Given that eligibility is explicitly scoped to queries with "explicit and direct user intent," ad copy that speaks precisely to a specific, well-defined need is more likely to perform well in this environment than generic, broad messaging built for shorter traditional search queries. If AI Mode's conversational surface is filtering for specificity on the query side, matching that specificity on the ad copy side is the sensible response, even before Google publishes more detail on how the matching works.

Check in with your account rep or Google support if you manage meaningful spend. Google Ads support teams sometimes have earlier or more granular visibility into experimental rollouts like this one than what shows up in public documentation. It costs nothing to ask directly whether your account is included in the test and whether any reporting improvements are on a near-term roadmap.

Document what you observe now, before Google's own reporting catches up. If you do notice a pattern shift in a priority keyword, save the data. Once Google eventually ships AI Mode-specific reporting, having your own before-and-after record from the pre-reporting period gives you a much stronger basis for evaluating whether the placement is actually working for you, rather than starting your analysis from zero once the tools exist.

Why This Matters Especially for Video and Image Ad Creative

There is a creative production angle to this story that is easy to miss if you only think about it as a targeting and reporting question. Every time Google opens a new automated or semi-automated ad surface, the practical bottleneck for advertisers who want to actually perform well on it tends to shift toward having enough distinct, well-targeted creative variants to test, not toward the targeting mechanics themselves. AI Mode's longer, more specific, conversational queries plausibly reward ad creative that speaks precisely to a narrower slice of intent rather than one generic ad running against every possible variation of a keyword. That means an advertiser reaching AI Mode inventory for the first time through exact and phrase match is going to want more creative variants tuned to more specific intents, not fewer.

That is exactly the kind of production bottleneck a fast AI creative pipeline solves well. Miraflow's Cinematic AI Video Generator, built on models including Veo3 and Veo3.1, lets an advertiser generate polished, on-brand video ad variants from a written prompt rather than booking a new production shoot every time a new intent-specific angle needs testing. For an advertiser suddenly needing three or four distinct creative angles to test against different high-intent AI Mode queries, rather than one generic ad running everywhere, that turnaround time genuinely matters.

The same logic applies on the static side. Miraflow's AI Image Generator covers text-to-image and image-to-image generation, along with inpainting and masking tools that let an advertiser adjust an existing product image, swap a background, change a color, update packaging, without a full reshoot every time a new query intent calls for a slightly different visual angle. An advertiser testing whether a more specific, benefit-led image performs differently against a longer, more conversational AI Mode query than the generic product shot that has been running for months can iterate through several versions in a single afternoon rather than scheduling a new shoot.

Neither tool changes the targeting or reporting questions raised above. But once an advertiser has actually confirmed they are eligible for this test and started watching for the performance signals described earlier, having a fast way to produce genuinely distinct creative variants is what turns that visibility into an actual competitive advantage rather than a passive observation.

The Broader Pattern This Fits Into

This test does not exist in isolation. Google has been steadily pushing more inventory through automated and AI-driven placements all year. The AI Max for Search auto-upgrade ran from September 1 through September 30, 2026, automatically shifting eligible Search campaigns toward AI Max's broader, more automated targeting model. The planned sunset of Dynamic Search Ads was pushed back after real advertiser pushback, landing at the delayed February 2027 migration date instead of the originally planned timeline. Performance Max itself picked up a round of new steering and reporting controls earlier in 2026, largely in direct response to advertiser complaints about losing visibility as automation expanded.

The pattern across all three of those stories, and now this one, is consistent: Google keeps testing whether more automated or AI-driven surfaces can absorb inventory and targeting decisions that used to sit firmly in the advertiser's control, then adjusts based on how much pushback it gets. What makes this specific test different, and worth tracking separately from those adjacent stories rather than folding it into the same general narrative, is that it runs the pattern in reverse. Instead of automating a previously manual campaign type, Google is opening an automation-only surface to the two match types that exist specifically to resist automation. Advertisers who have deliberately stayed on exact and phrase match, specifically to avoid exactly the kind of broad, automated targeting AI Max and Performance Max represent, are the ones this test affects most directly, and they are also the group least likely to be actively watching for changes like this one, precisely because their whole account strategy has historically depended on that automation staying opt-in rather than creeping toward them.

