AI summary
In 2026, Performance Max builds its listing groups from product feed attributes rather than keywords, Meta Advantage+ Catalog automatically assembles its ad creatives from the same catalog, and conversational agents like ChatGPT rely heavily on data already indexed by Google Shopping: according to a Peec AI study from March 2026, 83% of products present in ChatGPT shopping carousels already appear in the top 40 of Google Shopping. A well-structured product feed therefore benefits simultaneously to Shopping, Meta and visibility in AI.
A year ago, optimizing your product feed mainly meant: passing Merchant Center checks, filling in mandatory fields, and letting bids do the rest. That’s no longer quite true. Performance Max changed how Google targets, Meta Advantage+ Catalog brought its logic closer to Google’s, and a new category of players — conversational agents — now reads the same catalog to build its own recommendations. Here’s what concretely changed, and what it means for product data.
1. Performance Max changed the nature of targeting
On classic Search, the advertiser chooses a keyword and writes an ad. On Performance Max, these two decisions no longer exist in the same way: according to Google Ads documentation on managing Performance Max ad groups, asset groups inherit listing groups built directly from feed attributes — category, brand, custom labels, product type — and not from a manually selected list of keywords. Automated bidding then distributes the budget between these groups based on available performance signals.
This operation changes what matters in the feed. An imprecise product_type or empty custom_label no longer just limits reporting: it reduces PMax’s ability to distinguish products from each other, and thus to allocate budget well. Custom_label_0 to custom_label_4, which Google describes as free filters intended for reporting and bidding, remain the main free segmentation channel available for Performance Max — organizing them by margin tier, seasonality or best-seller status allows you to guide the algorithm without touching a single keyword.
The title remains the most heavily weighted field for matching. According to Merchant Center documentation on the title field, it accepts up to 150 characters, but only the first 70 are generally displayed before truncation on Shopping mobile: beyond that, the text continues to be indexed but is practically never read by the buyer.
2. Meta Advantage+ Catalog brought Meta closer to Shopping logic
Advantage+ Catalog Ads campaigns, described in the Meta Business Help Center, automatically assemble title, image and price from the catalog, without manual intervention per reference. The algorithm selects the image, associates it with the title, tests placements on Feed, Reels and Marketplace, then reallocates the budget to the combinations that convert best — a dynamic creative recombination mechanism that approaches, in its logic, what Performance Max does on Google’s side.
The technical constraint differs from Google: product data specifications for Meta catalogs recommend staying under approximately 65 characters for the title, versus approximately 70 on Shopping mobile, while accepting up to 200 characters at the technical level. A feed designed solely for Google therefore loses readability on Meta, where the title and main image carry most of the visible message. Hence the need to manage title overrides by channel rather than a single string applied everywhere.
3. Conversational agents already read the same catalog
This is the most recent change, and the least anticipated by feed teams. An independent study conducted by Peec AI and published on Search Engine Land in March 2026 analyzed approximately 43,000 products present in ChatGPT shopping carousels: 83% of these products correspond to results already present in the top 40 of Google Shopping, and 60% of these matches come from the top 10. In other words, the quality work done on feeds for Shopping already mechanically benefits visibility on AI surfaces.
Google formalized this link in May 2026, at Google Marketing Live, by introducing a series of optional conversational attributes in Merchant Center: question_and_answer to anticipate frequently asked questions, related_product for bundle reasoning, item_group_title and variant_option to help a system understand that this is the same product in variants, or document_link to allow an agent to directly cite a manual. These attributes are optional and do not affect approval of the main feed.
4. Measurement had to follow
These changes also complicated reading results. Feed enrichment can improve CTR and CPC without the dashboard reflecting it correctly if tracking is incomplete: auto-tagging disabled on Google Ads side, Consent Mode poorly configured, or variants poorly grouped that fragment performance between sizes and colors. The method remains simple on paper — a test segment, a comparable control segment, and tracking of CPC, CTR, ROAS and coverage over two to four weeks — but it assumes reliable tracking data upstream, which is not always guaranteed.
What it concretely changes
These four developments point to the same conclusion: product data is no longer an isolated technical input by channel, but a single source read by multiple systems — Shopping, PMax, Meta Advantage+, and now AI agents. A poorly filled title, attribute or highlight is no longer a “Google” or “Meta” problem: it’s a feed problem, which has repercussions everywhere the catalog is read.
We updated our guide on product feed optimization to cover these four points in detail — title structures by sector, attributes to prioritize, preparing the catalog for AEO, and measurement method.