Google Ads AI Max Ad Copy Test: Results From 3 Companies

Free Instant SEO Audit

By NewSunSEO July 30, 2026

Three companies tested Google Ads AI Max’s text customization feature across ecommerce, B2B lead generation, and B2C lead generation campaigns. The tests measured how well AI-generated ad assets performed against human-written copy on both highly optimized and long-tail campaigns. Results varied sharply by business model: AI assets helped lower-priority campaigns but hurt performance on carefully tuned ones, and at least one company turned the feature off after three weeks.

What text customization does in AI Max

Text customization is one of AI Max’s asset-generation capabilities. When turned on, it tailors ad copy in every ad group to the keywords in that group, which can take the strain off a PPC team by removing the need to customize ads for each ad group manually.

Because the system works in the background, it can sometimes generate assets that promote products or services the advertiser does not offer. To guide it, advertisers should apply messaging restrictions to auto-created assets, including brand and tone rules, so the output stays within approved language.

How to set up messaging restrictions

The setup process takes about an hour or two and follows a back-and-forth approach:

Use a prompt in Gemini to draft initial assets.
Prompt the system to make the ads overly promotional, including promises you would not approve.
Create messaging restrictions to block the unwanted styles.
Re-prompt until every generated asset aligns with company messaging guidelines.

Campaign selection criteria for the test

The tests covered three business types: ecommerce, B2B lead generation, and B2C lead generation. Within each account, the team selected campaigns that met several criteria:

No brand keywords.
Monthly spend of at least $20,000.
At least 100 ad groups.

Two campaigns per account were highly watched and frequently optimized. Two were neglected long-tail campaigns that performed well in aggregate but received less attention. To isolate the asset test, the team chose campaigns that did not rely heavily on pinning and excluded URL expansion, since most large accounts use pinning extensively in top campaigns.

How much auto-created asset review is required

Reviewers must change the default filter to include the ad itself, since AI-generated assets are not shown by default. Across the ecommerce and B2C accounts, roughly 19% of auto-created assets were removed because they did not align with brand messaging or offers. The companies monitored assets as they were created and pulled them before they accumulated many impressions.

Ecommerce campaign results

The ecommerce company sells more than 100,000 SKUs, and many shoppers revisit the site and search again if a landing page misses their target product. Early numbers made AI Max and text customization look highly successful. Deeper analysis showed AI Max was poaching impressions, clicks, and conversions from other campaigns, and total account revenue declined.

The team added many search terms as keywords so Google could prioritize the right ad group and campaign, then added negative keywords and audience lists to slow cannibalization. After rerunning the tests, AI text customization was not as effective as human asset management on highly optimized campaigns, but it did help the long-tail campaign.

B2B lead generation results

For B2B advertisers, RSA assets need to prequalify the audience, making ads unattractive to consumer searchers while appealing to business buyers. This company had used pinning extensively across assets to enforce that qualification, and it removed pins for the test to see how well Google could optimize.

Text customization did not understand the nuance of prequalifying a B2B audience. Click-through rates rose sharply, but conversion rates fell because the ads pulled in many B2C searchers. Messaging restrictions included B2B-prequalifying language, and some assets met it, but the overall ads shown to users did not properly appeal to business buyers.

After three weeks of poor results, the B2B company stopped the test, reverted to pinning, and removed auto-created assets. Performance returned to pretest levels within a week.

B2C lead generation results

The B2C lead generation company localizes ads through geographic ad copy and location insertion. Its optimized campaigns had tailored copy for the keywords in every ad group, while the long-tail campaign reused most assets across groups, with only a few keyword-tailored headlines per ad group. That pattern of formulaic copy in lower-priority campaigns is common, and it represented a clear opportunity for AI-generated assets.

AI Max’s auto-created assets delivered in those low-priority campaigns. They did not outperform the human-tested copy in the top campaigns, but they performed well in the long-tail campaign where opportunity was greatest.

Where AI Max automated assets work best

Across all three tests, AI was useful for brainstorming and drafting, but human-created assets still outperformed AI-generated ones on carefully managed campaigns. Advertisers who need specific assets, such as B2B prequalifying language, particular offers, or short-term promotions, should keep control of the assets and avoid handing them to AI.

When there is no time to fully optimize creatives, auto-created assets can help. AI can improve results on campaigns that receive little management when messaging is heavily restricted and reviewed often. AI requires human oversight and is not a set-and-forget setting. The most effective approach was using AI for heavy lifting in areas the team could not fully optimize, then spending human time reviewing and adjusting the output.

FAQ

What is text customization in Google Ads AI Max?

Text customization is an AI Max feature that automatically generates ad assets tailored to the keywords in each ad group. It can reduce manual work for PPC teams, but the auto-created assets may include promotions or offers a company does not actually provide, so messaging restrictions and regular review are required.

How did AI Max perform on ecommerce campaigns in the test?

Initial results looked strong, but further analysis showed AI Max was poaching impressions, clicks, and conversions from other campaigns and overall revenue declined. After adding keywords, negative keywords, and audience lists to slow cannibalization, the team found AI text customization was less effective than human management on highly optimized ecommerce campaigns, though it helped a long-tail campaign.

Why did the B2B company stop using AI Max text customization?

The B2B account’s click-through rates spiked but conversion rates fell sharply because the AI-generated ads attracted many B2C searchers rather than prequalifying business buyers. Messaging restrictions alone were not enough to maintain B2B qualification. After three weeks of poor results, the company reverted to pinning its ads and removing auto-created assets, and performance returned to pretest levels within a week.