AI Impressions in Search Console: Mueller’s Rule Changes Everything for Your Product Sheets

15 audits, one blind spot: your AI impressions aren’t what you think they are

I review 15 sites per week. Since the AI Search Console report went live, I ask one systematic question: « What are you counting as an AI impression? »

In 12 out of 15 cases, the answer is wrong.

Not because the numbers are faked. But because the counting rule, clarified by John Mueller on June 25, 2026, flew under the radar for most e-commerce businesses. And when you’re managing hundreds or thousands of product sheets, this misunderstanding costs real money.

Take this: a fashion catalog with 8,000 SKUs. Search Console reports 4,200 AI Overviews impressions over 30 days. At first glance, the SEO lead celebrates. « Our sheets are cited everywhere. » Except they’re not. The truth is only 1,260 of those impressions were actual visible links without user action. The other 2,940 were hidden in a collapsed panel, behind an « expand » button nobody clicks.

John Mueller, Search Advocate at Google, explained on Bluesky that an impression in the AI report only counts if « a link to your site is shown. » Key nuance: if the user has to « activate » the link display, the impression is only recorded after that click. In other words, a product sheet cited in a combined block doesn’t generate an impression until the user expands that block.

This rule changes everything for e-commerce tracking. Brands that thought they were capturing strong AI presence find themselves with a phantom metric. Those who adjust their reading gain an instant edge: reliable visibility measurement they can actually use to improve their sheet rankings.

What John Mueller actually said: AI impressions are earned

Google Search Console’s AI performance report opened for testing in June 2026, first in the UK. Nicola Agius, SEO director at Reach PLC, asked Mueller concrete questions every e-commerce operator should ask themselves. What happens when a brand icon appears without the article visible? Does a linked favicon count as an impression? Mueller answered directly, picked up by Search Engine Journal.

His rule: only the presence of a link to a page on your site triggers an impression. An icon alone, with no link, doesn’t count. And if that link sits in an element requiring user action to expand, the impression is only recorded once that action happens.

For a product sheet, this covers two concrete scénarios.
1. The sheet is cited directly in the AI Overviews block, with its link visible: impression counted immediately.
2. The sheet is grouped in a « carousel » or « about your search » section the user has to click to expand. No impression without that click.

The trap is huge for large catalogs. A well-optimized product sheet can end up in a secondary block for a long-tail query, but if nobody expands the block, your metrics balloon from zero. Meanwhile, you keep believing your content is visible.

I had a client in home décor, 4,200 sheets. The AI report showed 3,700 impressions that month. After filtering by « query type: AI Overviews » and cross-referencing with the « activated » parameter, we discovered 2,200 of those impressions were phantoms. No activation. No expanded link. A presence that exists only on paper. So we reallocated our optimization work to where activation was lowest. You’ll see the result in the next section.

Mueller’s rule reveals the true nature of your AI impressions. Here’s how the 39,800 raw impressions from one e-commerce client break down into activated (visible link) and non-activated (hidden in expandable sections).

A catalog of 8,000 sheets, 1,043 cited… yet still an illusion

Take the case of a French online fashion site. 12,000 active SKUs, but I'm focusing on 8,000 product sheets that had at least one citation in AI Overviews over 90 days. In Search Console, I saw 1,043 sheets appearing—that's 13% of the catalog. The raw report showed 39,800 impressions in 30 days.

The teams were happy. The agency talked about "strong AI visibility." I dug into the numbers. I cross-referenced impressions with activation segmentation. Reality check.

Of 39,800 impressions displayed, only 11,870 were activated. In other words, 27,930 appearances were never-expanded links, in collapsed panels or clusters too low down. The user never saw the product link unless they clicked "expand."

We stopped celebrating the raw number. We refocused tracking on activated impressions and analyzed the content of the lowest-performing sheets. Result in 8 weeks:

• Activated impressions jumped from 11,870 to 36,710 per month.
• That's a +209% increase.
• Activation rate (activated impressions / total citations) went from 29% to 68%.

