Guide · Google Ads automation
AI Max widens which searches Google answers. It doesn’t change what Google thinks a good lead is. Turn it on once the conversion you bid to reflects lead quality, and judge it on leads sales has had time to review.
The basics
Short answer
AI Max is a set of features inside a normal Search campaign, not a new campaign type. It widens the searches your ads can match, writes extra headlines and descriptions from your site, and can send clicks to other pages on your domain. Google says it adds conversions at a similar CPA. B2B teams should check qualified leads.
Google describes AI Max as a layer you switch on inside your existing Search campaigns, not a new campaign type (Google Ads Help). It has three parts:
When you turn AI Max on, text customization and final URL expansion are on by default, and final URL expansion only works while text customization is on. The features require conversion-based Smart Bidding (set up AI Max), and Google warns that AI Max won’t be effective if the campaign is limited by budget.
Google’s headline claim is that advertisers who activate AI Max “will typically see 14% more conversions or conversion value at a similar CPA/ROAS”, based on Google internal data from 2025 for non-retail advertisers (Google Ads Help). It counts conversions, not qualified leads.
This month
Short answer
Possibly. Google says Search campaigns using automatically created assets or the campaign-level broad match setting are auto-upgraded to AI Max starting in September 2026. Dynamic Search Ads campaigns follow from February 2027, after Google extended that timeline. Open each campaign’s AI Max settings to confirm, then review search terms, landing pages and generated assets.
Sources: Google’s AI Max upgrade announcement, updated 11 June 2026, and About Dynamic Search Ads. For Dynamic Search Ads campaigns, Google says dynamic ad groups move to standard ad groups, with settings and data carried into AI Max (Google announcement).
In Google Ads, open Campaigns › Settings, select the campaign, then choose AI Max in the section menu.
Log the upgrade date: the Google Ads audit checklist starts with changes like this.
Original framework
The three switches carry different risks for a form that feeds a sales team. Treat them as three decisions, not one toggle.
| Switch | What it changes | B2B failure mode | Control | Starting position |
|---|---|---|---|---|
| Search term matching | Matches searches beyond your keywords with broad match and keywordless technology | Job, training, student, retail and “free” searches that still fill in forms | Account-level and campaign negatives, brand settings, locations of interest; can be turned off per ad group | Test on |
| Text customization | Writes headlines and descriptions from your ads, landing pages and assets | Promises you can’t keep: lead times, certifications, a “free quote” where quotes are chargeable, regulated advice | Review assets marked Google AI; text guidelines where your account has them | On where claims are low-risk; reviewed first for regulated professional services |
| Final URL expansion | Sends the click to the page Google predicts will perform best | RFQ or demo intent landing on blog, careers or docs pages; 404s from incompatible tracking templates | URL exclusions, the landing pages report, a tracking template check | Off at first, or on with exclusions |
This is where most B2B drift comes from. An industrial keyword such as “hydraulic cylinder manufacturer” can attract searches from students, job seekers and people repairing one machine. Put the searches that can never buy on an account-level list before you start; the B2B negative keywords guide has a starting library. Keep exact match keywords for part numbers and specifications: an exact keyword identical to the search still takes priority.
Google offers text guidelines, an experimental beta for Performance Max and for Search campaigns with AI Max, with up to 25 term exclusions and 40 messaging restrictions per campaign (Google Ads Help). Where you don’t have them, review generated assets weekly.
Google sends people to “the landing page from your domain that is predicted to perform best”. For a lead form, the best-performing page by Google’s measure may be the one with the easiest form, not the one that qualifies the buyer. Exclude careers, jobs, blog, docs, investor, support and login paths, and confirm your tracking template uses a standard pattern such as {lpurl}? so expanded URLs don’t break (Google Ads Help).
