On September 1, 2026, Google Ads flips a switch you didn't ask for. Any search campaign using Automatically Created Assets or campaign-level Broad Match gets upgraded to AI Max automatically — and with it, new defaults for how queries match your ads and how your ad text gets rewritten.
If you manage paid search, that sentence should sting a little. Someone else is about to change your matching logic and your copy overnight. The good news: you have runway, and you have levers. This guide is the audit, the migration playbook, and the rollback plan — so September 1 is a controlled rollout instead of a surprise.
AI Max is Google Ads' campaign-level AI mode that centralizes automated query matching and creative customization. Starting September 1, 2026, campaigns using Automatically Created Assets or campaign-level Broad Match upgrade automatically, with search term matching on by default and text customization on for ACA campaigns. To keep control: audit affected campaigns now, export 90-day baselines and creative backups, then choose a path — opt in early with staged tests, opt out by disabling the legacy features, or accept the auto-upgrade with strict negatives, pinned assets, and conservative bids.
What actually changes on September 1
Four terms do most of the work in this migration. Learn them once and the rest of the plan reads itself:
| Term | What it is | What changes under AI Max |
|---|---|---|
| Automatically Created Assets (ACA) | System-generated headlines and descriptions built from your landing pages and existing assets | Migrated automatically, with text customization enabled by default |
| Campaign-level Broad Match | A setting that expands query matching beyond keyword-level match types at campaign scope | Migrated automatically, with search term matching enabled by default |
| Search term matching | The automated logic deciding which queries trigger your ad, using intent signals and context | Becomes the default matching mode — broader query sets are the norm |
| Text customization | Dynamic insertion or rewriting of headlines and descriptions to fit the triggering query | Enabled by default for ACA campaigns — your copy can now drift |
Two defaults carry the risk. Search term matching on by default means the set of queries that trigger your ads can grow beyond what your keyword lists imply. Text customization on by default (ACA path) means the platform may assemble headlines and descriptions you never wrote. Both are labeled "performance-preserving" — and both can quietly move CPCs, conversion rates, and brand messaging if you let them run unmanaged.
The 48–72 hour audit
Before touching any setting, build your evidence file. You can't measure drift without a baseline.
- Inventory affected campaigns — everything using ACA, everything using campaign-level Broad Match. Label them now so post-migration reports are easy to filter.
- Export 90-day baselines — impressions, clicks, CTR, CPC, conversions, CPA/ROAS, and the full search terms report.
- Back up creatives — every headline, description, path field, and pinning configuration.
- Capture targeting — audiences, remarketing lists, location targets, device modifiers.
- Snapshot keywords — keyword lists plus negative keywords at campaign and account level.
A CSV export of campaigns and asset libraries is your restore point. If migration goes sideways, you rebuild from it instead of from memory.
While you're in the search terms report, this is also the moment to mine negatives you already have but never formalized. A quick pass over the export surfaces the repeat offenders:
# Surface low-intent queries worth negating, ranked by spend
grep -iE "free|cheap|jobs|salary|tutorial|what is" search-terms-last-90d.csv \
| awk -F',' '{print $0}' | sort -t',' -k7 -rn | head -40
Your three options before the deadline
| Path | What you do | Best for | Watch out for |
|---|---|---|---|
| Opt in now | Enable AI Max yourself, configure toggles intentionally, test on low-risk campaigns | Teams that want rollout control and preflight data | Needs prep work before flipping it on |
| Opt out | Disable ACA and campaign-level Broad Match before Sept 1 | Brand-exact campaigns, compliance-heavy verticals | May reduce reach if those features drove coverage |
| Auto-upgrade | Do nothing; monitor closely after Sept 1 | Campaigns with strong governance already in place | You accept the defaults sight-unseen |
Opting in early is the only path that gives you timing control, audience allow-listing, and real test data before the platform-wide change. Opting out trades reach for exactness — reasonable for defense campaigns where the precise phrasing of your ad is the whole point.
The migration playbook
If you opt in, run it like a deployment, not a toggle.
1. Pick test campaigns
Choose 1–3 low-risk campaigns with varied objectives — one e-commerce, one lead gen, one brand. You want signal diversity, not your crown jewels.
