ChatGPT Shopping & Ecommerce Statistics 2026: Traffic, Conversion and Checkout

ChatGPT stopped being a research tool and started being a shop counter on 29 September 2025. That was the day OpenAI switched on Instant Checkout, letting US users buy from Etsy sellers without ever leaving the chat, and published the Agentic Commerce Protocol it built with Stripe.

Eleven months on, the picture is messier and more interesting than the launch hype suggested. Traffic from generative AI tools to US retail sites grew several hundred percent over the 2025 holiday season, and AI-referred shoppers now convert better than average. Yet buying inside the chat itself has been slow to catch on, and Walmart has said publicly that in-chat checkout converts worse than a simple click through to its own site.

This article pulls together the verified numbers on ChatGPT as a shopping channel specifically. For AI shopping behaviour across all platforms, see our broader AI shopping statistics piece. Here, the question is narrower: what is ChatGPT actually doing for, and to, ecommerce?

Key ChatGPT shopping statistics for 2026

• Generative AI traffic to US retail sites rose 693% year on year across November and December 2025, according to Adobe Analytics data reported by Digital Commerce 360.
• On Cyber Monday 2025 alone, generative AI traffic to retail sites was up 670% year on year, against record online spend of $14.25 billion, per Adobe.
• AI-referred holiday shoppers converted 31% more often than other traffic sources, and revenue per visit from AI referrals rose 254% year on year, per Adobe Analytics.
• Salesforce attributes $262 billion of global online holiday revenue, 20% of all orders, to AI and AI agents in November and December 2025.
• Shoppers arriving from AI search channels such as ChatGPT converted 9 times more often than social media referrals over the 2025 holidays, per Salesforce.
• ChatGPT's shopping interface triggers on roughly 9% of prompts overall: 12.1% of open-ended product prompts but only 3.1% of brand-name prompts, per a Profound study of 260 million prompt-level results.
• 9.5% of ChatGPT prompts carry commercial intent and a further 6.1% transactional intent, per a separate Profound intent study.
• Instant Checkout launched with Etsy sellers and a promise of "over a million" Shopify merchants, in front of what OpenAI said were more than 700 million weekly ChatGPT users.
• Only around 3% of transactions currently involve AI agents, but 33% of consumers expect at least a tenth of their purchases to be AI-driven within a year, per Checkout.com.
• 89% of merchants say they are actively preparing for agentic commerce, per Checkout.com.
• More than 21% of ChatGPT's outbound referral clicks go to Google rather than destination sites, per Semrush clickstream analysis.

What is Instant Checkout, and how big is its reach?

Instant Checkout lets a US ChatGPT user complete a purchase inside the conversation. The merchant stays the merchant of record, pays a small fee on completed sales, and fulfils the order through its existing systems. It launched on 29 September 2025 with Etsy sellers first.

OpenAI said at launch that more than 700 million people were using ChatGPT each week, and that "over a million" Shopify merchants, including Glossier, SKIMS, Spanx and Vuori, would follow Etsy [1]. Initially it handled single-item purchases only, with multi-item carts planned later, and it remains US-only at the time of writing.

The plumbing matters as much as the feature. Stripe co-developed the underlying Agentic Commerce Protocol with OpenAI and released it as an open standard, built around a Shared Payment Token that lets ChatGPT initiate a payment without ever exposing the buyer's card details, scoped to a specific merchant and cart total [2].

The strategic significance is the protocol, not the button. An open standard that any payment provider and any merchant can adopt is a bid to make ChatGPT the default storefront layer for the whole of ecommerce, in the way HTTP made the browser the default layer for content. Whether merchants want that is a different question, covered below.

How many ChatGPT prompts are actually about shopping?

A minority, but a commercially meaningful one. The best available evidence puts commercial and transactional intent at around 15% of prompts combined, and finds ChatGPT's shopping interface activating on roughly 9% of prompts.

Profound classified a large sample of real ChatGPT conversations in mid-2025 and found 9.5% carried commercial intent and 6.1% transactional intent, against 37.5% generative and 32.7% informational [10]. Profound describes that transactional share as roughly nine times what traditional search patterns would predict.

A later Profound study took a different angle, tracking when ChatGPT's shopping feature actually fires. Across 260 million prompt-level results spanning 13,000 product categories between September 2025 and January 2026, the shopping surface activated on about 9% of prompts [9].

