The Day AI Started Doing UK Buyers’ Shopping for Them —
What Changed on eBay, Amazon and Beyond

41% of consumers have already used AI platforms for product discovery. 56% of UK AI users now use it daily. The way UK buyers find products on Amazon, eBay, OnBuy and ASOS has changed structurally in the last 18 months — and most sellers haven’t adjusted to what that means. After 18 years on UK marketplaces, here’s what AI is actually doing to buyer behaviour, and what it isn’t.

Last month a customer messaged us through WhatsApp before placing an order. They wanted to confirm something about a wired earphone we list. The interesting part wasn’t the question — it was how they got to us. “ChatGPT recommended your store,” they said. Not Google. Not eBay’s internal search. Not a price comparison site. ChatGPT.

Eighteen years of selling on UK marketplaces and I’ve watched the buyer journey change three times. First when eBay opened to fixed-price listings alongside auctions. Then when Amazon’s logistics dominance shifted product discovery to the Buy Box. The third shift is happening now, in 2026, and it’s the most structural of the three — because for the first time, a substantial proportion of UK buyers are arriving at our listings with a recommendation already in hand, made by an AI system that read our content before they did.

This article isn’t about whether AI is good or bad for UK e-commerce. It’s about what’s actually changing in buyer behaviour on the marketplaces we work in — eBay UK, OnBuy, Amazon UK, our own site at maibo.uk — and how the shift is affecting both sellers and the buyers most haven’t noticed are being routed differently to the products they buy.

The Four Stages of the New UK Buyer Journey

The UK marketplace buyer journey in 2026 has reorganised into four stages, each with a different AI involvement. The change isn’t theoretical — it’s reflected in our analytics, our customer messages, and the kind of questions buyers ask before purchase.

Stage 1 — Problem framing in AI chat. The buyer no longer starts at a search engine. They start in ChatGPT, Gemini, or Perplexity, describing what they actually need rather than what they think they should search for. “I need wired earphones for office calls that block out background noise but cost under £20” — a query that wouldn’t have worked on Amazon’s search bar two years ago — is now the natural opening. The AI returns a shortlist of products with reasoning attached. The buyer takes that shortlist into the next stage.

Stage 2 — Verification on traditional channels. The buyer goes to eBay UK, Amazon UK, or OnBuy and searches for the specific product the AI named. Not for “wired earphones” generically — for the model the AI recommended. This stage is where marketplaces still play their original role: showing prices, seller histories, reviews, condition descriptions. But the discovery work has already been done. The buyer arrives with intent.

Stage 3 — Cross-checking with AI again. Before purchase, a growing share of buyers paste a listing URL or a product image back into an AI tool to ask “is this a fair price?” or “is this seller reliable?” The marketplace is no longer the final authority — the AI gets a second look. Buyers who weren’t price-checking three years ago are now AI-checking, which produces different decision patterns and different sensitivity to listings that don’t match what AI tools surface as fair.

Stage 4 — Purchase, still on the marketplace. The actual transaction is still happening on eBay UK, Amazon UK, OnBuy and direct sites — agentic commerce that fully completes purchases without a human in the loop is technically possible in 2026 but practically rare. eBay’s own user agreement now blocks LLM-driven bots from placing orders without human review. The purchase is human. The decision leading to it is increasingly not.

Across 40,000+ orders on our UK channels, the share arriving from this four-stage path has gone from invisible to substantial over the last 18 months. We can’t measure it precisely because the AI step happens off-platform. But the questions buyers ask in pre-purchase messages have shifted noticeably — they’re more specific, more product-aware, and more confident about what they want by the time they reach us.

“The marketplace used to be where buyers discovered products. In 2026 it’s increasingly where they verify what an AI tool has already discovered for them. That’s not a small change.”

— Halil Ibrahim Tutuncu, Maibo

What Each UK Marketplace Is Actually Doing About AI

The major UK marketplaces have responded to the AI shift differently — and the differences matter both for sellers and for buyers who use these platforms regularly.

1. 1
   eBay UK — image search, magical listing, agentic blockade

   eBay UK has integrated AI into discovery on the buyer side (image-based search lets you upload a photo and find similar products) and listing on the seller side (Magical Listing reduces listing time to roughly 30 seconds). But eBay has also drawn a sharp line against agentic commerce — its user agreement effective February 2026 specifically blocks “buy-for-me agents, LLM-driven bots, or any end-to-end flow that attempts to place orders without human review.” The position is consistent: AI to help the human buyer, not AI to replace them. For UK buyers this means eBay UK in 2026 still requires the human at the end of every transaction.

2. 2
   Amazon UK — Rufus folded into Alexa for Shopping

   In May 2026 Amazon retired the Rufus name and folded the assistant into Alexa for Shopping — a single AI agent integrated directly into the main Amazon search bar. The UK buyer typing into Amazon’s search now gets a conversational AI alongside the traditional product list. This is Amazon’s response to the threat that buyers might do their research in ChatGPT and never come to Amazon at all — bring the AI into Amazon’s own search so the discovery stays on-platform. For UK buyers it means the AI consultation step can now happen inside Amazon rather than outside it.

