Google launched AI Max for Search campaigns as the biggest update to Search Ads in years. The pitch: one toggle that gives Google's AI permission to expand match types, write headlines, choose landing pages, and basically rewrite your campaign on the fly.
Marketers had two reactions. The optimists called it "the future of Search." The pessimists called it "Performance Max wearing a Search costume."
I've spent the last 90 days running AI Max alongside controlled non-AI Max campaigns on real budgets, across three different industries. The actual results are more nuanced than either camp wants to admit.
Here's what I learned, what worked, and where I'd never enable it again.
The TL;DR: AI Max isn't magic and it isn't a disaster. It works brilliantly for accounts with clean conversion tracking and lots of buying intent in long-tail queries. It quietly destroys accounts with weak conversion signals, vague landing pages, or brand-protection concerns. The toggle isn't the strategy — your conversion stream is.
What AI Max actually is (in plain language)
AI Max bundles four things into a single Search campaign setting:
- Keywordless matching : Google's AI matches your ads to queries beyond your keyword list, using your landing page, ad copy, and other signals.
- Automated asset creation : Google generates new headlines and descriptions based on your existing assets and landing pages.
- Final URL expansion : The campaign can send traffic to different URLs on your site beyond what you specified.
- Text customisation : Google rewrites ad text per query to match search intent better.
You can enable some pieces independently, but the headline feature is the all-in toggle.
AI Max is essentially "give Google's AI broader permission across your Search campaign." Whether that's good depends entirely on what signal you're feeding it.
What I tested
Three accounts. Different industries. 90 days. Same overall budget.
- Travel / ticketing account : mature conversion tracking, profit feed in place.
- Healthcare lead-gen account : Cliniko-tracked bookings, value-adjusted for cancellations.
- B2B SaaS account : MQL → SQL → opportunity offline conversion pipeline.
For each, I duplicated the top-performing Search campaign, enabled AI Max on one, kept the original untouched, and split budget 50/50.
What worked
1. Long-tail capture in the travel account
The AI Max campaign captured queries the original keyword list never would have.
The travel account had 3 years of clean conversion data and a tight profit feed. AI Max had real signal to optimise against. ROAS came in 11% higher than the control campaign.
2. Headline freshness without manual A/B testing
The auto-generated headlines outperformed our best human-written headlines in 4 out of 7 cases. Not always — but often enough to be worth keeping enabled.
3. Lower CPC on returning visitors
In the healthcare account, AI Max consistently bid lower on returning visitors who had already filled the booking form once. The control campaign treated them identically. Net effect: ~8% lower CPC across the AI Max half.
Where it broke
1. The B2B SaaS account
Conversion lag in B2B SaaS is brutal. MQL fires within minutes. SQL takes 7-14 days. Closed-won takes 30-90 days. AI Max's learning was working off MQL signal almost exclusively.
Within six weeks, AI Max was pouring spend into broad keywordless matches that generated MQLs at 35% lower CPL — but those MQLs converted to SQL at half the rate of the control campaign. Net CAC was 28% higher.
Lesson: AI Max needs strong conversion signal close to the click. If your real conversion is weeks away, the algorithm chases shadows.
2. Brand bleed
AI Max matched our generic campaigns to branded queries the brand campaign was already running. Net effect: bid inflation in the brand auction, more spend, same conversions.
The fix: account-level negative keywords for our own brand terms on the AI Max campaign. Should have been there from day one.
3. Auto-generated copy that misrepresented the offer
Twice in 90 days, the auto-generated headlines made claims our landing page didn't support ("Same-day appointments available" when the practice only offered same-day for two services). One led to a polite legal email from compliance.
You must monitor the asset library weekly when AI Max is on. Disabling specific auto-generated assets is one of the most important workflows nobody talks about.
4. Final URL expansion sending traffic to the wrong page
In the healthcare account, AI Max started sending paid traffic to a blog post about back pain instead of the booking page. CTR was great. Conversion rate was zero. The blog had no booking CTA above the fold.
