Google’s own AI Mode push is driving a paid search cannibalization problem, and advertisers aren’t waiting for the earnings call to react. In November 2026, agencies told Digiday they’d cut Google and Performance Max spend by half, the same week Google told advertisers an 80 percent confidence interval is good enough for test results.
Key takeaways
- Agencies at Crowd Louder Media and ZGM Modern Marketing Partners cut Google and Performance Max spend up to 50 percent in 2026, moving budget to the open web, per Digiday.
- Google has acknowledged AI Mode earns less revenue per query than classic search, even while planning roughly $200 billion in AI infrastructure spending for 2026, per KPBS.
- Google’s Ads Decoded podcast told advertisers in November 2026 an 80 percent confidence interval is fine for directional calls, looser than the 95 percent its help docs specify, per PPC Land.
- Paid search isn’t dying. The baseline you’re testing against is moving, right as Google asks you to trust that baseline less.
What is paid search cannibalization, and is it really happening at Google?
Paid search cannibalization happens when a search engine’s own AI answer satisfies a query that used to send a click to a paid ad. Google built AI Mode to answer questions directly on its own interface, and that’s exactly the overlap that eats the clicks advertisers pay for.
Google hasn’t hidden this. Alphabet chief Sundar Pichai called it an innovator’s dilemma in remarks covered by KPBS in November 2026: “I think the dilemma only exists if you treat it as a dilemma.” Former Yahoo chief Marissa Mayer put the harder question more plainly in the same report: “How do I take that advertising spend and move it to a new format that keeps my advertising customers happy?”
The numbers are large. Google’s search ad business topped $280 billion in annual revenue by 2022 (an older figure, flagged since it predates AI Mode), and the company plans roughly $200 billion in AI infrastructure spending for 2026 alone, per KPBS. We’ve tracked the budget side before: most of the AI search budget companies are adding is old SEO budget with a new name, and paid search only absorbed the AI Overview hit for some brands.
Why are advertisers pulling spend out of Performance Max?
Advertisers are pulling spend because Performance Max, Google’s automated ad product, is getting harder to explain and costlier to justify. You hand it a budget and goals, and it blends search, Shopping, display, and video with little advertiser control over where the money lands.
At Crowd Louder Media, John Davis told Digiday some clients “have transitioned out of Performance Max completely, to a mix of Google ads, search campaigns” and open web buys. ZGM Modern Marketing Partners’ Sara Kerr cut a client’s spend by half once its cost per impression matched plain search.
TJ Kropp, head of search at Ramp97, told Digiday that “when campaign performance using their AI black box dips, there is no why.” You can’t fix what you can’t diagnose.
None of this stopped adoption. A Tinuiti report cited by Digiday found 95 percent of retail advertisers running Shopping ads had adopted Performance Max, 69 percent of median shopping spend. Wide use and wide distrust aren’t a contradiction here.
Why is Google lowering the confidence bar right now?
Google is lowering the bar because it wants advertisers making faster, more frequent calls on automated test results, not fewer of them. A confidence interval is a statistical way of saying how sure you can be that a result is real and not just noise, and a lower bar means advertisers act on weaker signals.
On its Ads Decoded podcast on November 18, 2026, Google said lower confidence is fine for everyday decisions. Host Ginny Marvin relayed it as “an 80 percent confidence interval for directional changes, and 95 percent for major structural ones.” Search experiments specialist Al Nevarez added that 80 percent “is typically good enough when combined with the advertiser’s judgment.”
Here’s the tension nobody at Google said out loud: its help docs still specify a 95 percent standard for ad variation testing, so this isn’t a clarification. It’s a lower bar, introduced the same month Performance Max’s reputation is pushing advertisers out.
| Signal | What Google tells advertisers | What advertisers report |
|---|---|---|
| Test confidence | 80 percent is fine for directional calls | Help docs still lean on 95 percent |
| Performance Max | An efficient way to reach shoppers | Costs now match plain search |
| AI Mode | A better way to answer queries | Earns less revenue per query |
Google wants you to trust a looser standard right when its own products are hardest to audit.
That’s not a coincidence of timing. It’s what happens when the company grading the test also owns the system being tested.
What should you actually do about it?
- Pull your CRM or sales numbers alongside Google’s reported conversions before calling anything a win.
- Keep your confidence bar at 95 percent for any Performance Max budget shift.
- Test a slice of budget on the open web, the way Crowd Louder Media’s clients did, for a benchmark Google doesn’t control.
- Ask your rep why a number moved. If “the algorithm learned” is the whole answer, treat it as unverified.
Frequently asked questions
What is Performance Max? Performance Max is a Google Ads campaign type that places ads across Search, Shopping, Display, YouTube, and Gmail from one budget and one set of goals. Advertisers get limited visibility into which exact placement drove a given result, which is the core complaint behind the pullback described above.
Why is Google’s ad revenue growing slower even as search volume rises? Google’s own leadership has said AI Mode queries, answered inside Google’s interface rather than on a traditional results page, earn less revenue per search. Search volume keeps climbing, but more of it lands on that lower-monetizing surface, so revenue no longer tracks volume the way it once did.
Is an 80 percent confidence interval reliable for Google Ads tests? For small, reversible changes, yes, that’s what Google’s own November 2026 guidance recommends. For a decision that moves real budget, such as shifting spend out of Performance Max, stick with the stricter 95 percent standard that Google’s own help documentation already uses for significance testing.
Sources: Digiday reported the Performance Max spend cuts and agency quotes in 2026. KPBS, an NPR member station, covered Google’s AI Mode monetization gap in November 2026. PPC Land reported Google’s confidence guidance from the Ads Decoded podcast on November 18, 2026. Treat the Performance Max adoption figures as vendor interest data, not an independent audit.