Common Mistakes Advertisers Are Already Making With This News

A few misreadings of this story are already showing up in early discussion, and each one is worth correcting directly.

Assuming this only affects broad match or automated campaigns. This is the single most common misunderstanding, and it is backwards. Broad match, AI Max, and Performance Max already had access to AI Mode inventory before this week. This story is specifically about exact and phrase match gaining access for the first time. If your account runs primarily on tightly controlled exact and phrase match campaigns and you assume this news does not apply to you because you are not running broad match or AI Max, you have the story exactly backwards.

Treating "small experiment" as meaning ignorable. Google's own framing calls this a small experiment, and it is reasonable to take that framing seriously in terms of current scale. But Anthony Higman's pushback is worth weighing here too: the previous access model to AI Mode was not a soft default, it was a hard structural restriction that a huge amount of advertiser strategy was built around for years. A restriction that firm being loosened at all, even in a limited test, is a bigger deal than the phrase "small experiment" makes it sound, regardless of how many accounts are currently included.

Expecting granular AI Mode-specific reporting that does not exist yet. Some advertisers, on hearing this news, immediately go looking for an AI Mode filter or segment in their reporting to see the impact directly. That reporting does not exist right now, as covered in detail above. Spending time hunting for a report that has not shipped yet is wasted effort; the comparative, trend-based monitoring approach described in the practical section is the realistic option until Google closes that gap.

Assuming a defined opt-out control exists. Google has not described any way to exclude specific keywords, campaigns, or an entire account from this test. Advertisers restructuring campaigns or adding aggressive negative targeting in an attempt to block AI Mode delivery are working against a control that has not been documented, and risk breaking normal Search performance in the process. If Google ships an actual opt-out mechanism, it will show up in official documentation, not through workaround campaign settings an advertiser guesses at.

Confusing this with the AI Max auto-upgrade or the AI Overviews ad rollout. This story involves standard Search campaigns using exact and phrase match, a specific surface (AI Mode, not AI Overviews), and a specific new eligibility rule. It is a separate story from the AI Max auto-upgrade covered above and from earlier AI Overviews ad placements. Treating all of Google's 2026 AI-and-ads news as one undifferentiated blob makes it much harder to figure out which specific change actually applies to your account.

Keeping Creative Production Ready as Automated Surfaces Expand

The underlying lesson from this test, and from the broader pattern it fits into, is that the list of surfaces where your ads can show is expanding faster than most advertisers' creative production pipelines are. A test like this one can move an exact match keyword's ads into a genuinely new placement within days of Google flipping a switch internally, with essentially no advance notice to the advertiser. Waiting until a new placement is confirmed and fully documented before starting to build creative variants for it means always being a step behind.

This is where having a flexible AI creative workflow built into your existing process pays off, rather than being treated as a nice-to-have. Miraflow AI's Cinematic AI Video Generator and AI Image Generator both work directly in the browser with no separate production setup, letting a performance marketer or agency spin up a handful of intent-specific creative variants the same afternoon a new placement opportunity shows up, rather than scheduling a shoot weeks out. For an agency managing multiple client accounts across a mix of match types and campaign structures, that speed difference compounds every time Google runs another test like this one. Anyone building out a fuller content pipeline, from script to visual to finished video in one place, can start from Miraflow AI's homepage and browse more Google Ads coverage on the Miraflow blog.

Frequently Asked Questions

Is this a permanent change to how AI Mode ads work?

Not as far as Google has said publicly. Ginny Marvin described it as "a small experiment that recently kicked off," with no announced end date, no confirmed permanent rollout, and no formal policy documentation published yet as of this writing. Treat it as an active test worth monitoring, not a confirmed permanent feature, though Anthony Higman's pushback is a reasonable reminder that even a labeled experiment can represent a real, lasting shift once it starts.

