If you run Search campaigns that use Dynamic Search Ads, automatically created assets, or the campaign-level broad match setting, Google is auto-upgrading them to AI Max between September 1 and 30, 2026, with the rollout expected to conclude by the end of the month (Google’s announcement, and Search Engine Land on the timeline). Most of the advice being written assumes you know the flip is coming and have prepared for it. This post is for the other case: an account that got swept mid-September without a deliberate migration, where the first job is not auditing the negatives — it is finding out the upgrade happened at all.
The flip is silent by design. There is no confirmation dialog, no email you have to acknowledge; the setting changes and the campaign keeps serving. That is a problem for search-term hygiene specifically, because AI Max widens the queries your campaign can match, and a negative list that comfortably covered the old surface can start leaking the day the wider matching turns on. This is a detection-first triage: confirm whether AI Max is already live on your campaigns, see what the wider matching has let in, and reconcile the gaps the same day — before a week of silent drift becomes a week of wasted spend.
Why you may not know it happened
The auto-upgrade is a settings change Google makes on your behalf, not an action you take, so it does not announce itself the way a campaign you built does. In-account banners inviting early upgrades have been appearing near the notifications icon for eligible campaigns, but the automatic flip itself does not require you to click anything (ppc.land on the September 1 auto-upgrade). If you were not watching the account daily during the rollout window, the change can land without you noticing until performance data drifts.
The affected set is at least knowable. Because Google closed the creation of new DSA and legacy campaign types earlier in 2026, the pool of campaigns eligible for the flip is fixed — you cannot have added a new one that surprises you (Search Engine Land on the auto-upgrade scope). That means you can enumerate the campaigns at risk before you even open change history: any Search campaign with the campaign-level broad match setting on, or with automatically created assets, or running DSA, is in the queue. Confirming which of those has actually flipped yet is the detection step. The full “am I in the queue” scoping is in the AI Max auto-upgrade scope check.
Step one: confirm whether AI Max is already on
Check three places, in order of speed. First, open each candidate campaign’s change history and filter it to September — the automatic upgrade is logged as a settings change on the day it happened, so a broad-match or AI Max setting change you did not make yourself is the timestamp of your flip. Second, open Campaigns → Settings and look for the AI Max control; if it is present and enabled, the conversion is done. You can switch it off there if you decide to, but detection comes first, because you cannot reconcile a change you have not confirmed.
The third check is behavioural and the most reliable. Add the Source column to the search terms report and look for a keywordless or broad-match source bucket that was not producing traffic before — its appearance is the clearest signal that AI Max is live and matching. Change history tells you the setting changed; the Source column tells you the matching behaviour changed, which is the part that actually affects your negatives. If change history shows the flip but the Source column shows no new buckets yet, the campaign was upgraded but has not accumulated wider-match volume, and you have a short head start to reconcile before it does.
Step two: add the Source column and read what it lets in
The Source column is the single most useful view for a post-flip account, because it splits your search terms by how Google matched them — keyword-driven traffic versus AI Max broad-match and keywordless matching. It is not shown by default, so add it explicitly: on the search terms report, open the columns editor and enable Source, then segment by it. Once it is there, sort the keywordless and broad-match buckets by cost and read them as their own report, because that is precisely the traffic your old negative list was never tested against.
Reading the new buckets in isolation is what makes the reconciliation finite. Under keyword matching you had a stable query surface; under AI Max you have that surface plus whatever the wider matching adds, and conflating the two hides the new leaks inside familiar totals. Segmenting by source isolates the delta so you can act on it directly. The mechanics of segmenting the AI Max report and negating from it are covered in segmenting the AI Max search terms report to negate, and the broader question of how much of your spend is even visible in the report — which AI Max complicates — is in the search terms visibility ratio audit.
What actually happens to your negatives
Your negatives do not switch off. Negative keywords are still respected after a campaign is upgraded to AI Max — account-level negatives, shared lists, and campaign-level negatives all keep filtering, and nothing is deleted in the flip (Google’s AI Max FAQ). The failure mode is subtler than deletion: the same list is now filtering a larger, more varied stream of queries than it was built for, so gaps that never mattered under narrow matching start letting spend through. Weak negatives become expensive quickly under wider matching precisely because they are now doing more work against more traffic.
There is a second-order effect worth naming. Broad-match-era accounts often used negatives as fences — blocking a term in one ad group so it served in another — and AI Max’s different routing can turn a fence into a wall, or make it irrelevant. So the reconciliation cuts both ways: some negatives now under-cover a wider surface, and a few may now over-block traffic you want. The conflict mechanics that decide whether a given negative even applies under AI Max — and how match types interact with the new matching — are in AI Max negative keyword conflicts. This post is the account-level detection and reconciliation that sits on top of those mechanics.
Step three: reconcile the gaps the same day
Start with the cheapest, highest-consequence check: confirm every upgraded campaign still has its negative lists attached. Attachment loss during a flip is rare, but a campaign running AI Max’s wide matching with no exclusions attached burns budget from the first impression, so it is worth the two minutes to verify before anything else. Open each upgraded campaign, confirm its shared lists are applied, its campaign-level negatives are present, and account-level negatives show active. If a list detached, reattach it now — that single fix outweighs the rest of the audit combined.
Then work the Source column’s new buckets by cost and group what you find rather than negating query by query, because AI Max will keep generating variants of each new intent. If a cluster of research-intent or adjacent-product queries has appeared, block the pattern with a phrase negative instead of chasing each string — the pattern-grouping discipline in grouping search terms into patterns matters more here because the query surface is wider. If you deliberately prepared for the migration and want the full structured pass rather than a same-day triage, the companion piece is the after-migration negative keyword audit; this post is the version for accounts that discovered the flip after it happened.
Step four: schedule the deeper pass
A same-day triage stops the immediate bleeding; it does not finish the job. Your first post-flip report only holds a few days of wider-match data, which is enough to catch the obvious leaks but too thin to show the stable pattern. Schedule a deeper review two to three weeks after you confirm the upgrade, once there is enough volume in the keywordless and broad-match buckets to negate structurally rather than react to noise. By then the new query clusters AI Max favours will be clear, and any negative that turned into an over-block will have shown up as a suppressed converting term.
After that, fold the upgraded campaigns into your normal weekly search-terms cadence and treat them as the higher-variance campaigns they now are. Because AI Max reaches wider, it produces new queries faster, so the migrated campaigns are the ones most worth reviewing weekly rather than monthly. The detection check in this post is a one-time event tied to the September rollout; the Source-column review it points you to becomes a permanent part of the cadence. Run the detection once to find out where you stand, then keep the segmented report in your weekly routine so the next expansion of matching never drifts silently again.