Key takeaways
- ChatGPT and Google AI Mode are the only two AI engines that reliably return structured shopping cards with buy links; Perplexity and Gemini mostly answer in prose, and Copilot does it about a third of the time.
- The two engines disagree on which brands to recommend nearly two-thirds of the time, according to BrightEdge's analysis of tens of thousands of identical prompts, and only 17% of queries produced the same brands everywhere.
- Google AI Mode shows noticeably pricier products than regular Google search (a median $149 vs $100), while ChatGPT leans on third-party review consensus rather than its own crawl of a brand's site.
- Both platforms pulled back their agentic checkout ambitions in 2026: OpenAI's Instant Checkout moved from product pages to retailer apps (Instacart, Target, Expedia), and Google's UCP checkout is still in early access for select US merchants.
- Promptwatch's broader citation data shows ChatGPT has almost stopped citing Reddit and social platforms overall (0.46% of citations), which complicates the popular claim that ChatGPT shopping answers are "sourced overwhelmingly from Reddit."
Two shopping surfaces, two very different machines
I'll say the thing most comparison posts gloss over: ChatGPT Shopping and Google AI Mode shopping look similar on the surface, a chat box, a few product cards, a buy button, but underneath they're running on almost opposite logic. One is a conversational layer bolted onto a brand-new product index. The other is a 50-billion-listing retail graph that Google has spent two decades building, now narrated by Gemini.
That difference shows up in the numbers. Across 3,312 product-intent prompts run through six AI engines, cloro's monitoring found ChatGPT returned a structured product card on 87% of prompts, and Google AI Mode on 91%. Perplexity and Gemini, by contrast, answered almost none of the same prompts with cards, just prose mentions of products. So if your shopping strategy is "optimize for AI," it's really a strategy about two platforms, not five.
How ChatGPT decides what to recommend
ChatGPT Shopping pulls product data from three places: merchant feeds (mostly Shopify, growing but still a fraction of the catalog), OAI-SearchBot crawling your site directly, and third-party coverage, reviews, listicles, forum threads. OpenAI doesn't charge for placement in the organic shopping cards, which is refreshing compared to a Google Shopping results page stuffed with paid listings. The tradeoff is that you have less control. You can't buy your way in.
Promptwatch's citation-type data backs up what that means in practice. Product pages are the single most-cited content type on ChatGPT overall, at 22.4% of all classified citations, and for commercial prompts specifically that rises to 26%, with landing pages at 20.6% and listicles at 15%. So four content types, product pages, landing pages, listicles, and comparisons, cover about 70% of what ChatGPT cites when someone's in buying mode. That's useful because it tells you exactly what to build: a clean product page, a landing page that states who the product is for, and enough third-party listicle coverage to get cited alongside your own content.
Here's where it gets messier, and where I think a lot of shopping-SEO advice floating around right now is already out of date. The cloro study found ChatGPT's shopping answers cite YouTube and Reddit each in about 19% of responses, with RTINGS close behind at 16%. That's the origin of the "ChatGPT shopping runs on Reddit" narrative you'll see repeated everywhere. But Promptwatch's broader citation tracking tells a very different story for ChatGPT's overall behavior: social platforms combined now make up just 0.46% of ChatGPT's citations as of early October 2026, down from 5.8% not long before. Reddit specifically fell from about 5.2% to 0.09% after a sharp collapse on August 14, 2026.
Both things can be true at once, and that's the nuance worth sitting with. Shopping-intent prompts may still pull disproportionately from UGC sources even as ChatGPT's aggregate citation mix moves away from social content for everything else. Or the cloro numbers, collected in an earlier window, are simply stale by the time you read this. Either way, don't build a GEO strategy on the assumption that Reddit presence alone gets you into ChatGPT's shopping cards. Track your own category's prompts directly rather than trusting a single snapshot study, because this is clearly a fast-moving target.
Review content itself is a smaller lever than people assume. Reviews as a standalone content type earn only 1.7% of all ChatGPT citations overall, though that climbs to 5.6% for branded prompts (someone already names your product), about three times the baseline. The takeaway: reviews matter more once a shopper already knows your brand than they do for discovery.
How Google AI Mode decides what to recommend
AI Mode is wired directly into Google's Shopping Graph, which holds more than 50 billion product listings and updates roughly two billion times an hour, interpreted through Gemini. If your structured data, Merchant Center feed, reviews, and price accuracy are already solid for regular Google Shopping, AI Mode mostly reflects that work back at you rather than requiring a separate playbook.
But AI Mode's selection behavior has a quirk that should worry anyone optimizing on price. A Productrise study found that when the same product appears in both AI Mode and traditional Google search, AI Mode shows it 21.6% more expensive on average, and across all listings the median AI Mode price is $149 versus $100 in regular search, a 49% gap. When matched products disagree on price, AI Mode is the pricier version 68.4% of the time. One SEO quoted in that report, Katelyn Geary of Break the Web, called it a reversal of Google's historical instinct to surface the cheapest listing: "Open market comparison is replaced by a curated path to higher-priced stock."
That's a genuinely strange turn for a company that built its shopping business on price comparison. It also means AI Mode shows far fewer products per query than classic search, about 3.9 on average versus 27.8, and only 1.28% of products ranking in traditional search even show up in AI Mode for the same query on the same day. If you're a budget or mid-tier brand relying on being the cheapest option, AI Mode may simply not be showing your category's bargain tier at all.
