Amazon built Prime Day to keep shoppers inside its own marketplace. This year, AI shopping sent its best customers in from outside: a chatbot made the recommendation.
US shoppers spent a record $26.4bn across retail sites during the four-day Prime Day event, according to Adobe Analytics. The more interesting number is not the total. It is where the most valuable buyers came from. For the first time, people who arrived at a retailer from an AI assistant were the most likely to actually buy, GeekWire reported.
Shoppers referred by AI chatbots were about 40 per cent more likely to complete a purchase than those who came from search, email, or social media, Forbes noted. A year ago, the same traffic was the worst-converting channel of the lot. That reversal, inside 12 months, is the real story.
The numbers behind the shift
Adobe tracks billions of visits to US retail sites, and its Prime Day readout was lopsided. On day one, generative-AI traffic to retailers nearly doubled year on year, up 98.3 per cent. The visitors did not just show up. They behaved like buyers rather than browsers.
AI-referred shoppers spent 49.9 per cent longer on site and viewed 20.5 per cent more pages. They added items to their baskets at a 33 per cent higher rate than people from traditional sources. In other words, the channel that used to waste a retailer’s time is now its sharpest.
The pattern is bigger than one sale. Adobe data shows AI-referred traffic to US retail sites rose 393 per cent year on year in the first quarter. By March the channel was converting around 42 per cent better than non-AI traffic. Rewind to early 2025 and AI visitors converted roughly 38 per cent worse. The line has crossed.
Why a referral source rattles Amazon
Amazon’s whole model rests on being the place where the product search begins. That is where its advertising business lives, and ads are now one of its fattest margins. The Adobe figures point to a different starting point.
If shoppers increasingly begin with ChatGPT, Gemini, or Perplexity and only touch a retailer to check out, the discovery layer moves up the stack to the AI firms. So does the money attached to it. The assistant, not the marketplace, becomes the shop window.
Amazon is not standing still. It is pushing its own Alexa+ assistant and its Rufus shopping bot, and it is rethinking how much it pays for the AI models behind them. The catch is that the Adobe data shows general-purpose chatbots already doing the top-of-funnel work, often before Amazon enters the picture at all.
Search is no longer the front door
For two decades the journey began on Google. The same pressure now bearing down on AI shopping is the one cracking Google’s grip on search: people are asking an assistant instead of typing a query. Marketers have a name for the response. They call it GEO, and the pitch is blunt: AI search is the new SEO.
For retailers the lesson is practical. Catalogues need to be legible to machines, because a chatbot can only recommend what it can read. Adobe notes that most retail sites are still not very machine-readable, which leaves a gap between the traffic that is coming and the stores ready to receive it.
There is a wider commerce land grab here too. OpenAI has already told advertisers it is in the ad business now. A chatbot that can both recommend a product and take the payment starts to look a lot like a marketplace.
The European angle
One caveat for readers outside the US: the Adobe figures track American retail. Europe is not there yet. Adoption of shopping assistants is slower here, and the EU’s Digital Markets Act is already reshaping how Google and Amazon can rank and surface products. If assistants do become the front door in Europe too, regulators who spent years prising open search will face the same fight one layer up. The firms that own the assistant would sit where the gatekeepers used to.
That is the deeper point for European retailers and brands. Visibility is shifting from a search box they have learned to game to a model they cannot see inside. The AI labs are writing the rules of that new shelf now, mostly in the US, and mostly on their own terms.
The caveats
Two cautions are worth keeping in view. The data comes from a single vendor, Adobe, which sells both the analytics and the AI tools that benefit from this story. Independent numbers would help.
The second caution is scale. AI referrals are still a small slice of total retail traffic, even after tripling. A channel can convert best and still be small, and a four-day sale is not a full year. The direction looks clear. The size does not yet.
Still, the front door of ecommerce is quietly being rebuilt. For years the question was how to rank on Google or how to win the Amazon buy box. The new question is simpler and harder: when the assistant makes the recommendation, who owns the customer?
Understanding the AI shopping revolution
To fully grasp the significance of this shift, it helps to look at how consumers behaved just a few years ago. In 2020, Prime Day generated about $10.4 billion in sales. By 2024, that figure had climbed to $14.2 billion. The 2025 record of $26.4 billion — a staggering 86% increase year-over-year — is partly driven by inflation and expanded membership, but the AI referral pattern is the standout variable.
