ChatGPT Shopping Research: What It Is & How Retailers Use It

The way people find products online is changing, and most retailers haven't caught up yet.

Shoppers used to start on Google. They'd enter a query, scan a results page, click around, open tabs, read reviews, go back, click again. It worked, more or less. 

But it was friction-heavy, especially for decisions that required comparing options, weighing tradeoffs, or figuring out which of twelve vacuum cleaners actually fit a small apartment.

ChatGPT shopping research changes that flow. Instead of links, you get a conversation. Instead of ten tabs, you get one buyer's guide built around what you specifically need.

For a growing number of consumers, that's a better experience. For retailers, it's a new discovery channel, with different rules than anything they've optimized for before.

This article explains how ChatGPT shopping research works, what the current data shows about how consumers use it, and what retailers can actually do to show up in it.

What Is ChatGPT Shopping Research?

ChatGPT shopping research is a product discovery feature launched by OpenAI in November 2025. It's available to all logged-in ChatGPT users (Free, Go, Plus, and Pro) across mobile and web.

The feature works differently from a standard ChatGPT response. When you ask a shopping question, it triggers a dedicated research mode. 

ChatGPT asks clarifying questions (budget, use case, who it's for, which features matter), then pulls data from across the web and returns a structured buyer's guide: multiple product options, key differences, tradeoffs, and up-to-date information from retailers.

It's powered by a version of GPT-5 mini trained specifically for shopping tasks. OpenAI's internal benchmarks put product accuracy at 52% on multi-constraint queries. 40% better than standard ChatGPT Search for the same type of question. (OpenAI, November 2025)

It's designed for decisions with depth. Electronics. Beauty. Kitchen appliances. Home and garden. Outdoor gear. Products where "what should I buy?" isn't a quick answer. For simple queries, like checking a price or confirming a feature, a regular ChatGPT response still handles it faster.

ChatGPT Pulse, available to Pro users, adds a proactive layer: if you've been discussing a topic in past conversations (say, e-bikes), a future Pulse card can suggest a relevant buyer's guide without you asking.

Why Shoppers Are Using It

Hundreds of millions of people already use ChatGPT to find, understand, and compare products. Shopping research is built on top of that existing behavior, not creating something new, but structuring a conversation that was already happening.

The growth numbers make it harder to ignore. Shopping queries on ChatGPT jumped from 7.8% to 9.8% of all searches in just the first half of 2025. A 25% category gain on top of a 70% overall increase in ChatGPT usage. That means shopping queries effectively doubled in six months. (Bain & Company / Sensor Tower, 2025)

Consumer surveys reinforce the shift. 64% of consumers plan to use AI chatbots for shopping in 2026 — with 26% planning to use them more, 24% using them at the same level as 2025, and 13% trying AI shopping for the first time. Nearly 1 in 4 say they plan to make AI their default way to shop. (PartnerCentric survey, 1,004 consumers, December 2025)

Something else worth paying attention to: the behavior of heavy ChatGPT users. Among daily AI users, 70% had already tried AI shopping in 2025, spending an average of $540 across 9 transactions, and 44% plan to use it even more in 2026. These are the early adopters who shape what mainstream behavior looks like two years later.

The reason shoppers gravitate to this feature isn't really about AI. It's about friction. The traditional journey looked like this: search query → list of sites → comparison → selection. 

Now: dialogue with AI → ready recommendation → confirmation or correction. Instead of "where to buy iPhone 15 Pro," a user asks "recommend a smartphone for photos under $1,000" and gets 1–3 options instead of 20 links.

That compression is what makes the channel interesting, and what makes the buyer who arrives at a product page via ChatGPT different from one who arrives via Google.

What the Conversion Data Actually Shows

There's a lot of noise about AI shopping. The conversion data is more complicated than the hype suggests, and it's worth looking at it honestly.

A 12-month GA4 analysis by Visibility Labs, covering January through December 2025 across 94 e-commerce brands, is one of the cleaner data points available. 

ChatGPT traffic converted at 1.81% compared to 1.39% for non-branded organic search; a 31% higher conversion rate. It outperformed organic in 10 of 12 months. 

The researchers attribute the gap to what they call "intent compression": users refine their needs through conversation before clicking, arriving at a product page closer to a decision than a typical search visitor.

That sounds like good news for retailers. But the same study found that non-branded organic traffic was 70x larger than ChatGPT traffic overall, narrowing to 47x in Q4 2025. 

ChatGPT generated $474K versus organic's $32.1M, just 1.48% of non-branded organic search revenue across the 94 stores. The quality is there. The volume isn't, yet.

A separate, larger study tells a less optimistic story at scale. Researchers at the University of Hamburg and Frankfurt School of Finance & Management analyzed 12 months of data from 973 e-commerce sites generating $20 billion in annual revenue. 

The results "contradict widespread expectations of LLM superiority." Affiliate links were 86% more likely to convert than ChatGPT referrals, while organic search outperformed ChatGPT by roughly 13%. (Kaiser & Schulze, SSRN, October 2025)

The same researchers found something reassuring, though: ChatGPT conversion rates improved steadily throughout the observation period. As traffic volume expanded, conversion rate and revenue per session rose, suggesting that shoppers referred by ChatGPT are gradually learning to trust and act on the platform's recommendations.