Where the two disagree
This is the part that should actually change how you think about "ranking" in AI shopping. BrightEdge ran tens of thousands of identical prompts through ChatGPT, Google AI Overviews, and AI Mode, and the platforms disagreed on which brands to surface 61.9% of the time. Only 17% of queries produced the same brands across all three. Google AI Overviews surfaced brands far more often (36.8% of queries) than ChatGPT (3.9%), and averaged 6.02 brands per query against ChatGPT's 2.37.
The gap between mentions and citations is also telling. ChatGPT mentions brands 3.2 times more often than it cites a source for them (2.37 mentions vs 0.73 citations), suggesting it's drawing on training data as much as live search. Google AI Overviews does the reverse: 14.30 citations against 6.02 mentions, meaning it shows its sources far more consistently than it name-drops brands directly.
| Dimension | ChatGPT Shopping | Google AI Mode |
|---|---|---|
| Product card rate on shopping prompts | 87% (cloro) | 91% (cloro) |
| Primary data source | Merchant feeds + crawling + third-party coverage | Shopping Graph (50B+ listings) via Merchant Center |
| Price tendency vs regular search | Not directly measured, tends to favor review consensus | 21.6% more expensive on average (Productrise) |
| Social/UGC reliance (overall citations) | 0.46% social share, Reddit collapsed to 0.09% (Promptwatch) | Social platforms at 12.2% of citations, YouTube leads at 4.28% (Promptwatch) |
| Agentic checkout protocol | ACP (with Stripe), scaled back to retailer apps | UCP, early access for select US merchants |
| Brands per query (vs AI Overviews baseline) | 2.37 average | 1.59 average (AI Mode specifically) |
Worth flagging too, since it changes where your content effort should go: AI Overviews leans on social media far more than ChatGPT does across the board, 12.2% of its citations versus ChatGPT's 0.46%, with YouTube its top social source at 4.28%. If you're deciding where to invest limited UGC or review-generation budget, Google's surfaces reward that work more consistently than ChatGPT does right now, at least for general queries outside the shopping-specific skew cloro measured.
The checkout story: both platforms pulled back
It's worth separating recommendation from transaction, because 2026 made clear these are different problems with different levels of maturity.
OpenAI's Instant Checkout, built on the Agentic Commerce Protocol with Stripe, launched with huge ambition, a million promised Shopify merchants. By February 2026, only about 30 Shopify stores were actually live, and by March OpenAI had moved in-chat purchases into retailer apps plugged into ChatGPT instead, Instacart, Target, Expedia, Booking.com. Shopify's own president said publicly the bottleneck was on OpenAI's side, not merchants'. OpenAI also apparently hadn't built out state sales-tax remittance by February, which tells you transaction volume never got close to what was promised.
Google's Universal Commerce Protocol launched checkout on standard Search listings in May 2026 and is extending into AI Mode and the Gemini app, paid through Google Pay with Wallet-stored shipping details. The merchant stays merchant of record, keeping customer data, fulfillment, and support. It's in early access for select US merchants as of October 2026, rolling out gradually to Canada and Australia. Shopify stores get UCP enabled by default.
Neither company has actually solved agentic checkout at scale yet. Three structural problems keep showing up across both protocols: merchant onboarding is a heavy lift for anyone not already on Shopify, product feeds go stale faster than anyone expects, and fraud-prevention frameworks only cover merchants who've opted in, leaving the long tail of smaller brands unreachable either way.
What this means if you're trying to get recommended
A few concrete moves, based on what the data actually shows rather than what sounds good in a LinkedIn post:
- Build product pages and landing pages that state, in the first sentence, what the product is and who it's for. These two content types alone cover roughly 47% of ChatGPT's commercial-intent citations.
- Don't assume Reddit presence alone gets you cited. ChatGPT's overall social citation share has collapsed; whatever boost UGC still gives shopping queries specifically needs to be verified against current data, not a single quarter-old study.
- If you sell on Google Shopping already, get your Merchant Center feed, reviews, and pricing accuracy locked down first. AI Mode largely inherits that foundation rather than requiring separate optimization.
- Watch your pricing positioning on AI Mode specifically. If AI Mode consistently surfaces pricier SKUs, a budget-tier product might simply not appear, regardless of how well-optimized the listing is.
- Monitor both surfaces separately rather than treating "AI shopping" as one channel. The two engines disagree on brand recommendations most of the time, so a win on one doesn't predict a win on the other.
For brands trying to track this across engines rather than guessing from anecdote, tools like Promptwatch monitor citation and shopping-card behavior across ChatGPT, Google AI Mode, Perplexity, and more, including its dedicated ChatGPT Shopping and Ads Radar tracking, so you can see which of your products get surfaced, where, and why, rather than relying on a quarterly third-party snapshot.
If you want to go broader and compare GEO platforms built for this kind of monitoring, the directory at bestgeosoftware.com is a reasonable starting point. And if the gap between what AI engines recommend and what your own analytics show is the real problem, that's less an SEO question and more a content and structured-data one, which is where an agency like 1001 SEO Media tends to get involved, building the product pages, comparisons, and schema that both engines are actually citing.
The bottom line
ChatGPT and Google AI Mode both show product cards on the large majority of shopping prompts now, but they're not the same channel wearing two outfits. One runs on a conversational layer still working out its data pipeline and checkout plumbing. The other runs on two decades of retail infrastructure with a chat interface stapled on top, and it shows in how differently they price, source, and select products. Treating them as a single "AI shopping" strategy is the mistake. Treat them as two separate recommendation engines, each with its own data source, its own pricing bias, and its own idea of what counts as a trustworthy signal, and you'll actually know what to build next.