When you upload product titles to Merchant Center, you assume that's what shows up in your Shopping ads. That assumption has been wrong for a while. Google's AI has been generating its own product titles when it decides yours aren't relevant enough for a given query. What's new as of August 26 is you can finally see it happening.
A new Product titles report is rolling out inside Google Ads, first reported by Search Engine Land. It shows which titles Google AI generated versus which ones came from your feed. Until now, that substitution was invisible.
Google has been doing this quietly
The product title in your Shopping ad isn't always yours. If Google's model determines a generated title would perform better against a search query, it swaps it in. Your original title stays in Merchant Center — it can still be served when Google judges it more relevant — but the AI version may be what shoppers actually see.
This creates problems beyond ad copy preference. Product titles carry brand name, model number, size, and variant details. If Google's model strips out a model number and replaces it with a more query-friendly generic phrase, you lose specificity. A shopper who sees a vague title, clicks, and lands on a product page with specs they weren't anticipating has a worse experience and a lower chance of buying.
Brands with regulated language have a separate problem. Supplements, cosmetics, and financial products often require specific qualifier language in titles. Google's AI doesn't know about your compliance requirements. A rewritten title that drops a mandated phrase creates legal exposure that won't show up anywhere in your ads reporting.
What the report actually shows
The Product titles report appears in Google Ads alongside existing Shopping reporting. It splits performance by AI-generated titles and your original feed titles, giving you a direct comparison of which version Google served and how each performed.
Four things to check when you pull it:
- AI-generated titles outperforming your originals tell you the feed titles are weak for those queries — close the gap in Merchant Center, don't argue with the result
- AI-generated titles underperforming your originals signal the model got it wrong — document these, especially on high-revenue SKUs, and take them to support
- Products where AI generates titles at high volume have thin original titles and the model is improvising across a wide query range
- Variant SKUs where AI is generating titles are most at risk of losing model numbers, sizes, and spec details that drive conversion
The report gives you comparison data that didn't exist before. If Google's version is winning on a product category, that's useful signal about what the algorithm values for those queries. You can close that gap in your feed faster than Google can keep improvising.
The feed quality problem this reveals
AI title generation doesn't happen randomly. It happens most often when your original titles don't match how users are searching. "Blue Sneakers - Men's" gives the model a lot of room to intervene. "Nike Air Max 270 React ENG Men's Lifestyle Shoes Blue Void/Black Size 11" gives it almost none.
This is the same dynamic that drives AI Max performance: specific attributes in the feed reduce the surface area where AI has to improvise. Thin titles, generic descriptors, and missing attributes all expand the space where Google generates its own copy — copy you may not agree with and definitely didn't approve.
The timing of this report matters. Starting September 1, Google begins mandatory migration of campaigns using automatically created assets and campaign-level broad match settings into AI Max. AI Max expands the scope of what Google can generate and serve across the entire campaign. The product title replacement you're now seeing in Shopping is one piece of a larger shift in who controls what the ad says.
What to do before September
Pull the Product titles report this week. If it's not visible in your account yet, it's still rolling out — check again in a few days.
Where AI-generated titles are winning, go back to the Merchant Center entry and ask what's missing. Add the specific descriptor, the variant detail, the attribute that makes the title query-relevant. Make the original specific enough that the model has no reason to substitute.
Where AI-generated titles are losing — especially on brand-sensitive or compliance-regulated products — note the exact titles Google generated. Reach out to Google Ads support with specific examples. There's no formal opt-out for title generation, but documented patterns of harmful substitutions on regulated SKUs are the basis for escalations.
Treat the report as a feed quality audit, not a creative dispute with Google's AI. The same SKUs generating poor replacement titles are the ones that will have the most exposure when AI Max expands its generation scope in September.
Your feed is the brief Google's AI works from. The better the brief, the less the AI needs to fill in.
The free Gromerce audit shows feed and signal quality gaps in under three minutes — worth running before the September migration window opens.
The Product titles report doesn't change what Google has been doing. It just means you can see it now.
Sources: Search Engine Land, August 2026