The most honest read on the data right now: ChatGPT is a discovery channel, not yet a conversion engine. Visitors arrive informed and engaged. They don't always buy in the same session; they often go back to Google to search for the brand directly. 

The ChatGPT funnel frequently looks like this: ChatGPT conversation → branded Google search → purchase. Those conversions show up as branded organic traffic in standard analytics, which means ChatGPT's influence is substantially undercounted in most reporting.

Post-purchase surveys that ask "how did you first hear about us?" are currently the most reliable way to measure AI's actual contribution.

What Retailers Are Doing About It

The retailer's response to ChatGPT shopping has gone through two distinct phases.

Phase 1: Instant Checkout (September 2025)

When OpenAI launched Instant Checkout in September 2025, retailers moved fast. Etsy, Walmart, and Shopify all lined up to let users buy directly inside ChatGPT. The idea was that removing the step of going to a retail website would increase conversions.

It didn't work as expected. Walmart found that conversion rates for products sold directly in ChatGPT were three times lower than those that rerouted users to Walmart's own website for checkout. 

OpenAI ended up ending Instant Checkout in early 2026. The problem, according to analysts, was infrastructure: OpenAI was scraping some retailer websites for product data, which meant stock levels, delivery estimates, and pricing were often inaccurate or out of date. Consumers noticed. Trust is hard to build and easy to lose.

Phase 2: Discovery First (2026)

OpenAI's current strategy, reflected in the March 2026 launch of the Agentic Commerce Protocol (ACP), positions ChatGPT as a discovery and comparison layer, not a checkout destination. 

Through ACP, merchants share product feeds and promotions so their catalogs are fully represented in ChatGPT. Leading retailers, including Target, Sephora, Nordstrom, Lowe's, Best Buy, The Home Depot, and Wayfair, have already integrated into ACP for discovery.

The architecture is cleaner. Retailers control the transaction, the customer data, and the brand experience at checkout. ChatGPT handles the research and recommendation. 

Companies, including Target, Nordstrom, Lowe's, Home Depot, and Wayfair, provide product data to ChatGPT while retaining payment authority within their own sites. 

Sephora offers personalized recommendations and membership integration through a dedicated ChatGPT app, with payment functionality deferred to its own platform.

For Shopify merchants, participation doesn't require any additional setup. Products from Shopify merchants are automatically discoverable in ChatGPT via Shopify Catalog, no opt-in required.

How AI Decides Which Products to Recommend

This is the part most retailers either don't know or underestimate.

ChatGPT doesn't rank products by ad spend. There's no auction. Recommendations are organic and based on publicly available data. OpenAI has confirmed it won't share user chat data with retailers.

What determines which products surface comes down to data quality and structure. Four critical signals determine whether a product gets included in a recommendation: 

structured data (Schema.org microdata for Product, Offer, and Review), completeness of product attributes, usage scenarios, and third-party validation signals like reviews and ratings.

The reasoning is simple: AI can only recommend what it can read. A store with less traffic but well-designed product cards is more likely to appear in AI recommendations than a large site with creative but poorly structured content. 

Schema markup matters. JavaScript-rendered content is a problem. Erlin's own data puts AI parsing success at 94% for static HTML with schema, dropping to 23% for JavaScript-rendered content and 7% for PDFs. (Erlin data, 500+ brands, 2026)

Product descriptions also need to work differently for AI than for SEO. An AI shopping model reads for specificity: who the product is for, what it does better than alternatives, and what constraints it fits. 

"Premium performance vacuum" tells the model nothing useful. "Cordless stick vacuum, under 6 pounds, 65 dB, suited for apartments under 800 square feet" gives it something to match against a user's stated requirements.

Erlin's benchmark data shows how much attribute completeness drives citation rates. Brands with 9+ structured product facts achieve 78% average AI coverage. 

Brands with 2 or fewer structured facts average 9% coverage. That's not a small gap; it's a different universe. (Erlin data, 500+ brands, 2026)

Third-party signals matter too. Shoppers coming from ChatGPT converted to Amazon purchases at 1.7 times the rate of those from Google (12% versus 7%), with an 11% higher average order value, and they stayed 46% longer on Amazon and viewed an average of five products compared to three for Google transitions. (RealityMine, January 2026) But to get there, your product needs to surface in the first place.

What Retailers Should Do Now

Most of what moves the needle on ChatGPT visibility isn't a new capability; it's fixing the basics that AI can actually use.

Get your product data structured 

Every product needs Schema.org markup for Product, Offer, and Review. This is the baseline. Without it, AI either ignores the page or uses it partially. 

Erlin's data shows structured data formats (comparison tables, FAQ schema, llm.txt) drive 28–34% coverage lift within 14–21 days of implementation. (Erlin data, 500+ brands, 2026)

Write product descriptions for constraint matching

Think about the clarifying questions ChatGPT asks users: What's your budget? Who is it for? What features matter most? Your product pages should answer those questions in plain, parseable language. 