3. 3
   OnBuy — machine learning for surfacing, lower visibility but real impact

   OnBuy uses machine learning to optimise product listings and improve how items surface in search results — a less visible deployment than eBay’s or Amazon’s, but one that meaningfully affects which sellers get visibility. For UK buyers using OnBuy, the practical effect is that the listings surfaced are increasingly the ones the ML model has identified as matching buyer intent — not necessarily the cheapest, not necessarily the highest-feedback. For sellers on OnBuy, this rewards listing quality and complete item specifics more than aggressive pricing alone.

4. 4
   ASOS — conversational shopping with Microsoft Azure OpenAI

   ASOS deployed conversational shopping tools built on Microsoft Azure OpenAI that help UK buyers narrow choices, explore styles, and receive tailored recommendations based on browsing habits and current trends. The fashion category is where this style of AI assistance fits most naturally — buyers describe a vibe rather than a SKU, and the AI translates that into specific products. For the buyer this works well. For sellers it means winning placement increasingly depends on how well your listings describe what your products are like, not just what they are.

5. 5
   Etsy — “algotorial” curation, AI extending human taste

   Etsy’s approach combines human curation with large language models — what they call “algotorial curation.” The human-built product collections get extended by AI to keep recommendations visually coherent and relevant to each shopper’s taste. This is a different model from the others: AI as taste amplifier rather than agent. For UK Etsy buyers it means the recommendations they see feel more curated, more handpicked, even though most of the work is now algorithmic. The trust pattern is older — Etsy buyers already trusted the curation idea. AI just scaled it.

The UK Buyer Behaviour Patterns That Are Genuinely New

Beyond the marketplace-level changes, the day-to-day patterns of how UK buyers actually shop have shifted in five specific ways since AI tools became part of the default research flow. These are observations from our customer messages and analytics — not predictions.

Higher question specificity at first contact. Buyers messaging us before purchase increasingly ask precise, model-specific, technically detailed questions rather than open-ended ones. “Does the Sennheiser CX 300 II you stock have the original 3.5mm jack or a USB-C adapter?” is a 2026 question. “What earphones do you recommend?” was a 2022 question. The AI has done the broad research for them; they want the specific verification we can provide that the AI cannot.

Lower price sensitivity at the lowest end. A surprising pattern from 2025 onwards: the cheapest listings on a search results page are converting less reliably than mid-tier listings. The AI tools that route buyers to specific products tend to filter out aggressive discounts as suspicious — implicitly or explicitly. The £6.99 “Sennheiser CX 300 II” listing that ranks first on a price-sorted search is increasingly flagged by AI tools as a likely counterfeit before the buyer ever clicks it. This protects buyers and changes the price floor of the listings that actually sell.

More cross-platform research per purchase. The same buyer now checks eBay UK, Amazon UK, and often a direct retailer before purchasing, with the AI sometimes orchestrating the comparison. The buyer who used to buy on whichever platform they happened to be searching now compares more deliberately — and the seller who appears on multiple platforms with consistent pricing and presentation has an advantage they didn’t need before.

Reviews matter differently. AI tools weight reviews into their recommendations, but the way they weight them isn’t the same as the way buyers used to. Detailed, specific, photo-supported reviews carry far more weight in AI summarisation than generic five-star reviews — which AI tools increasingly recognise as low signal. For sellers this means a smaller number of high-quality reviews now outperforms a larger number of generic ones in ways that simply weren’t true two years ago.

Trust in agentic checkout remains low. Despite all the AI involvement in discovery and verification, UK buyers are still not comfortable letting AI complete the actual purchase autonomously. Mintel’s 2026 research found that while 45% of UK consumers find AI features appealing, half believe its risks outweigh the benefits — and the discomfort sharpens specifically when AI makes important decisions without explicit human control. UK buyers in 2026 want AI to research for them. They don’t want AI to spend their money for them. This is a critical structural distinction.

Across the categories we sell — wired earphones, gaming peripherals, charging accessories, storage — these five patterns show up most strongly in the higher-consideration purchases. Lower-cost commodity items still follow more traditional buyer flows. AI’s impact has been heaviest where the buyer was already doing research; it’s been lighter where the buyer was already making impulse decisions.

What This Means for UK Buyers Right Now

The buyer side of this shift deserves direct attention. UK consumers using AI tools for shopping in 2026 are getting some clear advantages — and exposing themselves to some specific risks that the traditional marketplace flow didn’t carry.

The clear advantages: better product matching to actual needs (because the AI can parse a natural-language description that no search box accepts), faster filtering of obviously fake listings (because AI tools flag aggressive price anomalies), and access to product knowledge that previously required hours of forum reading. For higher-consideration purchases — electronics, appliances, technical products — this is a meaningful improvement to buyer outcomes.

The specific risks: AI tools can confidently recommend products that don’t exist in stock, prices that aren’t current, sellers that don’t ship to the UK. AI tools can be trained on outdated information that misses recent quality issues. AI tools can be gamed by sellers who optimise their content specifically for AI consumption — meaning the recommendation you get may reflect SEO investment more than product quality. The UK buyer in 2026 who trusts an AI shortlist without verifying on the marketplace is making a different mistake to the one who trusts the cheapest listing without thinking — but it’s still a mistake.

The practical recommendation: use AI for the research stage, use the marketplace for the verification stage, and use direct contact with the seller for anything where stock authenticity matters. For high-risk categories — discontinued products, branded electronics where counterfeits are common, anything where the product photo could be misleading — a pre-purchase message to the seller is worth more than any AI recommendation.