Key takeaways:
- Your Google Merchant Center feed has become the single control point: it powers AI Overviews, AI Mode, Gemini and even ChatGPT.
- 83% of products shown in ChatGPT carousels come from Google Shopping organic results, 60% from the top 10 alone (Search Engine Land, March 2026).
- A product disapproved in Merchant Center is excluded from AI answers, whatever the quality of your structured data.
- Merchant Center’s official “AI share of voice” measurement report exists, but not yet in France as of today.
For an online store, visibility in AI Overviews is not decided by the product page alone. It is decided in your Google Merchant Center feed, which now powers every AI shopping surface.
Does ChatGPT recommend your brand?Measure your presence and identify the brands cited in your place. No credit card required.
What changes for product listings with AI Overviews?
The decision point shifts from the product page to the data feed. On classic search, Google read your product page on your site to rank it. On generative surfaces, it draws first from your Google Merchant Center feed, the structured database describing each of your products by its attributes.
This feed no longer serves Shopping ads alone. It now powers free product listings, AI Mode in Search, product discovery via Gemini and conversational commerce. A single data source feeds all these surfaces at once.
The consequence fits in one sentence: your feed quality determines your AI visibility before your site content does. Two merchants selling identical products get different visibility depending on which has the most complete structured attributes, not on which has the best-written listing.
Why does your Merchant Center feed also power ChatGPT?
Because ChatGPT does not build its own product index: it queries Google Shopping. This is the most counter-intuitive discovery of the year for an online store, and it is documented.
A study published by Search Engine Land in March 2026, covering roughly 5,000 carousels and 43,000 products, establishes that 83% of products shown in ChatGPT carousels match the top 40 organic Google Shopping results, and 60% the top 10 alone. Researchers even found, in ChatGPT’s source code, Google Shopping parameters encoded in base64, proof of a direct data pipeline.
The mechanism rests on two parallel circuits. To write its answer text, ChatGPT runs classic web searches. To fill the product carousel, it sends a separate shopping query to Google Shopping and retrieves the organic listings. Your presence in the carousel therefore depends on your Google Shopping position, not on your standing with the model.
The reach extends beyond ChatGPT. The same Merchant Center feed serves as a source for Gemini, Perplexity, Copilot and shopping assistants. Optimising a single feed improves your visibility across all these surfaces at once.
| AI shopping surface | Main product source | Action lever for you |
|---|---|---|
| Google AI Mode | Merchant Center feed, direct | Feed compliance and completeness |
| Google AI Overview | Merchant Center feed and Shopping index | Feed compliance and Shopping position |
| Gemini (shopping) | Merchant Center feed, direct | Attributes and structured data |
| ChatGPT (carousel) | Google Shopping organic results | Position in the Shopping top 40 |
Sources: Google Merchant Center Help (2026) and Search Engine Land (March 2026). The Merchant Center feed is the common denominator of all these surfaces.
“ChatGPT Shopping is mostly Google Shopping wearing a conversational costume.”
Which product attributes matter most for AI?
Structured attributes take priority over prose. A shopping agent evaluates a product on rational criteria, the feed data, not on page design or brand storytelling.
Five attributes weigh more than others in selection by generative surfaces:
- The structured title: it carries the keyword and context. With product type, it is the field that most influences Google Shopping visibility.
- GTIN and product type: they categorise the item and let systems match it to a conversational query.
- Real-time availability and price: a mismatch between feed and page lowers the trust placed in the listing.
- Specification attributes: colour, material, size, style. Google now flags products missing these via a completeness score.
- Shipping and return signals: shipping cost, speed, return policy. Without these signals, an agent switches to a competitor whose feed provides them.
Google introduced conversational attributes in 2026, a new field to enrich descriptions for better matching with natural-language queries. This rollout is global, unlike the measurement report discussed below.
The 3 technical image requirements for 2026
- Minimum resolution raised to 500 × 500 pixels, with enforcement from 31 January 2027 (warnings from April 2026).
- Stricter field practice for AI commerce: at least three images, including one in use context.
- Multimodal vision models cross-reference text and image: a photo contradicting the listing creates a visibility gap.
How does a product get excluded from AI answers?
Through a feed compliance failure, before any question of optimisation. This is the point most guides overlook: a product disapproved in Merchant Center is removed from AI Mode and Gemini Shopping answers, whatever the quality of your structured data on the site.
Three causes of exclusion recur most often:
- A disapproved product for breaching Merchant Center rules, often linked to missing required data.
- A price or availability mismatch between feed and product page, which Google reads as an unreliability signal.
- Incomplete product data, filtered upstream before the generative surface even sees the product.
In my view, this is the most important reversal of priorities for an online store in 2026. Money goes into persuasive listing copy and design, when the first battle is won on feed hygiene. A catalogue with 15% disapproved products loses 15% of its AI surface before a single description line is written. The first task is not creative, it is technical: bring the disapproval rate to zero, align prices, complete missing attributes. This foundation is invisible on the site, but it decides everything upstream.
Does AI traffic convert better for e-commerce?
Available data suggests so, with a methodological caveat. Several analyses place the conversion rate of AI-engine traffic above that of classic organic traffic. The intent-based explanation is coherent: a user asking an assistant for the best product in a category expresses more advanced intent than a broad query.
The caveat concerns measurement. A large share of AI-referred traffic arrives without an identifiable referrer and falls into the direct traffic of your analytics. Conversion figures attributed to AI are therefore floors, calculated on the measurable fraction, not the real total.
