AI Search Data: What Google Hides (and How to Act Anyway)

When your catalog speaks to AI without telling you

I audit 45 e-commerce sites every month. Since June 2026, nearly all ask me the same question. « Stéphane, have you seen these new AI data in Merchant Center? What does it mean? »

A client calls me on a Tuesday morning. An organic cosmetics brand, 1,200 SKUs, 80,000 organic sessions per month. A new section appeared in their account. « AI performance insights. » 1,200 impressions. Supposedly from AI. But zero clicks. No exact queries. Just labels. « Vegan anti-aging skincare. » « SPF 30 hydrating cream. »

Intriguing.

Incomplete.

They ask what they can do with it. I tell them it’s like a treasure map without the route. But we can dig anyway.

Google opens the door… but not the window

Mid-July 2026, Google launched a pilot. Search Engine Journal reported it. The « AI performance insights » report appeared in Merchant Center under Analytics > Products, AI Performance tab. Brodie Clark, independent SEO consultant, showed the first screens from a client sub-account.

What you find there:

• Shopping questions grouped by type (category search, specs, reviews)
• Query frequency
• Purchase journey stage
• « Product terms » — words buyers use (e.g., « maximum cushioning, » « arch support »)
• Share of voice compared to competitors

What you don’t find there:

• Clicks
• Individual queries
• Conversion rate
• Exact position

Google says in its documentation: click on « product terms, » add them to your titles and descriptions, fill missing attributes. That’s it.

A first step, not a complete solution.

Grouped questions: the strategic fog

« That’s enough query data to tell you which attributes to add to your product listings, but it won’t tell you whether AI Mode sent anyone to your site. » – Search Engine Journal, July 2026

Google aggregates questions. A label like « organic facial care » can bundle 15 exact variants. You get a trend, not the phrase.

Is this enough to work on your attributes? Yes. To measure ROI? No.

I observe a massive gap. With my clients, 4 product sheets out of 10 miss attributes demanded by AI in these reports. Meanwhile, AI impressions explode. 20% of shopping impressions already come from these surfaces on certain catégories (rough estimate from my deployments).

You’re visible. But you don’t know it.

Avant de passer à l’action, voici la photographie des fiches produits d’un client typique : 60% sont complètes, 40% manquent d’attributs clés demandés par l’IA.

Action #1: plug the holes in your catalog

I did this for an organic cosmetics brand.

I opened the « AI performance insights » report. I sorted « product terms » by frequency. I listed recurring attributes: skin type, organic certification, SPF, texture, travel size.

Result: 14 missing attributes across 35 strategic sheets. None in the product feed.

We updated the feed in 48 hours. Enriched titles. Completed descriptions. A few additionalProperty structured data tags added.

Three weeks later: +60% AI impressions. No measurable clicks, of course. But organic traffic on the product pages involved jumped +12%.

Correlation? Yes. Causation? I can’t prove it. But the curve climbs.

Après avoir ajouté les attributs manquants dans le flux produit, voici l’évolution constatée : +60% d’impressions issues de l’IA et +12% de trafic organique sur les pages concernées.

And Search Console in all this?

In Search Console, it’s even more frustrating. You filter results by « AI type » in the Performance report. You see impressions. But zero clicks are attributed.

Zero.

Google separates these surfaces. No queries are shared. Just impression volume climbing. Like background noise getting louder without you knowing what’s being said.

I advise my clients two tracking points:

• If AI impressions spike strongly in GSC,
• and direct or brand traffic follows the same slope,
• you have a real impact signal.

It’s not exact science. But it’s an early indicator.

Actions #2 and #3: trace the untraceable

Add UTM parameters to your shopping feed URLs. Even if clicks from AI Overviews or AI Mode aren’t broken out, they can land in your Analytics via the Shopping channel. You’ll then see the source as « Google / CPC » or « Google / organic » depending on type. Distinguish these campaigns with specific UTM parameters. It’s a workaround, but it sheds light.

Monitor AI « share of voice. » The Merchant Center report gives your share of voice on « product terms. » If it drops 15 points on a category, competitors own the AI terrain. React. Enriching content is your only weapon.

These two levers don’t replace clicks. But they keep you from flying blind.

Will Google ever give us the clicks?

I’m not sure. Marie Haynes often reminds us: transparency isn’t Mountain View’s priority. These reports will stay intentionally vague to protect the ad ecosystem.

So the question is: will you wait for clicks to act? Or will you build a catalog that AI understands effortlessly, before your competitors do?

You have enough clues to start.

Stéphane Jambu

SEO & AI Engineer

I build growth systems / AI / Neuroscience | 650+ clients · 80 LinkedIn testimonials · 30 years of expertise · 15 years of systems running without me.

Follow on LinkedIn