If your non-brand prospecting is stalling but branded search is still printing conversions, AI Max can quietly blend the two—and make your pipeline reporting look “better” while getting less incremental. That’s the constraint. And it’s exactly why Google’s reported test of a dedicated Branded Searches control inside AI Max matters right now. (Source: [1])
In the early test, the control reportedly offers three modes: show ads on all relevant searches (default), control branded searches using brand inclusions/exclusions, or show ads only on unbranded searches. (Source: [1]) Read that again. It’s basically an admission that “broad-match + keywordless expansion” needs a steering wheel when brand demand is involved.
The timing isn’t subtle either. Reporting says legacy Search features like Dynamic Search Ads (DSA) will automatically upgrade to AI Max starting in September 2026. (Source: [2]) So even if AI Max isn’t on the roadmap, it may end up on the account.
If you only change one thing, change this: set up a controlled brand-vs-nonbrand experiment in AI Max with explicit guardrails, before the DSA→AI Max migration forces your hand.
Why this matters now: AI Max expands matching, and brand is the easiest place to “find” conversions
Google’s own help page says turning on AI Max activates search term matching by default, using broad-match and keywordless technology to expand reach. (Source: [4]) Expansion is the point. The problem is that brand queries are the lowest-friction place for the system to pick up extra conversions—especially in B2B where brand demand often correlates with late-stage buying.
Google’s positioning is that AI Max typically drives 14% more conversions or conversion value at a similar CPA/ROAS, and that campaigns relying mostly on exact/phrase can see a 27% uplift when activating AI Max. (Source: [3]) Those are attractive numbers. They’re also exactly why demand gen leaders need to separate two questions that dashboards love to blend:
1) Did performance improve? And
2) Was it incremental?
Branded search is where “incremental” goes to die—unless it’s handled with intent. A control that can force unbranded-only behavior (or at least govern branded matching) is a governance feature, not a nice-to-have.
The control is a test, but the direction is clear: governance is becoming a product feature
Google has said AI Max includes brand controls at both the campaign and ad group level. (Source: [3]) And a Google Help article describes brand controls as a way to specify which brands ads can be associated with—or prevent association with specific brands. (Source: [5])
So the “Branded Searches” toggle isn’t coming out of nowhere. It’s a more explicit interface for a long-running advertiser need: keep brand demand clean while still letting automation hunt for net-new.
But the context, however, is more complex. Hands-on testing feedback in the reporting is mixed: one tester said AI Max generated more headlines/descriptions and slightly improved CTR, but didn’t save meaningful time and didn’t produce major performance gains (yet). (Source: [1]) That’s a very operator-friendly takeaway: AI Max might add reach and creative variation, but it doesn’t automatically create incrementality.
A skeptical practitioner view goes further. A YouTube commentator argues AI Max isn’t recommended for most businesses yet, reporting better results from Dynamic Search Ads for keyword-less targeting and advising advertisers to build more conversion data before leaning heavily into automation. (Source: [2]) That’s not anti-automation. It’s sequencing.
One primary tactic: run a branded-search holdout inside AI Max
This is the cleanest move because it matches the risk: AI Max expands matching by default (Source: [4]), and the new “Branded Searches” options (if available) are explicitly about controlling that expansion around brand. (Source: [1])
The hypothesis (make it falsifiable): If AI Max is allowed to serve on unbranded searches only (or brand-restricted via inclusions/exclusions), then non-brand qualified pipeline will increase at a similar or slightly higher CPA because the system can expand reach without siphoning credit from branded demand. This should show up as a lift in non-brand conversion value/conversions (directionally) while branded conversion volume stays stable in the brand campaign.
When this is wrong: if the account doesn’t have enough conversion signal, or if non-brand intent is too thin in the category, forcing unbranded-only can just reduce volume without finding replacement demand. That’s not a failure. It’s an answer.
Run it this week (operator-ready)
Setup: Create a controlled test structure that prevents brand from “masking” results.
- Audience: Search traffic only; keep geo consistent with your current search footprint (AI Max supports location controls). (Source: [4])
- Campaigns: One brand campaign stays as-is (your baseline). One AI Max test campaign is configured to avoid brand blending using the “Branded Searches” control if present, or brand inclusions/exclusions/brand controls at campaign/ad group level where available. (Sources: [1], [3], [5])
- Budget range: Keep it boring. Start with 10–20% of existing non-brand search spend in the test to limit blast radius. (Directional guidance; spend depends on your baseline.)
- Timeline: 2–3 weeks minimum so the system exits the “new toy” phase; longer if conversion volume is low.
- Owners: Paid search owner + RevOps/analytics partner for attribution sanity checks.
Launch: Turn on AI Max in the test campaign, but treat brand as a protected asset. Google says AI Max includes brand controls and improved search-term visibility as it moves out of beta. (Source: [3]) Use that visibility to confirm what queries are actually coming through—don’t trust labeling alone.
Readout: Compare against the brand baseline and your historical non-brand baseline. Don’t declare victory because blended ROAS went up. That’s how teams accidentally buy their own demand back.
Next test: If unbranded-only is too restrictive, move one notch toward “control branded searches via inclusions/exclusions” rather than opening the default “all relevant searches.” (Source: [1]) Small steps. Fewer surprises.
Success metrics, guardrails, stop-loss
Success = lift in non-brand conversions or conversion value at similar CPA/ROAS and stable branded campaign volume (directional, not definitive, using platform + CRM checks). Google’s own benchmark claim is 14% more conversions/value at similar efficiency; use that as a reference point, not a promise. (Source: [3])
Guardrails = (1) branded campaign conversions don’t drop materially, (2) search term mix doesn’t drift into brand/competitor adjacency you wouldn’t approve. Brand controls are explicitly meant to manage brand association. (Source: [5])
Stop-loss = if the test campaign’s CPA rises sharply and branded campaign volume falls at the same time, pause and audit query mix and brand settings. That pattern often signals brand cannibalization or mis-scoped exclusions.
The open question at the start was simple: can AI Max expand reach without muddying brand attribution? Google appears to be building toward “yes,” one control at a time. (Sources: [1], [3], [4])
September 2026 is the real deadline—not because anyone needs to panic, but because automatic upgrades remove the luxury of ignoring governance. The teams that win won’t be the ones with the fanciest automation. They’ll be the ones who set the guardrails early, then let the machine work inside them.