Usage scenarios: "ideal for apartments under 60 square meters," "works for fine and thick hair," "suitable for beginners", are the most underdeveloped content blocks in most product catalogs.

Build your review presence

68% of AI citations come from third-party sources, not brand-owned websites. (Erlin data, 500+ brands, 2026) Reviews on G2, Capterra, Amazon, and similar platforms are what AI models treat as validation. A brand with 100 recent, specific reviews outperforms a brand with a polished product page and no external presence.

Connect your product feed through ACP or Shopify Catalog

If you're a Shopify merchant, your products are already discoverable in ChatGPT. If you're not, applying to participate through OpenAI's Agentic Commerce Protocol gives the system access to complete, accurate, real-time product data, which directly reduces the gap between what ChatGPT shows users and what's actually available on your site.

Track it

GA4 doesn't create a default channel grouping for AI traffic. Build a custom channel definition that classifies sessions from ChatGPT, OpenAI, Perplexity, Claude, and Gemini as a distinct AI Referral channel. 

Layer in post-purchase surveys to capture the full picture; many AI-influenced conversions are misattributed to branded organic search because users verify through Google before buying.

What This Looks Like in Practice

Walmart is the clearest example of a retailer navigating this well. Rather than treating ChatGPT as a checkout channel, they've integrated their AI shopping assistant "Sparky" into both their own app and ChatGPT. 

Even when users explore products inside ChatGPT, account linkage, reward program application, and payment all flow back through Walmart's own environment. Discovery happens in AI. The relationship stays with Walmart.

Sephora has taken a similar approach: personalized beauty recommendations and loyalty account integration through a dedicated ChatGPT app, with checkout handled separately. 

Target, Lowe's, and Home Depot have all connected product feeds through ACP without giving up transaction control. The pattern is consistent: AI for discovery, proprietary platform for the transaction. That's the model that's working.

The Honest Assessment

ChatGPT shopping research is real, growing, and worth preparing for. It is not, today, a revenue channel that replaces anything. Traffic volumes are small relative to organic search. Conversion data is mixed. The infrastructure is still being built.

What it is: an early-funnel discovery layer that's growing fast, attracts high-intent buyers, and is increasingly difficult to ignore as a category. During Black Friday 2025, shoppers coming from ChatGPT converted on Amazon at 1.7x the rate of Google-referred shoppers, with 11% higher average order value. That's the signal inside the noise.

The retailers doing the right work now, like fixing product data structure, building review presence, and connecting product feeds, are positioning for a channel that's compounding. The ones waiting for volume to justify the work will be optimizing after the category is already locked.

Only 16% of brands currently track their AI search performance systematically. (Erlin data, 2026) That gap is both the problem and the opportunity.

Frequently Asked Questions

What is ChatGPT shopping research?

ChatGPT shopping research is a product discovery feature launched by OpenAI in November 2025. It works as a conversational buyer's guide. ChatGPT asks clarifying questions about budget, use case, and preferences, then pulls product data from across the web to return personalized recommendations and comparisons. It's available to all logged-in ChatGPT users on Free, Go, Plus, and Pro plans.

How does ChatGPT decide which products to recommend?

ChatGPT shopping recommendations are organic; there's no paid placement. Products surface based on data quality and structure: Schema.org markup, completeness of product attributes, usage scenarios, third-party reviews, and pricing accuracy. Structured product feeds shared through OpenAI's Agentic Commerce Protocol (ACP) or Shopify Catalog give retailers a more complete representation in results.

Do ChatGPT referrals convert to purchases?

Yes, but with nuance. ChatGPT-referred visitors show higher engagement than typical organic search visitors. Across 94 e-commerce brands, ChatGPT traffic converted 31% higher than non-branded organic search in 2025. (Visibility Labs, 2025) However, the absolute volume is small; organic search was still 70x larger in traffic terms. Many AI-influenced sales are also misattributed to branded organic search because users search the brand on Google before completing the purchase.

Can I pay to appear in ChatGPT shopping results?

No. OpenAI has confirmed that shopping research results are organic and based on publicly available retail data. Retailers can improve visibility through product data quality, Schema.org markup, review presence, and connecting product feeds through ACP or Shopify Catalog, but not through paid placement.

What categories perform best in ChatGPT shopping?

ChatGPT shopping research performs best in detail-heavy categories where comparisons matter: electronics, beauty, home and garden, kitchen and appliances, and sports and outdoor. Simple purchases with low decision complexity (buying a specific known product) don't benefit as much from the research mode.

What happened to ChatGPT Instant Checkout?

OpenAI ended Instant Checkout in early 2026 after finding that in-ChatGPT conversion rates were far below those on retailer websites. Walmart found that in-ChatGPT conversions were three times lower than on their own platform. OpenAI has shifted to a discovery-first model through the Agentic Commerce Protocol, where AI handles product research and comparison while retailers manage the transaction.

Get your brand's AI visibility score across ChatGPT, Perplexity, Gemini, and Claude. 

Start Free Audit →