Episode 06 — Will AI replace traditional eCommerce search?

HOST A:  Last episode we ended on a question I want to actually answer today, or at least take seriously: is AI chat search going to replace traditional eCommerce search entirely?

HOST B:  That feels like the kind of question people ask in a way that’s designed to get a dramatic yes or no answer, and I’m not sure the real answer is that clean.

HOST A:  I don’t think it is either, but I think the data we covered last time makes the direction of travel pretty unambiguous. Nearly a billion prompts a day to AI assistants. 1.3 billion monthly referrals out to actual websites. Seventy-four percent of eCommerce businesses already adjusting their content to be AI chat-friendly, another twenty-four percent planning to.

HOST B:  So even if it’s not a full replacement, it’s clearly not a fringe behavior anymore either.

HOST A:  Right, and I think the more useful question than “replace or not” is: what is traditional search actually good at, what is AI chat search actually good at, and where does each one win.

HOST B:  Let’s start with traditional search, since that’s the thing everyone already understands intuitively. You type keywords, you get a ranked list of links, you click through and compare.

HOST A:  Right, and that model is genuinely good at certain things. It’s good when you already roughly know what you want and you’re comparison shopping across many options. It’s good for browsing — scrolling through a category page, seeing what’s out there. It’s good when visual comparison matters, like looking at a grid of furniture or clothing.

HOST B:  What’s AI chat search actually better at, mechanically?

HOST A:  It’s better when the shopper has a more complex or specific need that’s hard to express as a keyword string. Think about the difference between typing “waterproof hiking boots” into a search bar, versus asking an AI assistant “I need boots for a rainy week-long trek in Scotland, I have wide feet, and I want something under 150 dollars.” That second query is basically impossible to express well as search keywords, but it’s a completely natural question to ask a conversational assistant.

HOST B:  So it’s less about replacing search entirely, and more about capturing the queries that search was always bad at handling.

HOST A:  That’s exactly how I’d frame it. And there’s a trust dimension here too. When an AI assistant gives you a direct recommendation instead of ten links to evaluate yourself, it’s doing real synthesis work on your behalf — it read the reviews, it compared the specs, it’s handing you a conclusion. For a lot of purchase decisions, especially lower-stakes ones, people are going to take that.

HOST B:  Let’s bring accessibility back into this, because I don’t want this episode to drift into being a general AI commentary episode and lose the thread of the season.

HOST A:  Fair, let’s anchor it. The report’s framing here is precise: structured markup, alt text, and accessible FAQs allow products to surface in conversational recommendations. Without them, your catalog risks becoming invisible to both people and machines.

HOST B:  That phrase again — invisible to both people and machines.

HOST A:  And I think this episode is where that phrase really earns its weight, because we’re specifically talking about the mechanism by which an AI assistant decides what to recommend. It’s not magic. It’s reading your structured data, your markup, your descriptions, the same underlying content layer that determines whether a screen reader can make sense of your page.

HOST B:  So a retailer with genuinely inaccessible product pages isn’t just at legal risk and losing conversion at checkout, which we covered in episodes two and three — they’re also going to be functionally invisible in this new discovery channel.

HOST A:  Right, and I think that’s the part that should change how leadership teams prioritize this. If you frame accessibility purely as a legal or ethical issue, it’s easy for it to get deprioritized when budget is tight, because the consequences feel deferred and uncertain. But if you frame it as “this directly determines whether AI assistants can recommend your products at all,” that’s a much harder thing to deprioritize, because the business impact is immediate and measurable.

HOST B:  Let’s talk about what “AI can’t see you” actually looks like in practice for a retailer, because I think it’s worth making concrete.

HOST A:  Sure. Say your product descriptions live entirely inside an image — like a lot of fashion brands do with sizing charts or care instructions baked into a graphic. An AI assistant reading your page can’t extract that information at all if there’s no alt text or accompanying text content. Or say your product specifications are scattered inconsistently across the page with no semantic structure — no proper headings, no structured data markup — so a system trying to parse “what is the battery life of this product” simply can’t reliably find the answer, even though a human scrolling the page could eventually spot it.

HOST B:  So the AI assistant either gives an inaccurate answer, which is bad for you, or it just skips your product and recommends a competitor’s instead, which is also bad for you.

HOST A:  Those are genuinely the two outcomes. There isn’t a third option where the AI assistant “tries harder” to understand a badly structured page. It moves on.

HOST B:  Let’s talk about Amazon’s Rufus specifically, since it’s been mentioned a couple of times now and I think it’s a useful concrete example for listeners who haven’t used it.

HOST A:  Rufus is Amazon’s own AI shopping assistant, built directly into their app and site. A shopper can ask it things like “what’s a good gift for someone who likes hiking” or “does this blender crush ice,” and it responds conversationally, often pulling from product listings and reviews to answer.

HOST B:  Which means even within Amazon’s own ecosystem — arguably the most search-dominant retail platform that exists — there’s now a layer of AI-mediated discovery sitting on top of, or alongside, traditional search and browse.

HOST A:  Exactly, and that tells you something important: this isn’t only happening because shoppers are leaving retail platforms for ChatGPT. It’s also happening inside the platforms retailers already sell on. There’s no opting out of this shift by just staying focused on your existing channels.

HOST B:  Let’s talk about what a sensible response actually looks like for a retailer right now, because I think “redo your entire SEO strategy” is overwhelming and probably not even the right frame.

HOST A:  I don’t think it’s an either-or choice between optimizing for traditional search and optimizing for AI chat discovery. The good news, genuinely, is that a lot of the same foundational work serves both. Clean, well-structured semantic HTML helps traditional search engines crawl and rank your site, and it helps AI assistants parse your content. Accurate, detailed alt text helps screen reader users, and it’s also literally the data an AI system uses to understand an image. Structured data markup — schema for products, prices, availability — has been a best practice for traditional SEO for years, and it’s exactly what AI assistants rely on too.

HOST B:  So it’s less “build a parallel AI strategy” and more “do the structural content work properly, and you’re positioned for both.”

HOST A:  That’s the practical takeaway. The teams that are going to struggle here aren’t the ones who need to invent some brand-new AI-specific strategy from scratch. They’re the ones who’ve been cutting corners on basic content structure and accessibility for years, and now that debt is becoming visible in a new, more consequential way.

HOST B:  Let’s close with a direct answer to the episode title, since I did promise we’d take the question seriously. Will AI replace traditional eCommerce search?

HOST A:  I’d say: not replace, but rapidly become a parallel, and in some categories dominant, discovery channel that retailers cannot afford to treat as optional. The shoppers using it aren’t a fringe group anymore — we’re talking close to a billion daily prompts and over a billion monthly referrals. And the underlying requirement to succeed in that channel — structured, accessible, well-described content — is the same requirement that protects you from lawsuits and recovers abandoned cart revenue.

HOST B:  Which brings us right back to the season’s core idea. Different business problem, same underlying fix.

HOST A:  Exactly. Next episode, we’re shifting into regulation again, but going further than just litigation risk — we’re talking about the European Accessibility Act specifically, why the report calls it a clock that’s already started, and what seventy-one percent of eCommerce leaders are already grappling with as a result.

HOST B:  See you there.

HOST A:  Thanks for listening.