When Google launched AI Max for Search in mid-2025, the framing was simple. It was an optional AI-powered reach expansion layer. You could enable it on a Search campaign if you wanted more reach, at the cost of some efficiency, and leave it off if you preferred tighter control. The practitioner question of the day was “when is it worth enabling?”
That framing is no longer accurate, and if your agency or in-house team is still talking about AI Max as an on-or-off decision, the conversation is overdue for an update.
Three things have changed since launch: the independent research on how AI Max actually behaves has gotten considerably less flattering, Google has confirmed an architectural detail that quietly reshapes the decision, and on April 15, 2026, Google announced that AI Max is coming out of beta with an automatic migration path for Dynamic Search Ads and campaign-level broad match starting in September 2026.
We’ve been testing AI Max across several of our client accounts over the past several months. What we’ve found aligns closely with what the best independent PPC researchers have been publishing, and it’s given us a clearer read on what this feature actually is and what the decision to use it (or not) really costs.
When we started working through the AI Max question for our clients, the answers we needed weren’t easy to find. Google’s marketing explained what AI Max does and why to enable it. Practitioner write-ups covered the performance results and the cross-matching patterns. The basic eligibility guidance (“you need broad match for AI Overview placements”) was out there. But the questions we kept running into on real accounts were further downstream. What happens if precision matters and exact match is the right fit? What does the attribution actually look like when AI Max is overclaiming and opening new eligibility at the same time? How do you read your reports when two opposite behaviors are operating at once? And at the ecosystem level, what does it mean that every advertiser’s AI Max is pulling queries from every other advertiser simultaneously? We couldn’t find clean answers to those, so we ran our own tests and wrote this up.
What AI Max actually does
AI Max is a feature suite you can enable on existing Google Search campaigns. It’s not a new campaign type. When you turn it on, three capabilities become available. They’re enabled together by default but can be toggled on and off independently, with one exception noted below.
First, search term matching expands. AI Max takes your existing keywords and matches them against a much wider pool of queries than traditional exact, phrase, or broad match would. It does this in two distinct ways that matter for how you think about the feature.
The first way is what Google calls broad match expansion. Your existing keywords are used as a primary signal, but the system also pulls in landing page content, asset content, and user context to broaden matches into semantically related queries. The keyword is an input, not a strict filter. An important consequence: AI Max treats your keywords as broad match regardless of the match type you originally assigned them. If you have a keyword as exact match and no broad match version, AI Max effectively treats it as both exact match and broad match simultaneously. This is documented in Adalysis’s testing and confirmed in Smarter Ecommerce’s million-impression study, where exact match keywords accounted for between 27% and 89% of AI Max match type impressions depending on the account. The practical implication: you can use exact match keywords with AI Max enabled, but AI Max will not behave like a traditional exact match campaign. Your keywords become signals for expansion rather than precision filters.
The second way works very differently. It’s what Google calls “keywordless” matching. The system reads your landing pages and ad content directly and matches queries based on what it thinks your business is about, independent of any keyword you’ve entered. Smarter Ecommerce analyzed over a million AI Max impressions and found these two arms contributing roughly 50/50 to campaign traffic. That means half of AI Max’s matching isn’t driven by your keyword list at all. It’s driven by what Google’s model reads on your website.
Second, AI Max can generate ad copy automatically through a feature called text customization, producing headlines and descriptions tailored to specific search intent, pulled from your landing pages and existing ads. This is on by default when you enable AI Max, but you can toggle it off independently and keep your ad copy exactly as you wrote it.
Third, AI Max can redirect traffic to whichever page on your site it thinks is most relevant to a given query, rather than the landing page you specified in the ad group. This is called Final URL expansion, and it’s also on by default. It can be toggled off as long as you want to keep text customization on. Final URL expansion requires text customization to be enabled, so if you turn text customization off, Final URL expansion turns off with it.
The practical implication: if you want AI Max’s matching and AI placement eligibility without ceding control over your ad copy or landing pages, you can configure it that way. The search term matching component can be enabled on its own. This is useful for advertisers who want the reach and placement eligibility benefits without the creative and landing page automation, but it’s worth being clear that enabling any part of AI Max changes how your keywords behave in the auction.
The early claims versus what the data shows
When AI Max launched, Google’s performance claims were specific and appealing. Advertisers activating AI Max in Search campaigns would see 14% more conversions or conversion value at a similar CPA or ROAS. For campaigns still mostly using exact and phrase keywords, the typical uplift was even higher at 27%.
Then the independent research started coming in.