Either lock down final URL expansion to specific URLs, or audit landing pages weekly.
Account profiles where AI Max actually works
Based on what worked across the three accounts, AI Max is a good fit when all of these are true:
- You have clean, fast conversion tracking (conversion fires within hours of click, not weeks).
- Your conversion value reflects real business outcomes (profit feed, value adjustments for refunds/cancellations).
- Your landing pages are conversion-focused with clear CTAs above the fold.
- You have enough budget for the algorithm to learn (50+ conversions per month per campaign minimum).
- Brand protection is either not relevant or handled with account-level negative keywords.
- Your team has bandwidth to audit asset libraries at least weekly.
Account profiles where I'd never enable it
- B2B SaaS with long sales cycles and no offline conversion uploads to Google Ads.
- Accounts with noisy or inflated conversion signal (lots of micro-conversions counted as primary conversions).
- Regulated industries (legal, financial advice, medical) where auto-generated copy creates compliance risk.
- Brand-sensitive accounts where the wrong headline could damage trust.
- Accounts with broad, ambiguous landing pages (homepages instead of dedicated landers).
How to actually set it up if you're going to try
- Pick a control. Duplicate your best-performing Search campaign before enabling AI Max on either copy. You need a fair comparison.
- Fix conversion tracking first. AI Max is only as good as the signal you give it. Profit feed, value adjustments, offline imports — get them done first.
- Add account-level negatives for your brand terms and any competitor terms you don't want appearing in.
- Lock down final URLs to specific high-converting pages until you've earned trust in the AI's choices.
- Set up alerts for new auto-generated assets. Review them weekly. Disable anything that misrepresents.
- Give it 4-6 weeks before judging. AI Max needs a full learning period.
- Compare against the control on a real business metric — profit, real conversions after refunds, SQL-stage conversions for B2B. Not CPC.
Use case: an inbound travel agency selling Dubai experiences
A composite based on patterns from a few accounts.
A mid-sized travel agency was running 12 highly structured Search campaigns. Each campaign mapped to an attraction category — water parks, theme parks, museums, etc. Tight keyword lists, manual bid adjustments, the works. Spent ~3 hours a day on optimisation.
We tested AI Max on three campaigns. Same structure, same budget, AI Max enabled.
Weeks 1-3: AI Max underperformed the controls by 15%. Team was tempted to disable.
Weeks 4-6: Performance crossed over. AI Max started capturing long-tail queries the original keyword lists missed. ROAS climbed.
Weeks 7-12: AI Max campaigns finished 11% ahead on ROAS, with 22% less manual maintenance.
But: AI Max also captured brand traffic from sister-brand campaigns, double-charging the account. Once we added account-level brand negatives, the overlap stopped and net ROAS improved further.
The headline lesson: AI Max didn't replace the marketer. It absorbed the keyword-list work and freed up time. The strategic decisions (negatives, budget allocation, conversion tracking) still needed a human.
Common mistakes when rolling it out
- Enabling AI Max on every campaign at once. Test on one. Then expand.
- Disabling it after two weeks. Learning takes a full 4-6 weeks. Most people pull the plug too early.
- Skipping account-level negatives. Brand bleed will silently destroy your numbers.
- Letting auto-generated assets run without review. Compliance, brand, accuracy issues compound.
- Treating final URL expansion as harmless. It isn't. Lock it down or audit weekly.
- Comparing CPC instead of ROAS. AI Max often has higher CPC and better ROAS. CPC alone tells you nothing.
Bottom line
AI Max is genuinely useful for the right account. It's a disaster for the wrong one. The variable isn't Google's algorithm — it's the quality of the signal you give it.
If you have clean conversion tracking, a profit feed, conversion-focused landing pages, and bandwidth to monitor auto-generated assets, enable AI Max on one campaign and run it against a control for 6 weeks. Decide from real data, not from LinkedIn opinions.
If you have noisy conversion signal, long sales cycles, or compliance constraints, don't touch it yet. Fix the foundation first.
The toggle isn't the strategy. It never was.
Sources and further reading: