Website traffic falling? How much Google AI Overviews takes (and what to do)

Organic traffic falls month over month, the SEO team finds no technical errors, and nobody can really say why. Since early 2026 the most likely answer has a name: Google AI Overviews.

Good news: it’s measurable. Bad news: you don’t fix it by doing more SEO the old way.

Key takeaways:

• An AI Overviews traffic drop is diagnosed by comparing impressions and clicks in Search Console: if impressions hold but clicks collapse, the AI is answering instead of you.
• It mostly hits informational top-of-funnel queries, not transactional or brand ones. Segment before you react.
• Without a baseline of clicks PER query, any theory about the drop is a meeting-room opinion, not data.
• The countermove isn’t rewriting everything: it’s making content citable by AI engines and concentrating value on pages that convert.
• We rebuilt ILTEC’s AI-search-optimized site in 3 weeks. Measurement comes first, always.

What AI Overviews are and why traffic drops

AI Overviews are the answers Google generates at the top of the search results. The user reads the summary and often clicks no link at all. Your page can be cited as a source and still get zero clicks.

Result: you keep appearing (impressions hold), but people no longer reach your site. This is the pattern your analytics report flags as an “unexplained drop”.

Step 1. Build the baseline in Search Console

You can’t say “AI Overviews are taking my traffic” without a number. You need a baseline, exactly like any measurable project (we cover the same logic in how to measure the ROI of an AI agent for agents).

In Google Search Console, for the last 6-12 months:

• Export impressions and clicks per query.
• Compute the CTR query by query.
• Isolate queries where CTR collapsed while impressions stayed flat or grew.

Those are your AI Overviews candidates. Not all of the drop comes from there, but this pattern isolates it.

Step 2. Segment by query type

AI Overviews don’t hit everything equally. Split queries into three groups:

• Informational (“what is”, “how does it work”, “difference between”): the hardest hit. The AI answer satisfies the user before the click.
• Transactional (“price”, “quote”, “near me”): more protected, the user wants to act.
• Brand (your company name): nearly immune.

If most of your lost traffic is informational, don’t panic and cut the transactional content that still converts. That’s the classic mistake.

Step 3. Quantify the damage in euros, not visits

A lost visit isn’t equal to another. Take the informational queries in decline and ask: did that traffic convert?

• If those visits generated leads or sales, the damage is real and needs addressing.
• If they were “curiosity” visits that never converted, a traffic drop isn’t a revenue drop. Relax and focus your energy elsewhere.

This step separates the real problem from noise. Most companies skip it and panic over numbers that don’t matter.

Step 4. The countermove: from traffic to citability

You don’t recover the same visits: you change the goal. If the user reads the AI answer, you want that answer to cite you, with your brand and a link.

In practice that means:

• Structuring content into extractable answers (short paragraphs, clean definitions, verifiable data).
• Concentrating value on pages that convert, not on dozens of generic articles.
• Getting found not only on Google, but on ChatGPT, Claude, Perplexity.

That’s the work we do with AI Search Optimization: €2,000 audit, €3-15k sprint. For ILTEC, an office technology firm in Biella, we rebuilt the AI-search-optimized site in 3 weeks and added a sales agent via QR code on the machine fleet. Traffic wasn’t the KPI: qualified leads were.

If instead your problem is downstream, meaning the leads that arrive aren’t worked well, look at Sales & Marketing before touching content.

When it’s NOT (only) AI Overviews

Operational honesty: don’t blame everything on AI. First rule out:

• Seasonality (compare year over year, not month over month).
• A classic Google algorithm update.
• A technical problem (indexing, speed, a botched migration).

If you did the Step 1 baseline, you distinguish these cases in half a day. If you didn’t, you’re guessing.

The next step

If you see traffic falling and want to understand how much really comes from AI Overviews and what’s worth doing, let’s talk: 20 minutes, no pitch. Or start with the AI Search Optimization audit.