AI Is Now a Sales Channel: What ChatGPT, Perplexity, and Google AI Mode Mean for Your Ecommerce Brand
Twelve months ago, shoppers who landed on an ecommerce site from an AI assistant converted about 38% worse than everyone else. As of March 2026, that same traffic converts roughly 42% better. That's an 80-point swing in a single year, and most brands haven't changed a thing about how they show up to it.
AI has stopped being a search story and become an acquisition story. ChatGPT, Perplexity, Google's AI Mode, and Amazon's Rufus are now places where people decide what to buy, not just where they ask questions. We run Amazon, Google, and Meta for more than 100 brands, and across that portfolio we can see the same thing surface on marketplace listings and DTC sites at the same time: a new class of pre-qualified buyers arriving from machines that already did the comparison shopping for them. So let's get specific about what's actually happening, what's protected, what's bleeding, and what to do about it.
AI Quietly Became a Real Acquisition Channel
The volume isn't a rounding error anymore. AI-referred traffic to US retail and ecommerce sites grew roughly 393% year over year in the first quarter of 2026, according to Adobe, and retail AI-referred visits in May hit their highest share since Adobe began tracking the channel in late 2024. On Shopify storefronts, referral sessions from AI chatbots grew more than 8x year over year. It's a small base. But the slope is the whole story.
What makes this channel different is the quality of the visitor. AI-referred shoppers generate about 37% more revenue per visit and spend roughly 48% longer on site. More than half of AI-referred sessions start on a product detail page, versus about 20% for organic search. The assistant doesn't drop someone on your homepage to wander around. It sends them straight to the product it already decided fits their question. By the time they click, the comparison is finished and they're close to buying. Orders attributed to AI search carry around 14% higher average order values than organic.
The Three AI Surfaces Where Shoppers Find Products Now
Product discovery used to mean a Google search and a row of blue links. It's fragmenting fast across three different kinds of AI surface, and each one rewards different work.
Google AI Overviews and AI Mode
Google now answers a growing share of searches before anyone clicks. AI Overviews showed up on about 14% of shopping queries by early 2026, a 5.6x jump in four months across an analysis of 20.9 million shopping keywords. AI Mode, the full conversational interface, went out to all US users in March 2026, and inside it up to 93% of searches end without a single click to an outside site. The brands that survive this aren't the ones ranking number one. They're the ones cited inside the answer. Google's own data shows brands cited in an AI Overview earn about 35% more organic clicks and 91% more paid clicks than brands left out of the same answer.
LLM Shopping Assistants: ChatGPT and Perplexity
People now ask ChatGPT and Perplexity to recommend products the way they'd ask a knowledgeable friend. ChatGPT's product results are organic and unsponsored, ranked on relevance signals like availability, price, quality, and whether you're the primary seller. You can't buy your way in. Perplexity has pushed further into in-chat browsing and checkout through its PayPal integration. The conversion data backs the behavior: ChatGPT traffic has been measured converting about 31% higher than non-branded organic search, and some assistants convert dramatically higher than that on a tiny share of total volume.
One honest correction, because the hype got ahead of reality here. ChatGPT's in-chat Instant Checkout, launched with Shopify and Etsy merchants in late 2025, was pulled in early 2026 after fewer than 15 Shopify merchants ever went live. The lesson isn't that AI shopping fizzled. It's that the discovery layer, being the brand the assistant recommends, is what matters right now. The checkout mechanics will keep changing for years. Being inside the recommendation won't stop mattering.
Amazon Rufus
On Amazon, the AI surface is Rufus. It went from roughly 3% of sessions to about 38% over the course of 2025, drove an estimated $12B in incremental annualized sales, and now sits under the "Alexa for Shopping" umbrella reaching around 300 million customers. This matters because Rufus doesn't read your listing the way the old keyword-match algorithm did. It runs on COSMO, a knowledge graph that maps products to use cases, attributes, and customer intents. Listings with complete structured attributes, real material, use-case, and certification data, beat keyword-stuffed listings for getting surfaced. Across brands we've measured, Rufus-surfaced product pages convert meaningfully higher than the same ASINs reached through traditional search.
| AI surface | How shoppers use it | What gets you surfaced |
|---|---|---|
| Google AI Overviews / AI Mode | Ask a question, read a synthesized answer, rarely click out | Being cited as a source: answer-first content, structured data, topical authority |
| ChatGPT / Perplexity | Ask for a recommendation, get a short ranked shortlist | Strong availability, competitive price, real reviews, clean machine-readable product data |
| Amazon Rufus | Ask Rufus inside the app which product fits a specific need | Complete structured listing attributes mapped to real use cases (COSMO) |
Why AI Traffic Converts Better, and What That Changes
The reversal from converting 38% worse to 42% better in a year isn't magic. It's selection. An AI assistant absorbs the messy top of the funnel, the "which one is best for X" research, and only forwards the shopper once it has a candidate in mind. You're not paying to educate a cold browser anymore. You're receiving a warm buyer near the bottom of the funnel. That's why the revenue-per-visit and order-value numbers run higher than your other channels.
