Bottom line: Google’s current AI Max help documentation adds a keyword-prioritisation rule that matters during the September 2026 migration. When a campaign-level broad-match setting is disabled or a campaign moves to AI Max, broad keywords are not treated as exact-match keywords for prioritisation. If preserving the previous prioritisation behaviour is important, Google says to add copies of the exact-match keywords to the new AI Max campaign. Converting broad keywords to exact match can remove their historical keyword statistics. If the standard exact-match prioritisation is acceptable, Google says no action is required.
This is a configuration and measurement decision, not a performance result. The Google Ads Developer Blog says Phase 1 of the migration starts on 1 September 2026 and rolls progressively through 30 September for Search campaigns using campaign-level broad match or standalone automatically created assets. Dynamic Search Ads are a separate phase rescheduled to February 2027. Your account may not be migrated at the same time as another account, so use account-level read-back before changing budgets or drawing conclusions.
What the new AI Max rule changes
The new paragraph is easy to miss because it sits inside Google’s broader explanation of AI Max. It answers a narrow question: what happens to keyword prioritisation when the campaign-level broad-match setting is turned off or the campaign is moved to AI Max?
Google’s answer is that broad keywords are no longer treated as exact-match keywords in prioritisation. That means the fact that a keyword is written with the same words as a query does not, by itself, preserve the special prioritisation behaviour that existed under the campaign-level broad-match setting. Google’s documented remedy is to add copies of the exact-match keywords in the new AI Max campaign when that prioritisation is required.
The wording does not say that every account must add exact copies. It gives a choice. If the normal exact-match rules in the AI Max account are the intended design, Google says no action is required. The right choice depends on the account’s brand/non-brand structure, measurement plan, keyword governance and reason for enabling AI Max.
Why the September timing matters
Google’s 12 August 2026 Developer Blog divides the legacy migrations into phases. Phase 1 covers campaign-level broad match and standalone automatically created assets from 1–30 September, with a progressive rollout. Phase 2 covers Dynamic Search Ads and is rescheduled to February 2027. Do not combine the two dates in a client message or describe every Search campaign as part of the September phase.
The same notice says Google blocked creation of new legacy campaign-level broad-match configurations and legacy automatically created assets across the UI, Google Ads Editor and the API from 3 August. Existing campaigns still need an inventory. A date in Google’s announcement is a rollout window; it is not proof that a particular campaign has already changed.
Two valid prioritisation choices
| Account objective | What to verify | Documented decision |
|---|---|---|
| Preserve the previous broad-setting prioritisation for selected terms | Identify the exact terms whose routing matters, then compare the existing campaign structure with the new AI Max campaign. | Add exact-match copies in the new AI Max campaign, following Google’s guidance. Record the reason and any governance owner. |
| Use standard exact-match prioritisation in the AI Max account | Confirm the account’s Search and Performance Max structure and ensure identical exact keywords are eligible where the business needs that preference. | No special copy action is required under Google’s current wording, but normal keyword, brand and query controls still apply. |
Do not treat the first row as a universal recommendation. Exact-match copies can make ownership clearer for a protected brand or high-value query, but they also add objects to govern. Conversely, doing nothing may be correct when the account intentionally accepts the standard prioritisation logic. Write the objective before changing the campaign.
Historical statistics: the trade-off to record
Google says a broad keyword can be converted to exact, but that conversion removes historical keyword statistics. That is a data-continuity warning, not a statement that historical data disappears from every report. Before editing, export the relevant keyword, search-term, campaign and conversion views using the account’s existing date and attribution definitions. Keep the original match type, campaign, keyword text, status and export time in the change receipt.
If the team chooses exact-match copies instead of converting the broad keyword, record both the old and new objects. Define which object owns the query, how negatives and brand controls apply, and which report will be used for the comparison. Avoid counting the same business outcome twice simply because the account has two keyword objects with related text.
How Google’s priority ladder fits together
Google’s current keyword and ad-group prioritisation guidance still describes a ladder rather than a promise that a chosen keyword will always serve. At the top, an eligible exact-match keyword identical to the query is prioritised over broad or phrase keywords and a Performance Max campaign. Identical phrase and broad keywords, including AI Max, share a lower priority with an identical Performance Max search theme. AI-based relevance then narrows eligible ad groups, and Ad Rank resolves cases that remain tied.
Eligibility can override an expected preference. Google lists budget limits, low search volume, disapproved creatives or landing pages and unmet targeting requirements as examples. Therefore a test receipt should record eligibility state, budget status, approvals, targeting, query, landing page and Ad Rank context where available. A keyword-priority rule is not a guarantee that an ad will enter every auction.
Migration-readiness checklist
- Export campaigns that use campaign-level broad match or standalone automatically created assets. Keep account ID, campaign ID, campaign name, API version and extraction time.
