Google Ads tests AI-generated product titles for Shopping ads in beta

Google is testing a significant change to how Shopping ads can be presented: AI-generated product title variations that may be served instead of the title submitted in a merchant’s product feed.

The feature, described by Google as product title optimization within the AI Max for Shopping beta, does not replace the original feed title. Google says the submitted title remains the baseline, while its systems may generate and serve an alternative when they predict that version will perform better for a particular search.

That distinction matters. This is not simply another feed-editing tool. It is a serving-layer optimization: the product data an advertiser submits remains foundational, but the ad copy shown to an individual user may be adapted dynamically.

A new level of visibility into Shopping ad messaging

Google’s Product Titles reporting experience is particularly noteworthy because it gives advertisers a way to see both the original feed title and Google’s customized variation. The report can also show delivery and performance data associated with each title version, including core metrics such as impressions, clicks, CTR, cost, CPC, and selected conversion metrics.

For Shopping advertisers, this is a meaningful development. Historically, product titles have been one of the most consequential levers in feed optimization, yet advertisers have had limited visibility when Google’s automation used product data in unexpected ways. The new report does not eliminate that opacity, but it creates a much more useful audit trail.

The central question is not whether an AI-generated title looks more polished than the original. It is whether it earns incremental qualified traffic or conversion value without weakening product accuracy, brand positioning, or margin efficiency.

Google explicitly states that a customized title should serve only when its systems predict better performance than the original title. However, Google does not publicly detail the prediction model, the exact success metric being optimized, the minimum evidence required before a variation serves, or how the system balances click-through rate against conversion outcomes. Those unanswered questions are important.

A title that produces more clicks is not necessarily commercially better. In some categories, making a product title broader may increase query coverage and CTR while attracting less qualified traffic. The Product Titles report should therefore be evaluated through conversion and value-based metrics, not click metrics alone.

Product title optimization is not a substitute for feed quality

The beta reinforces a familiar principle: automation is only as strong as the underlying data available to it.

Google can potentially emphasize different attributes depending on search intent, such as material, color, compatibility, size, brand, or use case. But that does not reduce the need for accurate, complete, and structured product data. If a feed title is vague, key attributes are missing, or landing-page information is inconsistent, AI customization may have limited useful material to work with—or may surface messaging an advertiser would not have prioritized.

The original title remains strategically important because it is the submitted source data, the primary reference point in reporting, and the version Google compares against. Retailers should not interpret title optimization as permission to relax feed governance.

This is especially relevant for products with regulated claims, technical specifications, variant complexity, bundle configurations, or strict brand-language requirements. Better predicted ad performance does not automatically mean that every possible wording variation is appropriate for every business.

Controls deserve close attention

Google positions text customization as a configurable component of AI Max for Shopping. Separately, Google states that advertisers can opt out of Product Data Customization for Shopping ad title sthrough their Google account team or Google Ads support.

The terminology is worth treating carefully. “AI Max for Shopping,” “product title optimization,” and “Product Data Customization” are closely related in Google’s product communications, but advertisers should not assume that similarly named settings always have identical scope, eligibility, or operational behavior. Google’s public materials do not fully explain every relationship between these controls, particularly across beta access, campaign types, and account configurations.

That is not merely a documentation detail. Before enabling or assessing the feature, advertisers should establish which control applies to their account, whether it affects title generation specifically, and what happens to already-serving variations after an opt-out request.

The broader Shopping automation shift

Google has said product title optimization is intended for both Performance Max and Shopping campaigns. At the time of the beta announcement, however, advertisers should distinguish between Google’s stated direction and feature availability in their own accounts. Public announcements do not necessarily define universal eligibility, market coverage, or rollout timing.

Strategically, this development fits Google Ads’ wider movement toward query-level creative and product-data adaptation. In Search, Performance Max, and Shopping, advertisers increasingly provide the inputs and guardrails while Google determines the final combination most likely to perform in a specific auction.

The Product Titles report is therefore more than a reporting addition. It is a practical mechanism for judging whether that trade-off is working. If customized titles generate stronger conversion efficiency while preserving product precision, they may become a valuable incremental layer. If they mainly shift impression volume or increase low-intent clicks, the report should make that visible too.

Google’s official materials confirm the beta, the comparison between original and customized titles, and Google’s prediction-based serving logic. What remains less clear is how those predictions are calculated and how consistently controls apply across accounts. For experienced Shopping advertisers, that uncertainty makes measurement and governance more important—not less.

‍