Decision
Short answer
Only when Google bids towards a lead-quality signal, the campaign isn’t limited by budget, negatives and brand exclusions are in place and forms filter spam. Without those, wider matching finds more form-fillers, not more buyers. If the campaign produces few qualified leads a month, waiting is usually the better decision.
AI Max needs conversion-based Smart Bidding, so it chases whatever you count. Bid on a qualified-lead import, or at least observe one as a secondary action. See which lead stage to optimise for and offline conversion import.
Google says AI Max won’t be effective if it is. Sizing is in how much a B2B company should spend on Google Ads.
An account-level list for searches that can never become buyers, and brand exclusions for brands you don’t want to appear on.
Qualifying fields, reCAPTCHA or server-side validation, and spam removed before anything is imported. See improving Google Ads lead quality.
Google publishes no figure for experiments. Our rule of thumb: each arm should see at least 15 qualified leads over the readout period. Below that, wait or test longer.
If check 1 or 5 fails, turning AI Max on doesn’t give you a test, it gives you noise. Fix the signal first. Zephra’s Act, Review, Investigate and Wait decisions are explained in how it works.
Test design
Short answer
Run it as a Google Ads experiment on your largest Search campaign, with auto-apply off. Wait until leads from the test have had time to be qualified by sales, then compare cost per qualified lead between arms, not cost per form fill. Review AI Max search terms and Google-selected landing pages weekly.
An AI Max experiment splits one existing Search campaign: the control share runs with AI Max off, the trial share with it on. Setup offers an option to “Apply your experiment changes if results are favorable”, and if the experiment ends without being applied, AI Max settings revert to their pre-experiment state (Google Ads Help).
Leave auto-apply off in B2B. “Favorable” is judged on the campaign’s conversions. If those are demo requests or RFQs, the experiment can be applied on the metric sales doesn’t trust.
Pick the campaign with the most qualified leads, agree a lag window (typical days from lead to qualification, plus a buffer), keep bids, targets and budget stable, add negatives and URL exclusions to both arms, and read the result only after the lag window has passed for the last week of the test.
| Section | Record |
|---|---|
| Set-up | Start and end dates, split, lag window, readout date, auto-apply off, the decision rule |
| Each arm | Spend, leads, qualified leads, cost per lead, qualified rate, cost per qualified lead |
| AI Max share | Share of spend on the AI Max match type; share of leads that landed on Google-selected URLs |
| Decision | Apply, extend with new exclusions, or end and let settings revert |
Reporting
Short answer
Filter the search terms report by the AI Max match type, check the two AI Max total rows in the keywords report, open the landing pages report for URLs Google chose, and review assets marked Google AI. Segment each by your imported qualified-lead conversion, so you see which matches produced leads sales accepted.
Source: AI Max reporting. To see which terms produced qualified leads, segment by conversion action with a qualified-lead import in place, and pass keyword and match type into hidden form fields for your CRM. The method is in cost per qualified lead.
Worked numbers
A US B2B SaaS company pays $12 a click and converts 4% of landing page visitors, so a demo request costs $300. Sales qualifies 35%. Its main Search campaign spends $6,000 a week, and it runs a six-week AI Max experiment split 50/50, then waits three weeks for qualification before reading it.
| Arm | Spend | Demo requests | $ per request | Qualified | Qualified rate | $ per qualified demo |
|---|---|---|---|---|---|---|
| Control (AI Max off) | $18,000 | 60 | $300 | 21 | 35% | about $857 |
| Trial (AI Max on) | $18,000 | 72 | $250 | 20 | about 28% | $900 |
72 requests against 60, at $250 instead of $300. With demo requests as the campaign’s conversion, auto-apply could have rolled AI Max out.
20 qualified against 21, at $900 each against about $857. The extra requests came with a lower qualified rate.
The next step isn’t “AI Max fails”. Open the AI Max search terms and the landing pages report for the trial arm. If the unqualified requests cluster in fixable groups, such as tutorial and “free” searches or blog landing pages, add negatives and URL exclusions and extend the test. If they are spread thinly across many searches, end the experiment and let the settings revert. Both arms cleared the 15-qualified rule of thumb, so the result is readable; after two weeks it would not have been.