- Separate campaign-level broad match from ordinary ad-group broad keywords. The September Phase 1 notice does not make every broad keyword a migration candidate.
- List exact-match terms whose routing or brand protection has a business reason. Do not duplicate an entire keyword list without an objective.
- Choose the prioritisation path: exact copies to preserve the prior behaviour, or standard AI Max prioritisation with no special copy action.
- Record whether a broad keyword was converted, copied or left in place, and preserve the historical-statistics warning in the change log.
- Audit brand inclusions and exclusions, negatives, locations, language, budgets, bidding, approvals, final URLs and tracking templates before comparing results.
- Set a predeclared comparison window and one primary outcome, such as qualified leads or completed purchase value. Keep CRM qualification and attribution settings visible.
- Read back search terms, match-type/source columns, assets, selected landing pages and migration timestamps. A timestamp proves an implementation event, not incremental conversions.
Do not let URL automation contaminate the test
AI Max can use search-term matching, text customisation and final URL expansion. If final URL expansion is enabled, Google warns that pinned assets may not be respected when a more relevant URL is selected. Google also documents possible 404s when a dynamic landing page does not work with the account’s tracking template. Sample the primary page, a deep page, a location page, a redirected page and a URL with tracking parameters before treating an AI Max change as a keyword-only test.
Keep the landing-page offer, canonical, redirects, consent, form, analytics and tracking parameters comparable between control and treatment. If URLs, budget, conversion actions and keyword governance all change at once, label the result as a multi-variable change rather than attributing a movement to prioritisation.
Measurement and rollback
Use a receipt that separates implementation facts from outcomes. The implementation section should contain the migration timestamp, setting state, keyword objects, exact-copy decision, controls and approvals. The outcome section should contain the predeclared period, spend, clicks, conversions, qualified rate and revenue definitions. A modelled scenario or platform conversion is not proof of incremental or qualified business value.
Set a stop condition before launch: tracking loss, unexpected query mix, wrong landing page, brand-safety issue, qualified-rate deterioration, budget breach or material error. Assign a rollback owner and keep the previous export. If the account lacks a holdout or a clean comparison, lower confidence rather than turning a directional observation into a causal claim.
What this means for marketers
The immediate task is account inventory and query governance. Teams that relied on campaign-level broad match to give a broad keyword exact-like priority should identify those terms before migration and decide whether exact copies are worth the added governance. Teams that prefer the ordinary AI Max priority can document that choice and focus on eligibility, query review, brand controls and landing-page integrity.
Keep the broader AI for Google Ads guide as the context owner, and use the target bidding changes checklist when bidding is part of the test boundary. If language settings change, check the Google Ads language-targeting guide. Technical teams can use the Google Ads API v25.1 article as adjacent implementation context, not as proof that a specific account’s migration fields were queried.
For measurement after the campaign change, the Search Console AI performance-report owner covers organic AI visibility and should not be mixed into a paid-search attribution claim. Integration governance is covered separately in the Google Ads developer-policy update. These links are context, not evidence that those systems caused an Ads result.
Frequently asked questions
Does every broad keyword need an exact copy?
No. Google’s current guidance gives a conditional choice. Add exact copies when the account needs to maintain the prior prioritisation behaviour; otherwise the standard AI Max exact-match rules may be acceptable.
Does converting broad to exact preserve the keyword’s historical statistics?
Google warns that converting broad to exact removes historical keyword statistics. Export the relevant views first and document whether the team copied or converted the object.
Does a migration timestamp prove better performance?
No. It proves that a migration record exists. Compare a predeclared, like-for-like period with platform, site and qualified business outcomes.
Are Dynamic Search Ads part of the September migration?
No. Google’s Developer Blog separates DSA into Phase 2, rescheduled to February 2027.
Bottom line
Google’s exact-match prioritisation note makes September’s AI Max migration a governance decision as well as a settings change. Inventory eligible campaigns, decide which queries need protected prioritisation, record the historical-statistics trade-off and keep the test interpretable. Let account read-back and qualified outcomes—not a rollout date or a migration timestamp—decide what happens next.
Editorial method and source credits
Method: This refresh separates current Google documentation from secondary discovery, account-specific unknowns and performance inference. It preserves WP2839’s existing URL and owner. No Google Ads account or API query was accessed.
- Primary: Google Ads Help, “How AI Max for Search campaigns works”, accessed 1 September 2026.
- Primary: Google Ads Help, “About the broad match keywords campaign setting”, accessed 1 September 2026.
- Primary: Google Ads Help keyword matching and ad-group/asset-group prioritisation pages, accessed 1 September 2026.
- Primary: Google Ads Developer Blog migration notice, published 12 August 2026, accessed 1 September 2026.
- Discovery: Search Engine Roundtable’s 1 September 2026 AI Max prioritisation report; secondary context only.
- Visual credit: DMT-authored decision table and migration-readiness flow; no vendor screenshot or logo used.
