Quick answer
Exact and phrase match buy predictability. AI Max buys reach by matching queries beyond your keyword list using page and asset context. The decision is not which is better, it is which controls you need and whether you can measure qualified leads well enough to judge the extra reach.
- What each matching mode actually controls
- The migration inventory to take before switching
- A controlled experiment with lead-quality rules
Every few years Google widens how a keyword can match, and every few years the same argument follows. The useful version of that argument is not about trust. It is about which controls you keep, what you can observe, and whether your lead definitions are good enough to tell a wider net from a worse one.
The control matrix
| Dimension | Exact match | Phrase match | AI Max matching |
|---|---|---|---|
| Query range | Narrow, close variants only | Meaning-preserving variants | Broadest, extends past the keyword list |
| Primary input | The keyword | The keyword and its meaning | Keywords plus landing pages and assets |
| Spend predictability | High | Moderate | Lower, needs budget and negative discipline |
| Discovery of new terms | Minimal | Some | Its main argument |
| Main risk | Missed demand | Drift at the edges | Unqualified volume that looks like growth |
What Google clarified
Search Engine Roundtable reported Google's AI Max keyword-prioritization clarification on September 1, 2026. Current Google Ads Help documentation confirms that broad keywords are not treated as exact-matched keywords for prioritization after the broad match campaign setting is disabled or moved to AI Max, and that advertisers who prefer standard exact-match prioritization need to take no action. That is the whole of the confirmed change. It is a prioritization detail, not a statement about performance.
The migration inventory
Before changing a matching setting on a campaign that currently works, write down what you are about to put at risk.
- Every keyword with conversions in the last 90 days, and its match type.
- The current negative keyword lists, at account and campaign level.
- Location targeting and the presence or absence of location exclusions.
- Which landing pages the campaign is allowed to send traffic to.
- Current conversion actions and which of them count as primary.
- A baseline export of search terms, so you can compare the mix later.
Geographic and URL controls
For a Vancouver advertiser these two controls do most of the work. Set presence-based location targeting rather than interest, so you are not paying for people researching Vancouver from elsewhere, and exclude the regions you cannot serve. Then restrict the URLs the system may use, so wider matching cannot route a query to a page that was never meant to receive paid traffic.
Run it as an experiment
- One variable. The matching setting. Not the bidding strategy, not the budget, not the creative.
- A fixed split and a fixed run length, agreed before launch and written down.
- A lead-quality definition set in advance, in your CRM terms: what counts as qualified, who decides, and how quickly.
- A search-term review cadence during the run, with negatives added consistently to both arms or to neither.
- An evaluation date, so the decision is not made on the first bad week or the first good one.
Reporting that survives scrutiny
Report the query mix, not only the totals. Show which share of spend went to terms that existed in your keyword list before the test and which share went to new terms, then show qualified leads for each. That single split answers the only question that matters: did the extra reach bring work worth having.
Decision rules
| Observation | Decision |
|---|---|
| More qualified leads at acceptable cost | Keep, then tighten negatives and re-baseline |
| More conversions but qualification rate falls | Fix the conversion definition before judging the setting |
| Volume up, qualified leads flat | Revert, and record why in the change log |
| Result inside normal week-to-week variation | Extend the run rather than concluding |
Sources
- Search Engine Roundtable, AI Max keyword prioritization clarification, September 1, 2026.
- Google Ads Help documentation on keyword matching, AI Max for Search campaigns, and campaign experiments.
Google Ads account metrics, CPC, spend, conversions, CPA, ROAS and lead quality figures are not included here because they are account-specific and were not available for this article.
Our PPC advertising page explains how we structure and control paid search, and the Vancouver PPC agency page covers how that works for local advertisers. To review your own match-type setup, contact us.
Thinking about switching match types?
Book a free 30-minute consultation and we will map the controls and the test before anything changes in the account.
Book My Free Consultation →