A second movement is emerging, more structural than traffic itself. Agentic commerce, where an AI agent completes the purchase for the user, is progressing. These agents evaluate products on strictly rational criteria, namely structured data, availability, price and reviews, with no sensitivity to design or brand narrative. A clean feed then becomes the entry condition, not an advantage.
Gartner anticipates a 25% drop in traditional search volume by the end of 2026. For an online store, this shift does not mean fewer sales, but a move in the discovery point, from the classic results page to conversational surfaces powered by your feed.
How do you measure product visibility in AI in France?
With one important limit to know before investing: Google’s official tool is not yet available in France. Google launched an AI Performance Insights report in Merchant Center in 2026, showing a brand’s share of voice on its AI surfaces against comparable competitors, across the discovery, evaluation and purchase stages.
This report currently covers only the United States, Canada, Australia, India and New Zealand, with a pilot on a limited number of accounts. No European rollout is confirmed as of today. A French online store cannot yet rely on it, contrary to what some guides imply.
As with editorial-side presence measurement, this report shows only impressions, not clicks or sales. It answers “am I visible?”, not “does it convert?”. Share of voice remains a leading indicator to watch, not a commercial result.
While waiting for the French opening, three measurement levers are accessible right now:
- The feed compliance diagnosis in Merchant Center, free, listing your disapproved products and missing attributes.
- The manual test in neutral conditions, querying AI Mode, Gemini or ChatGPT on your categories to see which products and brands surface.
- An AI visibility tracking tool, for share of voice against competing brands on your strategic queries, across several engines.
On that last lever, this is the need I set out to cover by co-founding Cockpyt AI, which tracks a brand’s presence in ChatGPT, Perplexity and Gemini with share of voice against cited competitors. Dedicated tracking of Google’s AI Overviews and AI Mode is in development and will soon join those three engines.
Disclosure: I am a co-founder of Cockpyt AI. I mention it with first-hand knowledge of its current scope and limits, and I distinguish what the tool does today from what is coming.
Does ChatGPT recommend your brand?Measure your presence and identify the brands cited in your place. No credit card required.
Where do you start to make your product listings visible in AI?
With the feed before the listing, in an order that goes from blocking to refinement. The French rollout of AI Overviews, live since 22 July 2026, opens a window to take position before your competitors adjust their catalogue.
- Clean the feed. Bring the disapproval rate to zero and fix price and availability mismatches. This is the absolute prerequisite, no other effort compensates for a non-compliant feed.
- Complete the attributes. Fill in GTIN, product type, colour, material, size, and aim for a high completeness score across the catalogue.
- Enrich conversational attributes and images. Add conversational descriptions, move to three images minimum including one in context, align text and visual.
- Mark up the listings. Deploy Product markup with aggregate reviews on each product page, consistent with the feed.
- Measure and iterate. Track product share of voice and the feed diagnosis weekly, tightening on high-stakes categories.
This order is non-negotiable. An online store investing in images before cleaning its feed optimises products the AI surfaces do not see.
Frequently asked questions
Do you need a Google Merchant Center account to appear in ChatGPT?
In practice, it is the main lever. Since 83% of ChatGPT carousel products come from Google Shopping organic results, a complete, compliant Merchant Center feed is the most direct route to appear there. Other channels exist, but none covers such a large share.
Is Schema.org structured data enough for e-commerce?
No, it complements the feed without replacing it. Product markup with aggregate reviews strengthens understanding of your pages, but a product disapproved in Merchant Center stays excluded from AI answers despite perfect markup. The feed takes priority, markup consolidates.
Is the Merchant Center AI Performance Insights report available in France?
Not as of today. The pilot covers the United States, Canada, Australia, India and New Zealand. No European timeline has been confirmed. In the meantime, the feed compliance diagnosis and the manual test in neutral conditions remain accessible.
Can a small shop compete with big brands in AI?
On data completeness, yes. Share of voice in AI Mode is decided by feed quality and completeness against competitors, not by a bidding budget. Two merchants in the same category get different visibility depending on which has the most complete attributes.
Should you keep investing in paid Google Shopping?
The matter goes beyond advertising. The same feed powers Shopping ads and AI surfaces. Reducing attention to this feed means weakening the data source on which both your ads and your generative visibility depend. The budget trade-off is yours, but feed quality benefits both.
Do product images really influence AI visibility?
Yes, in two ways. The minimum resolution rises to 500 × 500 pixels with enforcement in early 2027, and vision models cross-reference the image with the listing text. A single, low-quality photo, or one contradicting the description, creates a visibility gap against a better-equipped competitor.
Primary sources
- Google Merchant Center Help, “Insights for AI-powered shopping experiences coming soon”, 27 May 2026 (share of voice, shopping funnel, attributes; pilot US, Canada, Australia, India, New Zealand), support.google.com
- Tom Wells, “ChatGPT sources 83% of its carousel products from Google Shopping via shopping query fan-outs”, Search Engine Land, 5 March 2026 (5,000 carousels, 43,000 products; 83% top 40, 60% top 10), searchengineland.com
- Google Search Central, “Optimizing your website for generative AI features on Google Search”, updated 10 July 2026, developers.google.com
- Google Marketing Live 2026, announcement of AI Performance Insights and conversational attributes in Merchant Center, 20 May 2026, relayed by Search Engine Land, searchengineland.com
- Google Merchant Center, 2026 product data specification update (image resolution 500 × 500, enforcement 31 January 2027)
- Gartner, forecast of a 25% drop in traditional search engine volume by the end of 2026