Comparison
Short answer
AI Max stays inside an existing Search campaign, keeping keywords, ad groups and search terms reports. Performance Max is a separate campaign type that spends across Search, Display, YouTube, Discover, Maps and Gmail from one budget. For B2B lead generation, test AI Max first, and add Performance Max only when a qualified-lead signal is in place.
| Feature | AI Max for Search | Performance Max |
|---|---|---|
| What it is | Settings inside a Search campaign | A separate campaign type |
| Inventory | Search | Search, Display, YouTube, Discover, Maps, Gmail and search partners |
| Query inputs | Keywords, match types, landing pages and assets | Search themes, up to 50 per asset group |
| Serving priority | Treated like phrase and broad match: shares priority with PMax search themes | Gives way to a Search exact match keyword identical to the query |
| Lead quality risk in B2B | Query drift and landing page drift | Form-fillers from non-search inventory, where negatives don’t apply |
Sources: keyword prioritisation, channel performance report, search themes. The full comparison, including when PMax earns a budget, is in Search vs Performance Max.
Context
| Size | Typical approach | Watch for |
|---|---|---|
| SMB | One campaign test at most; often wait until a qualified signal exists | Final URL expansion off; a small site has few safe landing pages |
| Mid-market | Product-line campaigns; experiment on the largest | RevOps confirms the qualification lag before the readout date |
| Enterprise | Several accounts and markets; change control for upgrades | Brand and legal review of generated text; URL exclusions per market site |
Wider matching can pull consumer “price” and retail searches in alongside genuine “price list” searches from business buyers. Judge those by outcome rather than negating them on sight. Hinglish and transliterated searches need their own review. Give notice under India’s DPDP Act before lead data is used for the qualified-lead import; see lead generation compliance.
Company set-up, free zone and visa searches are common drift for services firms. Google says that from September 2026 Search campaigns prioritise ads that match the language of the search term (Google Ads Help), so check how Arabic-language searches appear in your AI Max search terms if you advertise in English only.
Competitor and “alternative” searches surface through keywordless matching; decide on brand settings deliberately. Legal and financial advisory firms should review generated text before relying on it. Check consent rules before call or text follow-up.
Zephra creates and manages Google Search campaigns, keywords, creative and landing pages, and supports negative keywords, search-term promotion and waste detection as levers, reading them against lead-quality evidence from your CRM or a spreadsheet, where configured. Changes follow the permissions and approvals you set, and every change is recorded in an audit log. See Google Ads for B2B.
Questions, answered
Google’s set-up guide describes a toggle to opt out of all AI Max settings in a campaign’s AI Max settings, and search term matching can be switched off per ad group. Experiments that end without being applied revert automatically. For campaigns Google upgraded, confirm the options shown in your account.
Google says the features require conversion-based Smart Bidding, not a specific strategy. The bigger question for B2B is which conversion the bidding optimises to; see which lead stage to optimise for.
Start with it off, or on with URL exclusions for careers, blog, docs, support and login pages. Check the landing pages report weekly, and confirm your tracking template works with dynamic landing pages to avoid 404 errors.
Google doesn’t set a duration for AI Max experiments. In B2B, run it long enough for each arm to collect readable qualified leads after the qualification lag, often six weeks or more. Our rule of thumb is at least 15 qualified leads per arm.
Zephra creates and manages Google campaigns and supports keyword, negative keyword and waste detection levers, where configured. Low-risk changes can run automatically; everything else follows the permissions you grant or waits for approval, and every change is recorded in an audit log.
Your account
The free ad account assessment takes 10–15 minutes and reviews wasted spend, campaign structure, performance constraints and measurement gaps. No CRM needed, no campaign changes during the assessment.