The detail is where it gets useful for retailers. Open-ended prompts describing a need without naming a brand triggered shopping 12.1% of the time; prompts naming a specific brand triggered it just 3.1% of the time. And the behaviour is sticky: if a prompt triggered shopping one day, there was an 83% chance it triggered again the next.

Read together, these numbers say ChatGPT treats shopping as a discovery surface, not a brand search engine. If a shopper already knows your brand, ChatGPT mostly just answers. The commercial exposure, and the risk, sits in the unbranded "what should I buy" prompts where the model chooses whose products appear.

How much shopping traffic did ChatGPT send retailers over the 2025 holidays?

An enormous relative increase from a small base. Adobe's holiday data, drawn from more than 1 trillion visits to US retail sites, showed generative AI traffic up 693% year on year across November and December 2025, with the peak on the biggest trading days.

Figures reported by Digital Commerce 360 from Adobe Analytics show retail led every other sector for AI traffic growth over the holidays [5]:

Adobe's own Cyber Monday release put generative AI traffic up 670% year on year on the day, and up 760% for the season to that point, against record Cyber Monday spend of $14.25 billion [4].

Adobe groups all generative AI referrers together rather than breaking ChatGPT out, but every independent measure of AI referral traffic puts ChatGPT as the dominant single source, so the retail curve is substantially a ChatGPT curve. The caveat worth holding onto is the base effect: several-hundred-percent growth still leaves AI referrals as a low single-digit share of total retail traffic. The direction is what matters, not the current size.

Do ChatGPT-referred shoppers actually buy?

Yes, and increasingly at better rates than average traffic. Over the 2025 holiday season, Adobe found AI-referred shoppers converted 31% more often than other sources, a gap that had roughly doubled since the 2024 holidays.

The same Adobe holiday dataset showed revenue per visit from AI referrals up 254% year on year, AI-referred shoppers spending 45% longer on retailers' sites, viewing 13% more pages and bouncing 33% less often [5].

That is a striking reversal. As late as July 2025, Adobe's own reporting had AI traffic converting 23% worse than non-AI sources, itself an improvement from a 49% gap that January [6]. In roughly a year, AI referrals went from clearly inferior traffic to clearly superior traffic.

The likely explanation is qualification. A shopper who has already interrogated a chatbot about options, prices and trade-offs arrives at the product page with most of the funnel behind them. The retailer sees fewer visits than search would send, but each one is closer to a decision. For merchandising teams, that shifts the job from capturing attention to not fumbling a nearly-closed sale.

How much ecommerce revenue is AI actually influencing?

On Salesforce's measure, a fifth of global online holiday revenue. The company attributes $262 billion of November-December 2025 online sales, 20% of all orders, to AI and AI agents, within a record $1.29 trillion global holiday total.

Salesforce's holiday analysis, drawn from more than 1.5 billion shoppers across 89 countries, also found traffic from third-party AI search channels such as ChatGPT and Perplexity doubled year on year, and that shoppers arriving from those channels converted 9 times more often than social media referrals [7]. Retailers running their own branded AI agents grew online sales 6.2% year on year against 3.9% for those without.

During Cyber Week alone, Salesforce counted $336.6 billion in global online spend, with $67 billion of it AI-driven [8].

Treat the $262 billion with appropriate scepticism. "AI-influenced" is a generous attribution category that includes on-site AI recommendations and chat assistance, not just ChatGPT referrals, and Salesforce sells the AI agent software being credited. The cleaner signal is the relative one: AI search referrals doubled and out-converted social by 9 to 1. Whatever the true influenced-revenue figure, the channel comparison favours AI search decisively.

Are people actually buying inside ChatGPT?

Mostly not yet. Instant Checkout's first months underperformed the launch narrative, and the strongest criticism came from OpenAI's own retail partners.

Reporting by CNBC and Search Engine Land in March 2026, summarised by Digital Applied, quoted Walmart executive Daniel Danker saying in-chat checkout converted roughly 3 times worse than sending the same shopper through to the retailer's own site, whilst the same ChatGPT traffic delivered about twice the new-customer acquisition of search engines [13]. The same round-up cites a Forrester estimate that only around 30 Shopify merchants were live on Instant Checkout by February 2026, against the million-plus promised at launch.

Survey data points the same way. Checkout.com research published in June 2026 found only about 3% of transactions currently involve AI agents, 24% of consumers say they will never delegate purchases to AI, and 27% trust no organisation at all to operate a shopping agent on their behalf [12].