The strategic consequence is uncomfortable for anyone who built their playbook on rank tracking. Only about 38% of pages cited in AI Overviews rank in the traditional top 10. Citation runs on semantic passage retrieval, not position. So the metric that mattered for fifteen years, where you rank, is now a weak proxy for the metric that actually pays, whether you get cited and recommended. Brands still optimizing only for blue-link rank are optimizing for a screen fewer and fewer people look at.
What's Protected, and What's Quietly Bleeding
This is where most coverage either panics or pretends nothing changed. The truth sits in between, and it depends entirely on the query.
Transactional and product-name queries are relatively protected for now. Pure ecommerce queries carry an AI Overview only around 13 to 14% of the time, because when someone searches a specific product they want to buy it, not read an essay about it. Informational shopping queries are a different story. The "best [product]" style searches carry AI Overview presence around 83%. That's the traffic that's bleeding. The top-of-funnel research content that used to pull people onto your site is increasingly answered before they ever leave Google. Across query types, AI Overviews have been measured cutting organic click-through by anywhere from 15% to 46%.
So the honest read isn't "SEO is dead." It's that discovery is splitting in two. Bottom-funnel intent still flows to sites and listings. Top-funnel research is getting absorbed into AI answers, and the only way to capture value there is to be the brand the answer cites. Uncited brands don't get a warning. They just watch a slow, quiet decline in the research traffic that used to feed everything downstream, and they usually blame it on the wrong thing.
How to Get Your Brand Cited Across AI Surfaces
The discipline for this has a name now: generative engine optimization, or GEO. It overlaps with good SEO but optimizes for a different outcome. Classic SEO fought to rank among ten links. GEO fights to be one of the two to seven sources an AI engine cites in a single answer. A few things move the needle more than the rest, and they're the backbone of how we approach website SEO now:
- Answer-first content. Lead a page or section with a direct, standalone answer, then add the context. AI engines retrieve passages that resolve a question cleanly. Burying the answer under 600 words of preamble hides it from the retriever.
- Clean structured data. JSON-LD for Product, Offer, AggregateRating, and BreadcrumbList is the machine-readable backbone. Use ISO 8601 dates, prices as numbers rather than formatted strings, and explicit rating and review values. This lowers the chance an AI misreads or hallucinates your product details.
- Original data and clear authority. Engines preferentially cite content with proprietary research, real numbers, and a recognizable entity behind it. Recency is weighted, so stale pages lose citations to fresher ones over time.
- Complete attributes on marketplace listings. On Amazon specifically, fill every structured attribute field with true material, use-case, and compatibility data so Rufus and COSMO can map your product to the intents people actually ask about. We build this into listing optimization directly, and changes usually take four to eight weeks to show.
The measurement shift matters as much as the tactics. You can't manage citation share with a rank tracker. Start tracking how often each AI platform cites or recommends your brand, and your share of voice against competitors inside those answers. Most brands haven't started. The ones that move in 2026 become the brands these engines cite by default in 2027 and 2028, when the behavior is fully mainstream and the cost of catching up is far higher.
The Cross-Channel View Most Brands Can't See
Sitting where we sit, running Amazon and DTC for the same brands at once, the most useful thing about this shift is that it's happening everywhere simultaneously. The brand getting recommended by ChatGPT for its category is usually the same brand Rufus surfaces on Amazon and the same one Google cites in an AI Overview. The signals that earn those placements, clean structured data, genuine reviews, real authority, and consistent product information, are shared across surfaces. Fix them in one place and the benefit compounds across every channel at once.
That's the part a single-channel shop can't see. An Amazon-only agency tunes your listing for Rufus and never looks at your DTC schema. A pure SEO shop optimizes your site for AI Overviews and never touches your marketplace attributes. We model this across the whole stack because the buyer doesn't think in channels, and the AI engines increasingly don't either. If you want a team that runs your whole ecommerce stack with that cross-channel data in a single view, that's the work we do every day.
The Bottom Line
AI isn't replacing your sales channels. It's becoming one, and it's already sending some of the best-converting traffic you'll see this year. The brands that win it aren't the ones with the best rank. They're the ones the machines recommend. Figure out whether AI engines are citing you or your competitors right now, then close the gap before the behavior fully mainstreams. If you'd rather have an operator audit where you stand across Google, the LLMs, and Amazon, tell us what you're selling and we'll show you who's getting cited.