Google’s Merchant Center pivot quietly contradicts its GEO line

Google rolled out AI shopping visibility insights inside Merchant Center yesterday. Share of voice against similar retailers. Shopping funnel performance across discovery, evaluation and purchase. Product term insights showing what people actually ask conversational shopping interfaces. Product attribute completeness scoring.

It's a small product release. Most people in the industry will scroll past it. They shouldn't.

Because this is the first time Google has shipped a measurement product that quietly concedes the entire "GEO is just SEO" position the company has been arguing in public for the better part of a year. Not in a press release. Not in a Search Liaison tweet. In a shipped feature, which is the only place company positions actually exist.

The shape of the concession

Read the feature list carefully. Share of voice benchmarking against similar retailers — that's a recognition that AI shopping surfaces produce a ranked output where some brands appear and others don't, and that the appearance is measurable as a distinct signal from organic rank. Product term insights showing popular conversational queries — that's an acknowledgement that the query layer in AI commerce looks fundamentally different from keyword search and needs its own reporting. Product attribute completeness — that's structured data scoring rebadged as discoverability scoring.

None of this exists in Search Console. The metrics Google has refused to add to Search Console for AI Overviews and AI Mode for over a year — citation tracking, share of voice, conversational query breakdown — have just shipped inside Merchant Center for retail.

The difference is that retail pays per click directly. Organic publishers don't.

Why retail gets the tools first

This is the bit worth sitting with. Google has known how to build these reports for a long time. The infrastructure to count brand appearances inside an AI surface and benchmark them against competitors is not technically demanding for the company that operates the AI surface. The question was never *can they*. It was *will they*.

The asymmetry isn't a bug in Google's roadmap. It's the roadmap.

The answer, now visible, is: they will when there's a paid acquisition product attached.

Merchant Center sits next to Shopping ads. The retailers being given AI visibility insights are, overwhelmingly, the same retailers running Performance Max and standard Shopping campaigns. Google can give them measurement because measurement creates demand for bid adjustments, feed optimisation services, and ultimately more spend. The reporting layer feeds the auction layer.

Organic publishers don't have an auction layer. So they don't get the reporting layer.

The asymmetry isn't a bug in Google's roadmap. It's the roadmap.

What this tells you about the bigger position

For most of the last year, Google's public line on generative search has been some version of "the fundamentals haven't changed, just keep doing SEO." Liz Reid said it. Search Liaison said it. The keynote slides said it. And honestly, the fundamentals largely haven't changed — that bit is true.

But the measurement story has changed enormously, and Google has refused to acknowledge it in the surface where most businesses would want acknowledgement, which is Search Console. The Vanessa Fox interview that ran yesterday made this point in passing — even the person who arguably invented Search Console is now publicly frustrated that the tool has no metrics for featured snippets or AI Overviews. That's a remarkable thing for the originator of the product to say out loud.

Now Merchant Center ships exactly the metrics Search Console doesn't have. For retail. For paying advertisers. In the US, Canada, Australia, India and New Zealand first.

The position is no longer "AI search doesn't need new measurement." The position is "AI search needs new measurement, and you'll get it where you pay for it."

What it means if you sell things online

Two practical implications, neither of which is "rush to optimise for AI shopping surfaces."

First, if you run a product business with a Merchant Center feed, the attribute completeness reporting is genuinely useful and you should pay attention to it when it lands in your region. Not because of some new GEO optimisation theory, but because Google has just told you, in writing, which structured attributes their AI surfaces are reading and which ones you're missing. That's the kind of explicit signal we almost never get. Take it.

Second, and more importantly: the share of voice benchmarking will show most retailers a deeply uncomfortable picture. If you're a mid-sized brand competing in a category dominated by three or four large players, you will see, for the first time, in numbers, how much of the AI shopping conversation belongs to your bigger competitors. That number will be ugly. It will tempt you to chase visibility through tactics that don't work.

Don't. The same lesson applies here that applies everywhere in AI discovery: the brands that get cited are the brands the model has heard of. Building brand recognition through earned media, real reviews on platforms outside Google's index, and editorial coverage moves the share of voice number. Tweaking your product titles does not, beyond a point you'll hit in about a week.

The honest read

Google has shipped a measurement product that contradicts its own public messaging. That happens inside large companies — different teams have different incentives, and the team that builds Merchant Center is paid to make advertisers spend more, not to maintain consistency with what the Search team says at conferences.

But the contradiction is the signal. When the people building the actual tools ship features that imply AI shopping is a distinct surface requiring distinct measurement, and the people doing the public-facing communication keep telling you it isn't, trust the tools. The tools are what the company actually believes. The rest is just what the company has chosen to say.

The next twelve months of AI search measurement will be built where the money is, by the team that gets paid to build it. Retail gets the early version. Publishers and service businesses get to wait, or build their own, or use whichever third-party tool reverse-engineers what the first-party tools are measuring. As I wrote last week, the measurement layer for organic AI search is being built by outsiders for a reason. Yesterday's release is that reason, in product form.