Ads now appear in 25.5% of AI results, up 394% year over year. And yet Google just opened a side door most demand gen teams didn't see coming: traditional Search campaigns using exact and phrase match keywords can now serve text ads inside AI Mode. No AI Max. No Performance Max. Just your existing campaigns, if Google decides the user's intent is "explicit and direct."
Google Ads Liaison Ginny Marvin called it a "small experiment." That qualifier matters. But the signal underneath it matters more.
Why This Isn't Another Beta
Until this test, the working assumption across most B2B SaaS paid teams was clear: AI Mode ad visibility required AI Max for Search or Performance Max. Traditional keyword-targeted campaigns were locked out. That assumption shaped account restructures, budget requests, and roadmap conversations with RevOps throughout early 2026.
Now Google is testing whether tightly targeted exact and phrase match campaigns can participate too, at least for queries where the intent signal is unambiguous. The catch: Google's systems decide what qualifies. You don't opt in. You get to be eligible.
The Control Problem Gets Sharper
Exact and phrase match campaigns give you tighter keyword control. AI Max gives you broader reach and access to newer formats like Highlighted Answers. Google's own marketing materials claim advertisers using AI Max see 27% more conversions at a similar CPA or ROAS versus campaigns mostly using exact and phrase match.
"Similar CPA" is doing a lot of work in that sentence. Early performance data from AI Mode ads tells a more complicated story: 18% higher CTR alongside roughly 35% higher CPC compared to traditional search ads. If your unit economics are tight, that CPC inflation eats margin fast.
Then there's the zero-click problem. Reports from 2026 put AI Mode's zero-click rate between 92% and 94% of sessions. Your ad might show. It might even get clicked at a higher rate. But the denominator of sessions where anyone clicks anything is small. Position-one CTR for informational keywords dropped from 0.076 in December 2023 to 0.039 by December 2025. For keywords that later triggered AI Overviews, the decline was steeper: 0.073 down to 0.016 over the same period.
What to Do This Week
Don't restructure your account around this test. It's a limited experiment with no commitment to permanence. But don't ignore it either. Here's a diagnostic you can run in under an hour:
- Audit your exact/phrase match campaigns for high-intent queries. Pull your top 20 converting exact match keywords. Would Google classify these as "explicit and direct"? Bottom-funnel terms like "[competitor] vs [your brand] pricing" are more likely candidates than mid-funnel educational queries.
- Check Search Terms reports for AI Mode impressions. If you're in the test, you may already be getting AI Mode delivery without knowing it. Look for impression spikes on exact match campaigns that don't correlate with volume changes in traditional SERPs.
- Set a CPA guardrail before you see the data. If AI Mode impressions are flowing, your blended CPC will shift. Define your stop-loss now: a 20% CPA increase threshold is reasonable for a two-week observation window.
The hypothesis, stated falsifiably: if exact match campaigns are serving in AI Mode, then CPC will rise 15–35% on affected keywords while conversion rate holds or improves, because the user intent filter should pre-qualify traffic. If conversion rate drops alongside CPC increases, the intent gate isn't working for your queries and you should exclude or segment.
Measurement Has to Change Before Budgets Do
The real risk isn't wasted spend on a small test. It's measuring AI Mode the same way you measure traditional search and drawing wrong conclusions. When 93% of sessions produce no click, CTR becomes a misleading primary metric. Brands cited in AI Mode organic responses reportedly see 35% more organic clicks and 91% more paid clicks, according to one performance study. That's an assist effect, not a last-touch conversion.
Track downstream pipeline influence. Run a holdout if your volume supports it. Compare assisted conversion paths for queries where AI Mode ads served versus where they didn't.
Success = stable or improving cost-per-qualified-opportunity on affected campaigns. Guardrails = CPA and CPC monitored weekly. Stop-loss = pause AI Mode-eligible campaigns if blended CPA exceeds target by more than 20% for two consecutive weeks.
Google controls the intent gate. You can't see the classification logic. The experiment could disappear next month. That's the same asymmetry that's been building across every Google automation layer for years, now extended into AI Mode. The teams that'll handle this well are the ones already measuring incrementality rather than relying on platform-reported attribution. For everyone else, this test is less about AI Mode and more about whether your measurement stack can tell you what's actually happening when Google changes the rules underneath you.
