AI Overviews in Search Console: 3 traps that skew your decisions (and how to avoid them)

A phone call that smells like a trap

An SEO manager calls me on a Tuesday morning.
He’s seen his Search Console impressions jump +47% in two months.
He’s thrilled.

Yet there’s a catch.
His real organic traffic is down 18%. Sessions, product pages, shopping cart.
He doesn’t understand.
The site hasn’t changed.
No penalty.
Nothing.

I look at his Search Console with him.
63% of new impressions come from « Generative AI » — the AI Overviews.
And these lines show zero clicks.

It’s a new kind of anomaly.
The first major trap Google introduced in June.

Confirmation bias, which Guillaume Attias frames in the DOSE framework at BMO Academy, pushes you to treat this impression spike as a win.
You see a green number.
You validate your actions.
But the actual business is bleeding.

The Search Engine Journal article shows it: AI Overviews data in Search Console aren’t engagement metrics.
They’re designed to deceive.
If you make budget, content, or internal link decisions based on this, you’re flying blind.

I’ll show you the 3 main traps.
And most importantly, how to isolate them so you stop skewing your decisions.

Trap #1: impressions with zero economic value

Not every Search Console impression is a success signal.
But when you look at the « Generative AI » tab, it’s even more stark.

Google displays AI summaries at the top of the SERP.
The user reads the answer, gets their info.
They don’t click any link.
They leave.

Search Console logs an impression for pages whose content fed that answer.
Never reporting a click.
The result: impression curves climbing while generating zero visits.

On the e-commerce site I analyzed, AI Overviews drove 112,000 impressions in eight weeks.
0 clicks.
Zero conversions.
Impact on e-commerce KPIs? None.
Or worse: it prevented visitors from reaching product pages because the answer was already given.

This is what Dan Taylor, author of the Search Engine Journal piece, calls « an impression with no click delivers zero economic value ».
A hollow metric.

If you base your SEO investments on overall impression growth without separating AI Overviews, you risk scaling budgets on queries that bring you zero clicks.

The trap is subtle: your impressions rise, you sleep soundly, while your real traffic erodes.

Trap #2: the shared position 1 that distorts all your rankings

In a regular SERP, one page might be position 3.
Another in position 5.
In an AI Overview block, the rule changes.

Google assigns position 1 to all links in the generative box.
Even a secondary link, even an automatic link in a collapsible source.

Concretely, a page that never gets clicked, buried in an AIO dropdown menu, gets a position 1 in Search Console.
Another page, visible and clicked, gets the same label.
With no hierarchy of actual exposure.

This artificial leveling distorts your keyword tracking.
You see your product jump from position 6 to position 1 on multiple queries.
You think your optimization worked.
Except that « position 1 » generates only 2% of clicks — far from the expected 15% for a typical first position.

The bias is strong: everything looks like it’s improving while your organic positions haven’t moved.
Worse, your average position curve plummets, as we’ll see in the next trap.

Google’s official documentation confirms this behavior for AIO elements.
Search Engine Journal details it.
Your rank tracking tools don’t always make this distinction.

Trap #3: averages that shatter and hide the collapse

The direct consequence of trap #2?
Search Console average position becomes a statistical illusion.

In the client case mentioned, before AI Overviews appeared, average position hovered around 14.2.
A typical e-commerce profile, with lots of long-tail.
After two months of AI data, that average dropped to 5.7.
An apparent 60% improvement.
Problem: clicks fell 18%, and organic revenue dropped 22%.

Why?
Because thousands of AI Overview impressions inject a mass of « position 1 » entries with zero clicks.
They pull the overall average up and mask that your real positions on transactional queries (the ones that drive sales) are stagnant or declining.

You analyze your dashboard.
Average position: 5.7.
Congratulations, the team pats itself on the back.
Except when you isolate non-AIO segments, average position actually stayed at 14.1.
And your product audiences lose 8 positions on high-intent queries.
The masking effect is total.

If you plan your SEO strategy on these skewed averages, you’re making decisions blind.

How to isolate AI Overviews so you stop skewing your decisions

The first thing to do: segment.
In Search Console, use the « Appearance in results » filter → « Generative AI ».
Isolate these impressions in a separate view.
Never mix them into your standard traffic reports.

Here’s the method I use with all my e-commerce clients:

• Create a Data Studio (or Looker Studio) property with two segments: « Organic excluding AIO » and « AI Overviews ».
• On your main dashboard, hide the AIO segment from clicks, sessions, and conversion metrics.
• Track one KPI: the real organic click rate per thousand impressions excluding AIO, compared month over month.
• Keep the AIO segment on a secondary tab, only to monitor generative visibility without assigning it economic value.

For sites generating sales, I recommend going further:
Cross these views with Analytics data from product pages.
Identify pages appearing in AI Overviews where direct SEO traffic is falling.
This often reveals silent cannibalization.
The user sees the answer in the box and no longer clicks the natural link.

Also, brief your teams.
If your content lead optimizes using overall impression curves, they risk writing articles no one reads.
Reset objectives around clicks, not impressions.

What I observe with e-commerce sites: a 72% illusion

I review fifteen Search Consoles each week.
The same mistake everywhere.

A site with 800 SKUs, an outdoor gear specialist, sees overall impressions jump 52% in six weeks.
The owner is about to hire a writer to « ride the wave ».
I show him the breakdown by appearance type: 72% of new impressions are AI Overviews.
Clicks on product catégories are down 14%.
Investing content there would be throwing money away.

We reset the reporting in 48 hours.
We excluded AIO from decision metrics.
We looked at real click variations on purchase queries.
Result: +12% clicks in three months, by refocusing optimization on pages that actually convert.

Confirmation bias is the trap that costs the most.
You see green, you validate.
But the real signal is elsewhere.
Isolating AIO noise is how you take control back.

Reset your reporting to steer right

For me, this is the foundation of the DOSE framework.
Guillaume Attias repeats it: a decision maker’s first enemy is confirmation bias.
Green numbers telling you what you want to hear.

For an e-commerce operator — I’ve seen it a hundred times — every budget decision rests on metrics that deliver money.
AI Overview impressions are the opposite.
They inflate your KPIs and lull you to sleep.

My method:

• Physically separate AIO data from your performance reports.
• Set an objective around real organic clicks — not positions, not impressions.
• Track your non-AIO SEO conversion rates, month after month.
• Never make an investment decision based on an impressions curve that mixes everything together.

If you see your average position drop, ask yourself: are my clicks following?
If the answer is no, it’s the AI Overviews trap.
Not an improvement.
A mirage.

Are your SEO decisions based on impressions that deliver nothing?

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