Google has confirmed it is testing AI-generated descriptions on Shopping ads. This follows a similar test on standard sponsored search results seen earlier in the same month - so this is not an isolated experiment. There is a clear direction here, and it is accelerating.
The question for advertisers running Shopping campaigns is not whether this is good or bad. The question is: if Google is going to generate the description, what inputs is it drawing from, and how do you influence them? Because the answer to that determines whether AI-generated copy helps or undermines your campaign.
What Google Is Actually Pulling From
Google does not generate ad copy from thin air. For Shopping ads, the most logical source material is the product feed - specifically the fields you submit to Merchant Centre: product title, description, product type, brand, and any supplemental attributes. If your feed is sparse or generic, the AI has sparse, generic material to work with.
It will also draw from the landing page. Google's crawlers index product pages, and that content informs how products are categorised and described across organic and paid surfaces. A product page that leads with boilerplate copy or manufacturer descriptions gives the AI nothing distinctive to surface.
This is not a new dynamic - Performance Max has been generating ad copy from asset groups and landing pages for some time. What changes with Shopping ad descriptions is that the output appears directly alongside the product image and price, where it influences click-through decisions at the point of purchase intent. The stakes are higher than in a peripheral asset group headline.
The Feed Quality Problem This Exposes
Most product feeds are built for data compliance, not persuasion. Title optimisation gets some attention - advertisers know that keyword-rich titles affect Shopping ranking. But description fields are routinely neglected. Many are simply copied from the product page meta description, left at default, or filled with technical specifications that read like a datasheet rather than ad copy.
If AI-generated descriptions are derived partly from feed data, those neglected description fields now have a direct bearing on what appears in the ad unit. A description field that reads "100% polyester, machine washable, available in 3 colours" gives the AI very little to differentiate your product from the ten competitors above and below it.
Practically, this means feed audits should now include a review of description quality, not just title structure, GTIN compliance and image specifications. Are your descriptions written in a way that communicates a clear reason to buy? Do they include the product benefit, not just the specification? These are questions worth answering before Google answers them for you.
Control Is Narrowing - but Not Disappearing
There is a reasonable concern that AI-generated descriptions reduce advertiser control over messaging. That concern is valid. But the more useful framing is that the control is shifting - from direct copy input to upstream influence over the source material the AI draws from.
Advertisers running standard Shopping campaigns have always had limited copy control compared to search. You cannot write a Shopping ad headline the way you write a responsive search ad. The product title and price do most of the work. Descriptions, where they appear, are secondary. AI-generated descriptions extend this dynamic rather than fundamentally changing it.
Where this becomes more significant is for advertisers with brand-specific messaging requirements - legal disclaimers, regulated claims, specific positioning statements. If AI-generated descriptions can contradict or omit those requirements, that is a compliance issue that needs to go beyond the marketing team. Monitoring what descriptions are actually being generated should be part of any Shopping campaign review, particularly for categories like financial services, health products or regulated industries.
What This Means for Shopping Campaign Structure
For accounts running Performance Max with product feeds, AI copy generation is already the default. Product titles, descriptions and landing page content inform asset generation across all placements. This Shopping ad description test is consistent with that direction - Google is consolidating AI copy generation across more of its ad surfaces.
If you are running standard Shopping campaigns alongside PMax, keep monitoring the interaction between the two. Standard Shopping gives you more visibility into search terms and placement data. But if AI-generated descriptions start appearing more widely in standard Shopping, the distinction between managing a feed and managing ad copy becomes even blurrier. The feed is the creative brief.
One practical response is to treat your Merchant Centre supplemental feed as a creative asset, not just a data correction layer. Use it to push richer, more persuasive description content for your highest-priority products - particularly those with strong margins or high conversion rates where the quality of the description is more likely to influence the outcome.
What to Do Before This Rolls Out Widely
This is still a small test. Google described the equivalent sponsored search test in similar terms before it expanded. You have a window to put your product data in better shape before AI-generated descriptions become standard rather than experimental.
Start with your top 20 products by spend or revenue. Review the description fields in your Merchant Centre feed. Rewrite any that are purely technical or that read like internal catalogue copy. Make sure your product landing pages open with clear, benefit-led copy that a crawler can read and extract meaning from. These are not dramatic changes - but they directly shape the inputs the AI will use.
Also set up monitoring now. If you are not already pulling Shopping ad preview data regularly, start. When AI-generated descriptions appear on your products, you want to know what they say before your customers do - not after a campaign review two months later. The advertisers who will manage this well are the ones who treat feed quality as an ongoing discipline, not a one-time setup task.