Mike Ryan, Head of Ecommerce Insights at Smarter Ecommerce, documented that in one client account, a single competitor’s brand terms accounted for 69% of total Search impressions after AI Max was enabled. AI Max generated more than 8,000 impressions on competitor terms while traditional broad match generated just 77. Ryan estimated that if all competitor terms were included, the impression share for competitor queries would rise above 80%.
In a subsequent broader analysis covering 250+ campaigns, Smarter Ecommerce found median revenue uplift of 13% from enabling AI Max, largely consistent with Google’s 14% claim. But median CPA rose 16%, and ROAS ranged from +42% to −35% across campaigns. Ryan called it “essentially a coin toss: you may see a lift, but efficiency likely won’t follow.”
In November 2025, the ad verification firm Lunio published a controlled difference-in-differences analysis of 900 Search campaigns, comparing 404 AI Max-enabled campaigns against 491 controls. The finding: AI Max-enabled campaigns saw invalid traffic rise from a 3.7% baseline to 5%, a 35% relative increase. The mechanism isn’t fully understood, but the plausible read is that broader query matching pulls in lower-intent surfaces and more bot-adjacent inventory.
Named practitioners have been unusually direct about what they’re seeing. Brad Geddes, co-founder of the PPC testing platform Adalysis, publicly contradicted Google Ads Liaison Ginny Marvin’s messaging on AI Max attribution in late 2025. Xavier Mantica reported AI Max delivering conversions at $100.37 each after four months of testing, versus $43.97 on phrase match, 90% higher cost per acquisition. Andrew Goodman, founder of Page Zero Media, said he was waiting “to be proven wrong by any credible case study” on AI Max performance claims.
Adalysis’s testing surfaced a more specific problem with reporting mechanics. When AI Max is enabled, search terms that would have matched your existing exact match keywords can get attributed to AI Max in reports instead, not because AI Max did anything new, but because it’s being credited for matches your exact keywords would have caught anyway. The practical effect is that AI Max’s apparent contribution to your campaign performance can be inflated by reclassification, not real incremental volume.
This is not a fringe critique. It’s the documented consensus of the most rigorous PPC researchers publishing on the topic.
The eligibility shift that reframes the decision
The single most important development in the AI Max story happened in December 2025, and it got less attention than it deserved.
Ginny Marvin confirmed on Twitter/X that exact match keywords are not eligible to serve ads within AI Overviews. Exact match can still serve above or below the AI Overview, which is the standard ad block adjacent to the AIO box. But the placement embedded inside the AI Overview itself requires broad match or keywordless (AI Max) targeting. Marvin’s explanation: because the ad is matched to both the user query and the AIO content, only broad match or keywordless can trigger ads, even if an exact match keyword exists in the account.
The same rule applies to AI Mode, Google’s standalone conversational search surface.
This was not a policy change. It was a confirmation of how the system had been working. But the implications are significant.
Advertisers running exact-match-only campaigns have been locked out of within-AIO auctions without knowing it. Their campaigns can still serve above or below the AIO, which means they’ve seen some AI-related traffic. But the specific placement inside the AI Overview, which is where Google is increasingly putting ad inventory as the AIO expands, has been unavailable to them.
For advertisers who deliberately chose exact match for precision, this changes the calculus. The trade-off used to be “narrower reach in exchange for tighter control.” The real trade-off now includes structural exclusion from a growing share of inventory. As Google continues to expand AI-enhanced search surfaces, that exclusion will become more costly.
The only ways to be eligible for within-AIO placements are Performance Max, Shopping, or Search campaigns running either broad match or AI Max. There is no current path to within-AIO eligibility through exact match alone.
What we found in our own testing
We ran a test on one of our client accounts, a B2B advertiser in a narrow niche with specialized product terminology and a history of cross-matching problems even with extensive negative keyword lists in place (more than 12,000 negatives over the life of the account). The account had been running a single Search campaign with AI Max enabled. Our reading of the data: most of the conversions were showing up as exact match inside the campaign, so the exact match keywords were doing the real work and AI Max was layered on top for incremental reach.
We decided to split the campaign into two to validate this reading.
Campaign A kept AI Max enabled with the full keyword set. Campaign B used only the exact match keywords with AI Max disabled. Both campaigns had comparable budgets and shared bid strategy architecture. The hypothesis was that Campaign B would quietly deliver the exact-match volume that had been showing up inside the original campaign, and Campaign A would handle the AI Max expansion for incremental volume.
That is not what happened.
