AI Traffic Is Becoming a Real Retail Channel. Here's What John Lewis Is Doing About It.

Reuters reported this month that the share of John Lewis's website traffic arriving through AI tools, ChatGPT, Gemini and the like, rose from 0.3% to 2.5% over the past year. Managing director Peter Ruis called the growth "exponential" and said it's happening across all age groups. John Lewis's response is a new content studio at its Oxford Street flagship, with influencers recording daily, plus original video including a celebrity mini-series, aimed at getting its products surfaced by AI tools rather than found by people typing into a search box. The company is framing this against what it calls a difficult economy, with discretionary spend under pressure and every channel that can bring in a sale worth pursuing.

Start with the honest read of that number. 2.5% of traffic is small. It is nowhere near the scale of organic search, paid social or email for a retailer John Lewis's size, and John Lewis hasn't published how it defines an AI-referred visit, so the figure is directional rather than audited. If this were the only data point, a leadership team would be entitled to file it under "watch, don't act."

What earns it attention is the curve behind it: an eightfold increase from a standing start in twelve months, which is what exposure looks like early, before it's large enough to force a decision. I've watched retail leadership teams handle structural shifts like this the same way more than once, waiting for the number to become impossible to ignore before agreeing it deserves a response. By the time AI-referred traffic is 15 to 20% of a major retailer's mix, whoever got their product data right first will already have banked the advantage of being findable while everyone else was still debating whether the channel was real.

The shopper who never sees your shop window

An AI agent shopping on someone's behalf doesn't browse the way a person does. It doesn't scroll a category page, notice a display, or respond to a headline on a landing page. It queries whatever structured data and product text exists, reasons over it against the shopper's stated need, and returns a shortlist or a single recommendation. If your product data is thin, inconsistent, or spread unevenly across your catalogue, the agent has nothing to reason with, and it will surface a competitor's product instead, regardless of how strong your physical merchandising or brand story is in a shop or on a well-designed product page.

That puts a premium on something retailers have always underinvested in: complete, accurate, consistently structured product data at scale. I wrote about a related test in the Migros product descriptions case study, where AI-generated copy lifted conversion by nearly a quarter once it was in front of a human shopper. The John Lewis story sits one step earlier in the funnel, at the point where an agent decides whether to find and recommend the product before a human shopper ever reaches the page that copy sits on.

Brand equity when an agent is choosing for you

The other exposure worth naming is what happens to brand equity when an intermediary is doing the comparing. Retail brands are built through a mix of visual merchandising, staff interaction, and emotional storytelling that assumes a human is present to receive it. An agent compressing a shopping decision down to price, specification, stock availability and delivery time reads none of that unless it's been translated into text and structured data the agent can parse. A retailer with a strong brand and thin product data risks losing to a competitor with a weaker brand and a cleaner feed, because the agent never got to see the thing that made the strong brand worth choosing.

That's a genuinely uncomfortable trade-off for a retailer like John Lewis, whose brand has historically been carried by trust, service, and in-store experience. A content studio producing video and influencer content is a reasonable instinct: more content, more places to be found. Whether that specific content is legible to the systems doing the recommending is a separate question, and one neither John Lewis nor Reuters has answered. Video optimised for a human scrolling a feed and product data structured for an agent comparing SKUs are not automatically the same investment.

Where this sits for a mid-size retailer

None of this needs to be decided in a boardroom on instinct, and it's exactly the kind of question the playbook's Section 2: Strategic Context is built for. The Competitive AI Exposure Assessment (Tool 02b) gives a structured way to check where your business and your competitors actually stand in AI-agent visibility today, rather than debating it from impressions. Treat it the same way you'd audit paid search performance or marketplace presence: a channel that either has traffic moving through it or doesn't, checkable now.

Once you know where you stand, the AI Strategic Positioning Tool (Tool 02d) forces the sharper decision: is this an optimisation problem, cleaning up product feeds and structured data within the operating model you already have, or a transformation problem, a genuinely new distribution channel with its own rules for what gets surfaced and no settled playbook yet for winning at it. Most mid-size retailers will land on optimisation for now: better data, better feeds, no reorganisation required. That's a legitimate answer. What isn't legitimate is drifting into one by default, leaving the choice to whichever team happens to own the product catalogue.

The lesson from John Lewis's 2.5% is to find out, this quarter, whether your own products would be found by the shopper who never visits your website at all.

Sources

• UK's John Lewis looks to harness AI agent shopping in 'difficult economy' (Reuters, via WKZO, September 2026)
• UK's John Lewis looks to harness AI agent shopping in 'difficult economy' (Reuters, via AOL, September 2026)