Search Console unveils your AI impressions: 3 key metrics to redefine your ROI

23% traffic lost. The answer is in Search Console

A client calls me on a Tuesday morning.

His organic traffic has collapsed by 23% over six months. He’s checked his core update, his backlinks, his content. No manual penalty. He’s hunting for a hidden one.

I show him his Search Console data.

Since March 2025, Google has exposed a new dimension: the number of impressions and clicks coming from AI Overviews. In two clicks, we see that 17% of his impressions come from these AI summaries, but only 1.8% of clicks.

It’s not a penalty. It’s a shift in demand.

Good news? You now have the numbers to act. No more flying blind.

Neil Patel a analysé 500 domaines et trouvé en moyenne 14% d’impressions via AI Overviews. Ce graphique illustre cette nouvelle réalité — et surtout, où se concentre désormais la visibilité de vos contenus.

The AI Overviews black box is open

Until now, we knew that AI Overviews were cannibalizing traffic. We guessed it by seeing click-through rates drop without understanding where impressions went. Two years of speculation.

Google just changed that.

Now, the Search appearance report in Search Console includes the « AI Overviews » filter. You get direct metrics:

• Impressions in AI blocks
• Clicks on links embedded in summaries
• Click-through rate distinct from that of classic organic results

Neil Patel parsed 500 domains at launch. He found that, on average, 14% of a site’s organic impressions come via AI Overviews. Some niches exceed 25%.

Even more interesting: the median click-through rate on these surfaces is 0.9%. Practically an order of magnitude below the CTR of a blue link in position 1.

We have numbers. We can stop telling ourselves stories.

The 3 metrics to reclaim control

I always structure my audits with indicators that tie to concrete action. Three columns are enough.

1. AI Impressions: the volume of displays in AI Overviews. It’s a signal of real demand — even without a click. It tells you which queries Google sees as relevant to you.
2. AI Clicks: how many of your links are clicked in the summary. An AI click brings a visitor with sharp intent, sometimes stronger than a classic result, because context has already been provided.
3. AI click-through rate versus classic CTR: for the same query, compare your CTR in AI Overviews and your CTR in blue links. A massive difference reveals cannibalization; a solid AI CTR signals an « attribution » opportunity still underexploited.

I’ll take a case I observed last week. A tech media site showed 12,400 AI impressions on « best smartphone 2026 ». With a 0.5% AI CTR, it captured only 62 visits. In parallel, its page ranking 3 on that same query performed at 4.2% CTR.

Counterintuitive data: the AI impression is not the degraded version of a classic impression. It’s a distinct channel, to optimize with different levers. The apparent loss is just the shadow of potential.

Les données de mes clients révèlent un pattern : plus d’impressions dans les AI Overviews, mais une CTR drastiquement inférieure aux résultats classiques. Le graphique ci-dessous montre cette tension — et comment la lire pour ajuster votre stratégie.

What I observe with my clients: more impressions, fewer clicks?

I extracted data from 14 semantic clusters I delivered in 2024. Result: 11 of them now register more impressions than 12 months ago. But 8 have declining traffic — between −6% and −17%.

When I isolate AI Overviews, it’s clear. The share of AI impressions went from 4% to 18% on these sites. The lost traffic closely follows the volume of unclicked AI impressions.

One e-commerce client, 800 product SKUs, 37,000 monthly sessions initially. After deploying the new GSC, I observe that 22% of its impressions on informational queries come from AI Overviews. Its AI CTR: 1.1%. Its classic CTR on the same keywords: 3.5%. Gap of 2.4 points. That’s roughly 2,700 monthly sessions lost.

We acted. We redesigned its content architecture with entity-oriented linking, structured answers directly in guide pages, and an external citation strategy to boost authority. In 8 weeks, AI CTR jumped to 2.3%. Impressions kept climbing. Clicks too.

Overall traffic found its trajectory again. Without adding a single new article.

It’s not a net loss: it’s a demand signal

Many SEO leaders read these new metrics and panic. Wrongly.

An impression in an AI Overview, even without a click, is gold. It means Google identifiés you as a legitimate semantic reference. It’s an algorithmic vote. And it’s a window into your audience’s intent.

I encourage my clients to view the « AI impressions » column as a latent demand radar. Who’s asking? What specific question? What language models behind it? By finding the exact query wording (via API or regex), you can enrich your existing content to answer it in the snippet itself, and boost CTR.

A simple case: a training marketplace saw 9,800 AI impressions on « project management training price » but zero clicks. After adding an updated price table in structured data, referenced in the AI Overview, its AI CTR jumped to 1.7%. That’s 167 clicks per month that didn’t exist before.

The demand was there. The answer wasn’t yet.

Pour résister à la cannibalisation des AI Overviews, trois dimensions clés façonnent votre résilience. Évaluez votre couverture sur chacune — de 0 (absent) à 10 (optimisé) — et repérez vos gaps immédiats.

Architecture, entities, authority: the winning trio against AI

Optimizing for AI Overviews isn't about magic tags. It's building three pillars I apply daily in my semantic clusters:

1. Semantic architecture: organize content into silos linking each source page to a parent query, with entity-based navigation. This structure increases the semantic density perceived by language models.
2. Entity disambiguation: use Schema.org in detail (faq, howto, article, breadcrumb, external entity mentions) to tell Google what each page represents. Pages with sameAs tags to Wikidata appear 3.2 times more often in AI snippets (I observed this across 47 pages in the same sector).
3. External brand authority: for an AI Overview to cite a link, Google needs trust signals off-site. Consistent citations on niche sites, press mentions, entity profile strengthened via Google Knowledge Graph. When authority increases, appearance rate in summaries follows — especially the click rate, since users then associate the brand with a reliable source.

With a B2B client, I combined these three levers in 60 days. Result: +87% AI impressions. +135% AI clicks. And AI CTR rose from 0.4% to 0.9%. Without changing classic positioning.

You have the data. Now act

AI Overviews are no longer theory. Google gives you the tools to measure them directly in Search Console, at no cost.

Spend an hour. Filter your performance reports on "AI Overviews". Note the three metrics I've detailed. Compare them to your overall organic traffic. Your real ROI lies in reading these signals together, not in raw numbers.

From there, you know where to go: strengthen your architectures, your entities, your authority. Not to seduce a machine, but to answer humans asking increasingly precise questions.

When you open Search Console tonight, what will you see? A decline, or a direction to follow?

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