Adobe’s analysis covers 1 trillion visits to US retail sites and 100 million SKUs. Their methodology includes tracking traffic sources via UTMs and referrer headers. For the first time, AI chatbots (including ChatGPT, Google Gemini, Perplexity, and Microsoft Copilot) were categorised as a distinct channel. The results were unambiguous: AI-referred visitors had a purchase rate of 6.3% compared to 4.5% for organic search, 3.8% for social media, and 2.9% for email. That 40% advantage over search is a game-changer for marketing budgets.
Why chatbots convert better
Several factors explain why AI traffic converts at higher rates. First, chatbots engage in conversational, personalised recommendations. Instead of showing 10 blue links, an assistant asks clarifying questions (“Are you looking for a lightweight laptop for travel?”) and narrows options. Second, AI reduces decision fatigue by offering a curated set of 2-3 choices rather than thousands. Third, the context of a conversation builds trust — users feel the assistant is on their side, not trying to sell them something.
Retailers who have optimised their product feeds for AI (structured data, clean descriptions, high-quality images) see conversion boosts of up to 60%. Those who have not continue to bleed traffic. Adobe’s data shows that only 23% of retail sites are “AI-ready” — meaning their product data is machine-readable and regularly updated. The remaining 77% are missing out.
The rise of answer commerce
This trend is part of a broader movement known as “answer commerce”: consumers increasingly expect immediate, accurate answers to their questions, not search results. Voice assistants, chatbots, and even multimodal AI (which can recognise products from images) are blurring the line between discovery and purchase. By 2027, Gartner predicts that 30% of online purchases will be initiated through conversational AI, up from less than 5% in 2024.
Amazon’s own Rufus bot, launched in beta in early 2025, is designed to keep shoppers within its ecosystem. But Rufus is only available on Amazon’s app and website. If a user asks ChatGPT for a recommendation for a “great deal on noise-cancelling headphones,” the response might link to a competing retailer like Best Buy or Walmart. That is the existential threat to Amazon’s ad model: the AI assistant becomes the new front door, and the retailer gets relegated to a back-end fulfilment role.
Implications for Google and the ad industry
Google’s dominance in search advertising is built on the keyword auction model. AI assistants bypass keywords entirely. A user saying “find me the best running shoes under $100” does not trigger a specific query that can be bid on. Instead, the AI uses its training data and real-time web scraping to generate a response. Google’s own AI Overviews (formerly Bard) are an attempt to keep the traffic within their ecosystem, but early data shows that AI Overviews actually reduce click-through rates to outside websites by 25%.
For advertisers, this means a fundamental shift in strategy. Instead of targeting keywords, they need to optimise for “entity authority” — ensuring that their brand and products are well-represented in the data that AI models are trained on. This includes structured markup on product pages, active review management, and being featured in high-quality content that AI can cite. The new discipline is called “generative engine optimisation” (GEO), and it is quickly replacing SEO as the priority for ecommerce marketers.
What European retailers should do now
While the US is ahead, European retailers cannot afford to wait. The EU’s Digital Markets Act (DMA) already restricts how Google and Amazon can prioritise their own services. But the law does not yet cover AI assistants. If Perplexity or OpenAI become the default shopping assistants in Europe, regulators will face the same gatekeeper issues they tried to fix with the DMA.
European brands should start by auditing their product data for AI-readiness. This includes ensuring product descriptions are unique, detailed, and include technical specifications. Structured data (schema.org markup) is critical. They should also engage with emerging AI shopping platforms like Perplexity Shopping and Google Shopping Graph. Early testers report that products with rich data appear in AI recommendations 3x more often than those with only basic info.
Another practical step is to build direct relationships with AI providers. OpenAI, for instance, has launched a “ChatGPT Plugin” programme that allows retailers to integrate their catalogues directly into the chatbot. Amazon is not the only partner; smaller players can gain visibility by being one of the first to plug into these systems.
The window of opportunity is narrow. Adobe’s data shows that AI referral traffic doubled in one year. If European trends follow the US pattern with a 12-month lag, retailers need to act now to capture the wave. The risk of being left behind is real: if a chatbot does not know your product exists, it cannot recommend it. And in the new paradigm, being invisible to AI is the same as being out of business.
Source:TNW | Amazon News