Does this mean my exact match keywords are definitely showing ads in AI Mode right now?

Not necessarily. Eligibility is limited to queries Google's systems judge to show "explicit and direct user intent," a standard Google has not defined publicly, and the test itself has not been confirmed as running across every account or every country. Your keywords being technically match-type eligible does not guarantee any specific query has actually triggered an AI Mode ad placement for your account.

Which countries is this test running in?

Google has not named specific countries as part of this confirmation. No public source, including ppc.land's original coverage and Search Engine Roundtable's write-up of Marvin's statement, has published a country list as of this writing.

Can I opt my account out of this test if I don't want my exact match keywords showing in AI Mode?

Google has not described any opt-out mechanism for this test at the time of writing. Building campaign changes around a control that has not been documented is more likely to disrupt your normal Search performance than to reliably block AI Mode delivery specifically.

How is this different from AI Overviews ads or the AI Max auto-upgrade?

AI Overviews and AI Mode are different Google surfaces: AI Overviews are the AI-generated summary panels that sit above traditional search results, while AI Mode is the separate, fuller conversational search experience. The AI Max auto-upgrade, covered in Miraflow's breakdown of that rollout, automatically shifted eligible Search campaigns toward AI Max's broader targeting model, a different mechanism from this test, which specifically opens AI Mode text ad eligibility to exact and phrase match keywords for the first time.

Will I be able to see AI Mode-specific performance data for my keywords soon?

Google has not announced a timeline for adding an AI Mode reporting segment or breakout to standard keyword reporting. Given how long it took for even basic segment-level reporting to catch up after Shopping ads launched inside AI Mode in February 2026, advertisers should not assume this gap closes quickly, and building your own comparative monitoring in the meantime, as described above, is the realistic near-term approach.

Does this affect Performance Max or AI Max campaigns at all?

No, this test is specifically about extending AI Mode text ad eligibility to exact and phrase match keywords in standard Search campaigns. AI Max and Performance Max already had access to AI Mode inventory before this test, and Marvin noted they retain exclusive access to newer AI Mode ad formats like Highlighted Answers that are not part of this specific change.

Should I shift budget toward exact and phrase match now to take advantage of this?

That is a bigger strategic decision than the current information supports making confidently. The test has no confirmed scale, no defined eligibility criteria beyond the vague "explicit and direct user intent" standard, and no reporting to measure results against yet. A more measured approach is watching your existing exact and phrase match keywords for the performance signals described above before making a deliberate budget shift based on a placement you currently cannot directly measure.

Conclusion

Google confirming that exact and phrase match keywords can now serve text ads inside AI Mode is a genuinely new development, not a rewording of an existing policy. For years, reaching AI Mode ad inventory required using one of Google's more automated products: AI Max, Performance Max, broad match with Smart Bidding, Dynamic Search Ads, or Shopping. That list never included the two match types advertisers reach for specifically when they want tight, predictable control over their targeting. Now it does, at least inside a test Google is calling small, with an eligibility standard, "explicit and direct user intent," that has not been publicly defined, and with no way yet to separate AI Mode performance from standard Search performance in your own reporting.

The right response is neither panic nor indifference. Audit your highest-value exact and phrase match keywords, watch for performance shifts that do not track normal seasonality, resist the urge to build campaign changes around an opt-out control Google has not actually documented, and keep your ad copy and creative tightly matched to specific query intent, since that is likely to matter more, not less, as more of your inventory potentially runs through AI Mode's longer, more conversational queries. And as this test and the broader pattern of automated surfaces expanding continue through the rest of 2026, having a fast, flexible way to produce genuinely distinct ad creative, whether that is polished cinematic video or quickly adjusted product imagery, is going to matter more for advertisers trying to actually perform well on each new placement Google opens up.