2. Enable AI Max explicitly
Turn it on deliberately rather than waiting for the auto-upgrade. Confirm the toggles you're accepting: search term matching on, and text customization on if the campaign used ACA.
3. Build the control layer
This is where the migration is won or lost:
- Upload a prioritized asset library of approved headlines and descriptions. Pin critical messaging where the platform allows.
- Add exhaustive negative keyword lists — brand protection, compliance exclusions, irrelevant intent.
- Provide audience signals and first-party data where supported to shape the automated matching.
Give the negatives structure and they scale; dump them in a flat list and they rot:
# negative-keyword-taxonomy.yaml — organized by intent, not alphabet
brand-defense:
- "competitor names"
- "reviews of [competitor]"
- "vs [competitor]"
low-intent:
- free
- cheap
- jobs
- salary
- "what is"
compliance:
- "[regulated terms per vertical]"
geo-mismatch:
- "[regions you don't serve]"
4. Set conservative bidding
Start looser-than-historical on target CPA, or use target ROAS with guardrails. Add a max CPC cap so a matching expansion can't spike spend before your daily review catches it.
5. Review queries daily for two weeks
The search terms report is your early-warning system. Flag unfamiliar queries, negate fast, and track unit economics — not just CTR.
6. Check creative drift
Compare dynamically customized headlines against your brand standards. Deviations get fixed by pinning text or tightening asset controls — not by hoping the model learns your tone.
7. Expand in stages
If CPA/ROAS hold and nothing embarrassing ships, roll to high-value campaigns incrementally.
What to expect in the numbers
- Query breadth: matching expands beyond strict keyword matches — more reach, possibly less relevance per click.
- Creative alignment: text customized to the query can lift CTR on relevant queries — and produce awkward or non-compliant copy on edge cases.
- Measurement noise: an influx of broader queries shifts conversion rates and attribution. Expect fluctuation in CPA and ROAS while behavior settles.
Scenario recommendations
| Scenario | Recommendation |
|---|---|
| E-commerce (low AOV, high volume) | Start in mid-funnel campaigns; tighter CPA targets; aggressive negatives for low-intent queries; watch average order value drift |
| Lead generation (quality-sensitive) | Only after strict audience signals and pinned CTAs; track sales-qualified leads, not raw form fills |
| Brand defense | Keep exact phrasing out of AI Max's hands — disable ACA and campaign-level Broad Match to retain exact-match behavior |
Privacy, compliance, and creative safety
- Verify dynamic text customization can't insert or echo user-submitted PII into ad copy; block PII-shaped patterns via negatives and asset controls.
- Pin mandatory disclaimers and mandated language for regulated verticals — health, finance, legal.
- Audit consent settings before feeding audience signals or customer lists into AI Max.
Rollback plan (write it down before you need it)
- Pause the affected campaign.
- Restore from your exported asset CSV; rebuild with legacy settings — turn off ACA and campaign-level Broad Match before reactivating.
- Reapply conservative bidding and re-check negative keyword coverage.
- Document what broke: example queries, creative screenshots, timestamps. These artifacts are how you escalate with platform support or recreate a pre-migration state.
Monitoring timeline
| Window | Focus |
|---|---|
| Days 1–7 | Spend spikes, unusual queries, creative mismatches — daily review |
| Weeks 2–4 | Stabilization; keep adding negatives, adjust bids — review a few times a week |
| Weeks 4–12 | Long-term ROAS and funnel impact; confirm attribution settings still fit |
Do these things now
- Identify every campaign using ACA or campaign-level Broad Match.
- Export settings, creatives, and 90-day performance baselines.
- Decide your path: staged opt-in, opt-out, or monitored auto-upgrade.
- If opting in: conservative bids, strong negatives, pinned assets, daily query reviews for 14 days.
- Keep the rollback plan in the same folder as your exports — not in your head.
AI Max is a platform-level change, not a feature toggle. Matching, messaging, and measurement all move together. Negatives, pinned assets, and bid caps are the levers you still own — use them, and September 1 becomes just another rollout you ran on purpose.