I've had four separate conversations this month with founders convinced they need an "AI search strategy." Usually this means they want to know if they should be worried, or if there's some new channel they're supposed to be dumping budget into before their competitors get there first.

Shopify just put out data that's going to accelerate that panic. AI-referred sessions to Shopify stores grew 197% year over year in Q2. On spec-heavy products, shoppers coming from AI converted at roughly double the rate of shoppers coming from organic search. Those are real numbers, not vibes, and I understand why they're setting off alarms.

Here's my actual take: the data is useful, the panic is misplaced, and the thing worth doing about it isn't a new strategy at all. It's a chore most brands have been skipping for years.

The framing I keep seeing is AI search versus organic search, like you have to pick a side or build two separate playbooks. Shopify's own data undercuts that framing. Organic search grew too, 12% year over year, and it's still sending more total traffic than every AI platform combined. Nothing's being replaced. Search is just spreading across more surfaces than it used to.

What actually explains the conversion gap is more mundane than "AI is smarter." When AI systems pulled from clean, structured product data instead of scraped or third-party feeds, the shoppers they referred converted twice as well. Twice. Same AI, same platform, same shopper intent. The only variable was whether the underlying product data was actually usable.

That's not an AI story. That's a data hygiene story wearing an AI costume. Shopify's own team described the underlying shift as buyer journey compression, where "AI search collapses discovery and consideration into a single conversation." Fine, but a compressed conversation still needs accurate inputs. Garbage in, garbage recommended.

I've been doing retention and lifecycle work long enough to know which brands have their product data actually in order, and it's a short list. Most catalogs I look into are held together with inconsistent titles, thin descriptions written for a person skimming a grid, missing attributes, and metafields nobody's touched since launch. It's worked fine for years because a human shopper could still figure out what they were looking at even through the mess.

AI can't do that. It doesn't infer the way a human browsing your site does. It reads whatever structured signal you've given it, and if that signal is thin or inconsistent, it either gets your product wrong or skips it for a competitor whose data is cleaner. Shopify's numbers are just the receipt for that.

So the founders asking me about AI search strategy are usually asking the wrong question. The right question is whether a shopper's actual language, the way they'd describe what they need out loud, shows up anywhere in your product data. Compatibility, use case, materials, who it's for, what it's not for. Most catalogs answer none of that. They answer "what category is this" and stop there.

This is the part that makes me confident this stays an advantage instead of turning into table stakes anytime soon. Catalog cleanup is unglamorous. It doesn't have a launch date, it doesn't make for a good slide, and nobody gets excited pitching "we restructured our metafields" the way they get excited pitching a new ad channel or an influencer push. It's the kind of work that only pays off in aggregate, quietly, over months, which makes it exactly the kind of work that gets deprioritized every single time something louder shows up on the roadmap.

I'd put money on most brands reading about this Shopify data spending their energy trying to figure out how to "show up" in ChatGPT or Perplexity, chasing some imagined AI SEO trick, instead of doing the unglamorous version: sitting down with actual customer questions, from support tickets, on-site search, chat transcripts, and turning that language into structured product attributes. That's the whole unlock. It's not sexy. It's also the only part of this that's actually in your control.

If you run a brand doing real revenue and you want to take this seriously, skip the strategy deck. Go look at ten of your best-selling PDPs right now and ask honestly whether they'd answer a specific, spec-driven question a shopper actually has, not the question your product team assumed they'd ask. Then go look at your metafields and see how much of that answer is sitting anywhere a machine could read it, versus buried in a paragraph a human has to parse.

If the answer is "not much," you don't have an AI problem. You have a homework problem you've been putting off since before AI search existed. AI just made the cost of skipping it visible for the first time.

That's the whole opportunity right now. Not a new channel. Just finally doing the boring thing while most of your competitors are busy chasing the shiny one.