Campaign A continued spending its budget and converting at the expected rate. Campaign B could not spend its budget. Impression volume fell sharply. The conversion volume that had previously been attributed to exact match in the original campaign largely did not migrate to Campaign B. It stayed with Campaign A, still reported as exact match, because Campaign A still had AI Max enabled.
Campaign B’s exact match impression share remained high, around 90%+, which made the situation initially hard to diagnose. If the metric had simply dropped to 40%, we would have immediately understood that we were missing impressions to competitors. Instead, the metric looked healthy, while actual spend and conversion volume collapsed.
We contacted Google to understand what was happening. The clarification that came back, consistent with what we later confirmed from Marvin’s December 2025 eligibility statements, made the full pattern coherent.
A substantial share of the exact-match-labeled conversions in the original campaign weren’t conventional exact match serving. They were exact-phrase user queries that had served inside AI Overviews or AI Mode, placements that required AI Max eligibility to reach, but that Google was reporting as exact match because the user’s query was an identical match to an exact keyword in the account. The reporting label reflected the query-to-keyword relationship, not the placement the impression served in.
When we removed AI Max from Campaign B, we removed its eligibility to serve in those within-AIO auctions. Campaign B could still compete for traditional SERP exact match and above/below AIO inventory, and it did, which is why its impression share stayed high. But the within-AIO volume had been keeping the original campaign fed, and that volume couldn’t reach Campaign B once AI Max was off.
On closer inspection of the search terms data, two distinct behaviors had been operating simultaneously in the pre-split campaign, and they’re worth naming separately because they’re often conflated.
The first behavior is AI Max overclaiming. Some impressions labeled “AI Max” in the match type column were for search terms identical to targeted exact match keywords. These are cases where the user’s query and the exact keyword matched, but AI Max was credited for the match anyway. This is the pattern Adalysis documented. It inflates AI Max’s apparent contribution without representing new volume.
The second behavior is AI Max as an eligibility gateway. Some exact match labeled impressions only served because AI Max was opening access to within AIO placements that exact match keywords alone can’t reach. When AI Max came off, these didn’t shift columns in the report. They disappeared.
Without a controlled split test, both behaviors are largely invisible. You see the final reported numbers but not the mechanics underneath them. One behavior flatters AI Max; the other flatters exact match. The net effect depends on which dominates in a given account.
Why this is a bigger dynamic than any single account
The spillover dynamic between advertisers is the part of AI Max that’s getting the least attention in published analysis, and it’s the one with the biggest implications. The pattern we observed in our test is not unique to our account. Variations of it are showing up across the published research.
Smarter Ecommerce’s 69% competitor-impression case. Adalysis’s documentation of brand terms matching to non-brand queries, non-brand terms matching to competitors, and brand queries matching to competitor terms. Jyll Saskin Gales reported that the queries her coaching clients are seeing under AI Max are “often very generic, one-or-two-word searches, and a significant number of competitor terms.” These are all individual-account observations of what is almost certainly a symmetric dynamic. If Advertiser A’s AI Max is aggressively matching Advertiser B’s brand terms, Advertiser B’s AI Max is almost certainly doing the same thing to someone else.
Google Ads CPCs rose 10-25% across most industries in 2026, the steepest annual rise since 2021. Multiple analyses attribute this to increased competition for AI-optimized ad placements and automation-driven auction density. This is the price signature you would expect if bidirectional query-pool expansion were happening at ecosystem scale. It doesn’t prove the mechanism, but it’s consistent with it.
The implication for an individual advertiser is uncomfortable. AI Max’s expansion isn’t just bringing in incremental relevant volume. It’s also bringing in competitor queries and tangential searches, and the bidirectional pressure is pushing auction density up across the platform. “Being more careful with negatives” only partially addresses this, because half of AI Max’s matching doesn’t use keywords in the first place.
What this means for how you should think about your campaigns
The question “should we enable AI Max?” isn’t the right question anymore, because the default is changing. Starting in September 2026, Google will automatically migrate all Dynamic Search Ads, automatically created assets (the legacy name for what AI Max now calls text customization), and campaign-level broad match campaigns to AI Max. After September, new DSA campaigns won’t be creatable in Google Ads at all.
The better questions, in order of priority:
What does our AI placement exposure actually look like? AI Overview trigger rates vary enormously by query type. Informational, research-heavy queries trigger AIO most often, up to 74% for research questions, 60% for educational content. Shopping and local queries are much lower, 3-7%. Specific technical B2B queries sit somewhere in the middle. The cost of losing within-AIO eligibility depends heavily on where your query profile sits on this spectrum. An advertiser with mostly transactional, low-funnel query traffic may be losing very little by running exact-match-only. An advertiser in a category where AIO triggers frequently is losing meaningful inventory.