Yet demand is forming underneath the hesitancy. A third of consumers in the same study expect at least 10% of their purchases to be AI-driven within a year, 57% would let an agent switch brands for better value, and shoppers would tolerate agents spending up to £177 per purchase without asking. The pattern resembles early mobile commerce: discovery moved to the new surface years before checkout did.

OpenAI has kept building through the stumble. OpenAI launched a dedicated shopping research feature on 24 November 2025, timed for the holidays, running on a version of GPT-5 mini post-trained for shopping tasks, and noted that "hundreds of millions" of people already use ChatGPT to find, understand and compare products [3].

Where do ChatGPT shopping clicks actually go?

Disproportionately to Google, homepages and big platforms rather than straight to product pages. That routing shapes who benefits from ChatGPT's growth.

Semrush analysed 17 months of US clickstream data, more than a billion records from October 2024 to February 2026, and found over 21% of ChatGPT's outbound referral traffic goes to Google, with online services, software and media outranking retail among destination categories [11]. A meaningful slice of ChatGPT's "shopping" value therefore arrives at retailers looking like ordinary Google traffic.

For attribution, this is a quiet nightmare. A shopper who researches in ChatGPT, clicks out to Google to sanity-check, then buys, shows up in analytics as search traffic. Adobe and Salesforce both only count referrals they can see, which means every figure in this article probably understates ChatGPT's real influence on purchases.

How are merchants responding to agentic commerce?

Faster than their infrastructure can follow, by their own admission. Nearly nine in ten merchants say they are preparing for agent-led shopping, whilst conceding that consumers will adopt it faster than they can get ready.

Checkout.com's June 2026 research found 89% of merchants actively preparing for agentic commerce and 72% acknowledging consumers will move faster than merchants can [12]. On the platform side, Digital Applied's round-up notes Shopify's self-reported, unaudited figures of 7x growth in AI-driven traffic and 11x growth in AI-attributed orders since January 2025 [13].

The key milestones so far:

The emerging merchant consensus is "discover in AI, buy on site". Retailers want ChatGPT's qualified traffic and its new-customer reach, but they want the checkout, the customer data and the upsell on their own property. Unless in-chat conversion improves dramatically, the protocol's future may be less about buying in the chat and more about structured product data feeding the chat.

How to read these numbers

Every figure here needs a health warning of some kind, and the warnings differ by source.

Adobe's numbers come from server-side analytics on over a trillion US retail visits, which is excellent for measuring what arrived, but it groups all generative AI referrers together and cannot see dark traffic, such as a ChatGPT user who reads a recommendation and later types the retailer's URL directly. Semrush's finding that a fifth of ChatGPT clicks route through Google means referral-based counts systematically understate ChatGPT's influence.

Salesforce's "AI-influenced" revenue is an attribution model built by a company that sells AI commerce software. It bundles on-site recommendation engines with ChatGPT referrals, so it is not a ChatGPT figure, and the 20% share should be read as an upper bound on a broad definition.

Profound's prompt studies measure real behaviour at serious scale, but its panel skews towards users whose activity it can observe, and intent classification is itself a model output with error bars nobody publishes.

Survey figures, such as Checkout.com's, report what people say, not what they do. Stated willingness to delegate purchases to an agent has historically run well ahead of actual delegation. And several-hundred-percent growth rates all share the same property: the base was tiny, so percentages flatter.

What this means going into 2027

ChatGPT is a discovery channel first and a checkout second, and that is unlikely to flip soon. With shopping triggering on 12.1% of unbranded prompts against 3.1% of branded ones, the battleground is being the answer to "what should I buy", not being findable by name.

Optimise the landing, not just the listing. AI-referred shoppers convert 31% better and are worth 254% more per visit than a year earlier. They arrive pre-qualified; the retailers who win will treat them as bottom-of-funnel traffic rather than top.

Expect the protocol to outlive the feature. Instant Checkout's first year disappointed, but the Agentic Commerce Protocol gives every merchant a standard way to expose structured product data to agents. Feeds, availability and pricing accuracy become AI visibility work.

Attribution will get worse before it gets better. With a fifth of ChatGPT clicks laundering through Google and dark traffic invisible entirely, boards should assume ChatGPT influence exceeds anything the dashboard shows.

Watch holiday 2026 as the real test. If Adobe's retail AI traffic growth holds anywhere near triple digits from the 2025 base, and in-chat conversion closes the gap Walmart described, agentic checkout revives. If not, "discover in AI, buy on site" hardens into the industry's settled model.

If you want to make sure your product data and landing pages are ready for ChatGPT's shoppers before the next holiday peak, you can book a call with me.