- Heidi Schwende7 days ago4 min read
Things are moving fast. Too fast to write something once and walk away.
Back in September, I wrote about ChatGPT's Instant Checkout launch—Etsy sellers, the Agentic Commerce Protocol, and a promise about Shopify expansion. In that post, I said this was worth watching but not worth panicking over.
Three months later, the landscape has already shifted. New retail partners, new features, and actual conversion data are forcing me to update what I told you then.
The Conversion Numbers Are Ridiculous
Seer Interactive's recent analysis found ChatGPT shopping sessions convert at 15.9% compared to 1.8% for Google Organic. That's nearly 9x higher.
Before you get too excited: the traffic volume is still tiny. Most retailers are seeing well under 1% of sessions from ChatGPT. But the pattern is consistent across industries—low volume, disproportionately high conversion.
What's driving this? Intent compression. By the time someone asks ChatGPT "best carry-on under $300 with good spinner wheels," they've already done the mental work of defining what they want. The AI conversation filters out browsers. What's left are buyers.
The Attribution Problem Is Still Real
Here's what most analytics dashboards miss: Similarweb data shows journeys that start in ChatGPT and end in a purchase elsewhere happen about 5x more often than direct ChatGPT-to-retailer clicks.
Someone asks ChatGPT for recommendations, gets three options, then Googles the specific product name to read more reviews before buying on Amazon. Your GA4 shows that as an organic or direct conversion. ChatGPT influenced the entire decision but gets zero credit.
This is exactly the measurement problem I flagged in my post on how to measure AI search visibility. If you're only measuring referral traffic from ChatGPT, you're seeing maybe 20% of its actual impact on your sales.
What's Changed Since September
- Walmart and Target are live
Walmart and Target are live
This isn't just indie Etsy sellers anymore. Mass-market retail is in the mix, which means ChatGPT's product recommendations now span price points and categories that matter for mid-market brands competing in those spaces. (I covered Google's parallel moves in AI shopping—the convergence is real.)
- Shopping Research launched in November
Shopping Research launched in November
OpenAI built a dedicated shopping model (based on GPT-5 mini) that generates personalized buyer's guides. It asks clarifying questions, pulls real-time data from retailers, and presents tradeoffs—not just product cards. This is the conversational shopping experience they've been building toward.
- The accuracy is getting better, but it's not great
The accuracy is getting better, but it's not great
OpenAI's own testing shows Shopping Research hits 64% accuracy on product recommendations. Better than their other models, but still means roughly 1 in 3 suggestions might miss the mark on price, features, or availability. Users will learn to verify.
What Actually Drives Visibility
I've covered schema markup and structured data before—that foundation still matters. But three months of data is revealing what separates brands that show up from brands that don't:
- Third-party mentions carry more weight than your own content
Third-party mentions carry more weight than your own content
ChatGPT pulls from reviews, forums, and publisher content to build product labels like "durable" or "budget-friendly." Brands with coverage in buyer's guides, Reddit threads, and niche review sites get cited more often than brands with polished product pages but no external validation.
- Variant data is a mess
Variant data is a mess
Search Engine Land's analysis found users asking for "black sneakers" sometimes see navy. "King-size sheets" pulls Cal King. If your size, color, and variant information isn't crystal clear and consistent across every data source, you'll get filtered out or misrepresented.
- Server-rendered schema beats JavaScript injection
Server-rendered schema beats JavaScript injection
Sites using JSON-LD that's present in the initial page load perform better than those adding structured data via scripts that run after page render. ChatGPT's crawlers may not be waiting around for your JavaScript.
The traffic is small. The conversion rates are high. The attribution is broken. The technology is improving but imperfect.
So what do you actually do with this?
If you're an e-commerce brand: treat ChatGPT as a canary, not a channel.
The optimization work that makes you visible in ChatGPT—clean data, natural language descriptions, recent reviews, external mentions—is the same work that will matter for Google's AI shopping features, Perplexity, and whatever Amazon is building.
This is what I mean when I talk about where AI search is actually headed—the platforms are converging on the same signals.
If you're not seeing ChatGPT in your analytics at all, that's a signal. It means your product data isn't structured well enough for any AI system to recommend you. Fix that problem and you fix visibility across every platform moving in this direction.
The brands treating this as "wait and see" are the same ones who ignored mobile optimization in 2014. By the time the volume gets big enough to show up on their dashboards, the early movers will own the AI recommendation landscape.
Not sure where to start? That's what we do. Let's talk about what AI visibility looks like for your business.
Sources:
Seer Interactive – conversion rate data
Similarweb – attribution/journey data
OpenAI – Shopping Research and Instant Checkout announcements
Search Engine Land – variant data and visibility analysis
Digital Commerce 360 – Walmart/Target partnerships and accuracy stats
- eCommerceAdaptive SEO
