Agentic commerce won't save a weak product foundation

There's a new pitch making the rounds, and if you sell products online you've probably heard a version of it. Hand over your product data, let an AI enrichment layer rewrite every title, description, and attribute, and you'll dominate agentic commerce. Just give the tool the keys.

It sounds efficient, and that's the problem. Enriching a whole catalog with AI, at scale, with no oversight and no strategy behind it, is one of the fastest ways to turn a distinct product range into generic filler that reads like everyone else's. And the AI surfaces these tools promise to win pull from the same product data Google already uses, so weak data doesn't just fail to help. It can drag your products down everywhere at once.

So it's worth separating two things that get sold as one. Being enriched is not the same as being recommended. You can push AI-generated data into every field you own and still watch an assistant recommend a competitor. Worse, you can degrade your brand while you do it. This post is about why that happens, and what enrichment looks like when it's built on a foundation instead of used in place of one.

We have been here before

When AI visibility first became a thing a couple of years ago, dozens of tools appeared almost overnight. Some tiny, some backed by serious money. The message was loud and consistent: SEO is dead, and the only thing that matters now is optimizing for AI. Or "GEO," a term I still can't quite bring myself to like.

I say this as someone who was in that market. I built and sold an AI visibility tracker called Trackerly.ai. But I never told users that handing over their keys would win them AI. Even then, I argued that building a brand worth finding was the number one lever, and that tracking existed to measure whether you were getting there, not to replace the work.

Look at where that wave landed. Plenty of those platforms have since disappeared, and AI optimization never replaced SEO. The "SEO is dead" crowd softened to "SEO plus GEO," which is closer to right but still undersells the relationship: SEO is the foundation everything else gets built on. GEO is one thing you can layer on top, but only if that foundation is solid. And even then, from what I've seen working with brands, the share of revenue coming from AI answers tends to be small, often a rounding error next to organic search.

Then there were the tactics that got yelled into existence. Self-promotional listicles. Buying citations on spammy websites. Turning all your pages into .md versions and llms.txt files. The people selling this were either very inexperienced, or they were selling snake oil. I don't know which is worse.

The tactics that aged badly

Take self-promotional listicles, where a brand publishes its own "best X" ranking and quietly puts itself at number one. It was sold hard, and it worked for a while. Then Lily Ray documented what happened next: across a recent bout of Google ranking volatility, sites leaning heavily on these listicles lost significant visibility, several dropping by more than 40%. And because AI answers lean on Google's results, those losses carried straight into AI Overviews and, in many cases, ChatGPT.

Glenn Gabe found the same pattern with scaled AI content. A site that had published hundreds of thousands of AI-generated pages got hit with a manual action, collapsed in Google, and then watched its ChatGPT citations fall too, because ChatGPT often grounds its answers on Google's index.

The lesson in both is the same. AI shopping and AI answers sit downstream of Google. A shortcut that gets you penalised there doesn't stay contained. It follows you into the AI surfaces you were chasing. As Lily Ray likes to put it: it works, until it doesn't.

Same movie, new poster

Ever since Google announced and started rolling out the Universal Commerce Protocol, I've watched the exact same pattern repeat. Only this time the label isn't GEO, it's agentic commerce. Platforms are promising users agentic commerce domination: hand over the keys, let the tool enrich every part of your brand, and it'll get picked up by AI.

This narrative misses two things. The first is that agentic commerce and UCP are an evolution of organic shopping, not a separate planet. The playbook is mostly the one that already decides how your products show up in Google. These tools have a commercial incentive to frame it as an entirely new discipline, because a new discipline justifies a new budget line. But treating it as a standalone thing you can win in isolation, while neglecting the organic shopping opportunity that's already sitting in front of you, gets the priorities backwards.

The second is that these features are barely rolling out. The traffic and revenue flowing through them today is a fraction of the main event. Even now, more than a year after ChatGPT announced its own version of a product feed portal, that portal still isn't broadly available. It's real and worth preparing for. It is not a place you can go win while ignoring the foundation underneath it.

The real job hasn't changed

Carrie Rose put the underlying point better than I could, writing about the SEO versus GEO debate. As she argues:

"But the job was never really 'rank the website.' The job was always: make the market want you, know you, and trust you. Search was just where that showed up."

Being cited, or being enriched, is not the same as being chosen. In shopping the risk is sharper, because agentic and AI surfaces draw on the same product data Google already uses, much of it flowing through the Shopping Graph. Feed it thin or misaligned data and you don't just fail to help yourself, you risk your products being misread. Callum Lockwood, Director of Organic Search at Re:signal, made this point in his Organic Shopping Talks interview: without a foundational understanding of your products from source, you risk them being misrepresented or misunderstood.

What good enrichment actually looks like

None of this is an argument against using AI to improve product data. I built AI-assisted enrichment into Productrise. The difference is governance: the guardrails that keep enrichment grounded, accurate, and aligned with your brand instead of drifting into slop.

The most important guardrail is data, because you can't enrich blind. Productrise measures, every single day, which of your products actually show up in Google's organic shopping results, for which queries, and where you're losing ground to competitors. That daily visibility data is exactly what should feed your enrichment: you fix the products and attributes that are underperforming, then check whether the change moved anything, instead of rewriting everything and hoping. And, as Callum advises, you test before pushing mass-scale feed changes, not after.

Jamie D'Alessandro, SEO Manager at JAKALA, framed the healthy version of this in her Organic Shopping Talks interview. Her view is that the brands making product data easy for Google to extract, and as enriched as possible while staying genuinely valuable to the user, are the ones that gain an edge. What she struggles with most is bad data at the source, and her fix isn't to patch a few products by hand, it's to correct the process systemically. That's governance.

Beyond the data, good enrichment starts from source rather than thin air, keeps your feed and structured data in sync (they're separate systems that need to match, as I cover in my piece on why your product feed is an SEO asset), and protects your brand voice instead of flattening it. If enrichment makes your listings indistinguishable from every competitor running the same tool, it's working against you.

A feed audit is the most practical place to see where you stand: which titles are too short or stuffed with duplicate words, where structured data is missing, where prices don't match between your feed and your site.

Data enrichment is a multiplier, not a strategy

AI enrichment is a multiplier. Applied to a strong foundation of clean feed data, matched structured data, clear positioning, and real SERP insight, it can genuinely help. Applied on its own, with the keys handed over and no governance, it multiplies whatever is already there. And if what's already there is thin, it multiplies slop, which is how a brand ends up washing out with everyone else running the same playbook.

The good news is that the foundation is measurable today. Organic shopping in Google runs on the same product data that AI and agentic surfaces draw from, which makes it the most practical place to start. A useful first step is seeing where your products currently appear, and where you're not yet visible for the categories you care about.

If you want to start from data instead of a slide about the future, you can create a free Productrise account and track where your products rank in Google's organic shopping results, no credit card required. Get the foundation right first. Then let enrichment do what it's actually good at.