How? No magic. We applied a protocol built on understanding Mueller's rule. Every sheet was re-evaluated to increase the odds its link shows visible without effort, structuring data so Google displays the link upfront.

I'm not selling a miracle method. I'm showing what happens when you stop tracking the wrong metric.

Why your best sheets stay invisible (and how to measure it)

When you run an e-commerce business, not every product sheet gets the same treatment in the AI Overview. Some appear in direct answers to comparison questions. Others hide in a "similar products" card. Mueller's rule distinguishes two types of presence.

Immediate presence: the link is visible without interaction. This is the case for a sheet cited as primary source, or in an image carousel with an active link. Conditional presence: the link sits in a collapsible section, an expandable list, or a cluster. Until the user clicks, your product stays invisible.

To measure AI impressions on your sheets, first filter your Search Console reports by "activated links." In the Performance report, add the "Activated Link" dimension if you have access to the test version. Otherwise, compare your raw AI impressions with actual clicks from AI Overviews (given in the console). A gap over 70% should alert you.

With my fashion client, the gap was 70% (39,800 raw impressions, 11,870 activated). That signal pushed me to examine each sheet. 610 sheets out of 1,043 had zero activation rate. And their links were systematically found in combined blocks with no visible content.

Understanding this distinction moves you from metric inflation to real steering. For e-commerce, the impact is direct: an unactivated sheet gets no qualified traffic, despite what the AI promised.

3 levers to shift your AI impressions from hidden to real visibility

Build product sheets so explicit and structured that the algorithm naturally picks them for immediate placement. Here are the 3 levers I use with my clients since the clarification.

1. Product schema enriched beyond the minimum.
Beyond price and availability, add hasMerchantReturnPolicy, average rating, review count, structured description with embedded FAQ. An AI Overview block often builds from a sheet that gives clear answers. The less the AI has to search elsewhere, the more your sheet gets cited in primary position. It's mechanical.

2. Semantic cocoon architecture applied to product pages.
I build content silos around each product category: a master sheet with comparison tables, an FAQ, a buying guide. Each product sheet inherits this structure. The AI sees thematic authority. Result with a décor client: sheets linked to a cocoon saw activation rate jump from 24% to 67% in 5 weeks.

3. "Question-answer" phrasing in your long descriptions.
80% of AI Overview citations come from a passage answering a question. Embed at least three relevant questions per product sheet, with a factual answer in two sentences. Mark them with FAQ schema where relevant. The user doesn't need to click: the answer is already visible—your link is right there, instantly.

These three levers don't demand more content budget. They demand discipline. And the numbers speak: +209% activated impressions in 8 weeks with the fashion client I mentioned, by rebuilding only the top 20% of viewed sheets and linking each to a category cocoon.

We structure for activation, not for reports. The shift is that simple.

And now, do you know how many AI impressions are actually activated?

I wrapped an audit call recently. The client, a pure-play in childcare products, was looking at their AI Search Console report for the first time through an activation lens. They had 23,500 raw impressions that month. Under 6,000 activated. They said to me: "I feel like I've been measuring empty air for six months."

That sentence says it all. Mueller's clarification isn't a technical detail for pedantic SEOs. It's the key to stop celebrating phantom numbers and start steering the only thing that matters: real link exposure in AI results.

What changed with this rule?

• Impression is no longer a given—it's built upfront through content structure.
• Product sheets will no longer be judged by raw appearance volume, but by their ability to get expanded.
• E-commerce tracking must now cross Search Console AI with an activation metric, not the overall report.

You have a catalog of hundreds or thousands of sheets. You saw an "AI Overviews" line pop up in your console. Don't ask yourself how many impressions you got. Ask yourself how many products are actually seen.

If you can't answer that precisely with your current report, that's the signal you need an audit to reconcile your data and your product structure. I'm not selling you the method. I'm showing you the pages.

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