Is our account structurally prepared for AI Max expansion? If you enable AI Max on an account with incomplete conversion tracking, weak negative keyword coverage, inconsistent landing page quality, or insufficient conversion volume for Smart Bidding to stabilize, the expansion becomes a liability rather than a reach multiplier. The pre-flight list matters: accurate conversion tracking including enhanced conversions; comprehensive account-level negatives covering bargain-intent, career, educational, and irrelevant categories; brand exclusions on non-brand campaigns to prevent AI Max from cannibalizing your own brand campaigns; automated bidding already in place with enough historical data to convert; evidence that broad match works acceptably in the account with proper exclusions; and budget headroom. Missing any one of these and the expansion will find queries you don’t want to pay for.
How will we know if AI Max is earning its place? Google has confirmed no segmented AI Overview reporting is available, and no timeline for adding it. You cannot directly measure how much of your AI Max traffic came from within-AIO versus traditional placements. This matters for measurement: the AI Max match type label in the search terms report tells you which matching mechanism surfaced a query, not which placement served the impression. A query labeled “AI Max” could have served in traditional SERP, above/below AIO, within AIO, or in Search Partners. Without segmented reporting, the only way to size AI Max’s real contribution is through controlled testing: split tests, budget-matched comparisons, or disciplined before-and-after windows with proper guardrails.
What’s the plan if AI Max underperforms in the first 60-90 days? Smart Bidding learning periods for AI Max run longer than most advertisers expect, particularly on accounts with moderate conversion volume. The first month is noise; the second month is signal; the third month is where you decide. If your agency or team is enabling AI Max without a defined kill-switch threshold (CPA floor, ROAS floor, or spend ceiling triggering review), they’re flying without instruments. Ask what the specific failure criteria are and what the response will be when they’re hit.
Is there an internal architecture that gives us AI Max access without full exposure? A dual-campaign structure, with strict exact match as the primary campaign plus a smaller AI Max companion with aggressive fencing (comprehensive negatives, brand exclusions, text customization and final URL expansion disabled where needed, kill-switch floors in place), preserves measurement cleanliness on the primary campaign while providing a measurable test on the companion. It’s more work to maintain than either pure option. For accounts where precision matters and cross-category matching has been a documented problem, the extra complexity is often worth it.
What we’d want you to take away
Three things.
First, AI Max is no longer optional in any meaningful sense. The migration announcement in April 2026 confirmed it. Whatever you think about the feature, your portfolio will be exposed to it by September, and the decision has shifted from “whether” to “how.”
Second, the published independent research on AI Max is consistent and not flattering. 69% competitor-impression cases. 16% median CPA increase. 35% increase in invalid traffic. Named practitioners publicly contradicting Google’s performance claims. These are not anomalies. They are what the data shows when AI Max is tested rigorously and measured against its stated benefits.
Third, running exact match only campaigns in 2026 is not a neutral choice anymore. It trades away access to within AIO inventory for reporting clarity and matching precision. For some accounts, that trade still makes sense. For others, the structural exclusion is a larger cost than they realize. Either way, the choice should be made consciously, with a clear read on the account’s AI placement exposure and the measurement mechanics we’ve described.
The agencies and in-house teams worth working with right now are the ones actually testing AI Max on their clients’ accounts, measuring against Google’s claims, and pushing back where the data contradicts the marketing language. Google’s performance claims should be treated as hypotheses to test, not specifications to implement. The independent research overwhelmingly suggests those hypotheses fail in a meaningful share of accounts.
If you want to have this conversation with your team, whether that’s your own team, an agency, or a consultant, bring these questions. Ask what controlled testing they’ve done. Ask what their kill-switch criteria are. Ask how they measure AI Max’s incremental contribution when Google doesn’t provide segmented placement reporting. The quality of the answers you get will tell you a lot about whether your Google Ads spend is being managed by people who are reading the system clearly or people who are implementing Google’s recommendations without examining them.
The September migration makes this urgent for any business running Google Ads. Dynamic Search Ads and campaign-level broad match campaigns will auto-migrate to AI Max, which means advertisers who haven’t thought through how AI Max will behave on their account are about to be exposed to it by default. If you want a clear read on your own account, whether that’s your AI placement exposure, your current campaign architecture, or what to ask your own team, we’re glad to take a look.
