The pattern shows up more and more often in Search Console. Organic traffic is flat. Rankings look fine in Ahrefs. Nothing in the Performance report screams disaster. Yet click-through rates on the pages that make money are quietly sliding. The cause, increasingly, is Google's AI Overviews pulling the content into synthesised answers, listing the site as a source, racking up impressions, and sending almost nobody through. The site gets cited. It does not get visited.
That situation is now playing out across thousands of sites. And the data to understand it is sitting right inside Google Search Console, mostly ignored, because the new AI Overviews reports are genuinely confusing if you haven't spent time with them.
Let me walk you through what's actually there and how to read it without losing your mind.
What Google Actually Added (and When)
Google started rolling out AI Overviews in the US in May 2024, at Google I/O. The Search Console reporting came later, trickling out through late 2024 and into 2025. It's not a separate product. It lives inside the Performance report you already use every day.
A lot of people miss this: there are now two distinct "Search Type" filters. You had Web, Image, Video, News, and Discover before. Now there's a specific toggle that surfaces AI Overviews appearance data layered on top of standard web search results. Google calls the underlying metric an "AI Overviews" filter within the Search Appearance dimension. It's not immediately obvious because it doesn't live where you'd expect it.
To find it: go to Performance, select "Search type: Web", then look for the Search Appearance filter in the top row of filters. You'll see an option to filter specifically for queries where your site appeared inside an AI Overview. That's your starting point.
The Four Metrics and What They're Actually Telling You
The familiar four: impressions, clicks, CTR, average position. Same columns. Completely different meaning in this context.
Impressions in AI Overviews
An impression here means Google served an AI Overview for that query and your site was cited as a source inside it. Not that your blue link appeared below. Not that you ranked first. The AI module itself surfaced your URL as a reference. This is a new kind of visibility that has no real predecessor in the old reporting.
I've seen sites with 40,000 monthly AI Overview impressions generating fewer than 600 clicks from those impressions. That's a CTR of roughly 1.5%. Compare that to a top-three organic result which historically pulls 10-30% CTR depending on the query type. The gap is enormous.
Clicks
A click registers when someone actually taps or clicks through from the AI Overview citation to your page. This does happen. It's just rare. Users mostly get their answer from the synthesised text and move on. The citations are there partly for credibility signalling to the user, not primarily to drive traffic.
CTR
This is where it gets uncomfortable. AI Overview CTR benchmarks are still forming and Google publishes none. Expect AI Overview impressions to convert to clicks at a fraction of your standard organic rate for the same queries, and treat your own historical organic CTR, not an industry figure, as the baseline. If you go in expecting traditional organic CTR numbers, you'll spiral into unnecessary panic.
Position
Position within AI Overviews reporting reflects your standard organic ranking for that query, not a "position inside the AI module." Google hasn't given us a slot-number for where your citation appears within the generated text. So position 1 in this view means you ranked first organically for a query where an AI Overview appeared. It doesn't mean you were the first or most prominent citation in the AI answer itself.
How to Pull a Report That's Actually Useful
Raw numbers aren't where the insight lives. You need to cross-reference.
- Open Search Console Performance, filter to "Search type: Web."
- Add the "Search Appearance: AI Overviews" filter.
- Export to a Google Sheet (the built-in export button, top right, downloads a CSV).
- In a second tab, pull your standard web performance data for the same date range with no appearance filter.
- Use a VLOOKUP or XLOOKUP on the query column to match rows.
- Add a calculated column: AI Overview CTR divided by standard CTR for the same query.
What you're looking for: queries where you have high AI Overview impressions and a significantly lower CTR than your standard organic CTR for similar query types. Those are your problem queries. Those are the pages where AI Overviews are intercepting intent that used to convert into clicks.
An illustrative example of what this exercise surfaces: a set of informational blog queries where AI Overview impressions run into the thousands a month while clicks stay in double figures. The AI is summarising the answer completely, so users have no reason to click. The usual response is to restructure those pages around content that cannot be fully answered in a three-sentence synthesis: comparisons built on your own data, tools, pricing, or a decision the reader has to make. Measure clicks on those pages before and after. Do not assume the lift.
Segmenting by Query Type: The Part Nobody Talks About
Not all queries behave the same inside AI Overviews, and the report doesn't segment this for you automatically. You have to do it yourself.
Broadly, queries break into three buckets for AI Overview behaviour:
- Informational queries ("how does X work", "what is Y") get the highest AI Overview impression rates. CTR from these is brutal. Expect sub-2%.
- Navigational and brand queries rarely trigger AI Overviews, so you'll see fewer impressions here. If they do appear, CTR is usually much higher because users already wanted a specific site.
- Transactional and commercial investigation queries ("best X for Y", "X vs Y pricing") are the interesting middle ground. AI Overviews do appear for these, but users are often still in decision mode and click through at higher rates, sometimes 4-8% in my observations.
Segment your exported data manually by labelling query types. I usually just use a keyword filter column with simple tags: INFO, NAV, TRANS, COMM. Takes about 20 minutes for a 200-row export. Worth every minute.
What "Cited But Not Clicked" Actually Means for Your Site
Being cited in an AI Overview is not nothing. I want to be clear about that. There's a brand-impression effect. Users see your domain name as a credible source. That matters for brand recall, particularly in B2B contexts where the buying cycle is long and touchpoints add up.
But the honest read is this: if a page's primary job is to generate organic traffic that leads to a conversion action, and that page is now getting the bulk of its impressions inside AI Overviews, the page's business model is broken. The content is working hard. The business isn't seeing the return.
The Google Search Central documentation has some guidance on how Google selects content for AI Overviews, though it's fairly high-level. What it does confirm is that cited content tends to be well-structured, authoritative, and directly responsive to query intent. So at minimum, showing up there means your content quality is reasonable. That's not nothing.
Changes Worth Making Based on What You Find
Look, I'm not going to tell you to block AI Overviews entirely. You can attempt to opt out of AI training datasets via robots.txt (using Google-Extended as the user-agent), but that's a separate thing from whether Google cites you in a live AI Overview result. The robots.txt spec for Google-Extended exists, but blocking it doesn't guarantee removal from AI Overviews. Google is pulling from its index, not just recent crawls.
What I'd actually do, based on what the data shows:
- Identify high-impression, low-click pages. These are candidates for content restructuring toward deeper analysis, proprietary data, or interactive tools that can't be summarised.
- Add genuine conversion pathways higher on those pages. Email capture, a related tool, a product link. If people aren't scrolling from AI Overview traffic, you need the hook before the fold.
- Build content that requires the user to be on the page. Calculators, comparison tables, downloadable templates. Synthesised text can't replicate these.
- Watch position vs. CTR for transactional queries. If you're ranking position 1-3 for a commercial query that's triggering AI Overviews and CTR is still reasonable, hold the line. Don't restructure what's working.
- Flag queries where you appear in AI Overviews but rank outside position 10 organically. This is rare but happens. The AI is citing you as a source even though you're not appearing prominently in the blue-link results. That's a signal your content is authoritative on that topic even if your overall page SEO isn't optimised yet.
One more thing: check date ranges carefully. Google's own rollout of AI Overviews was uneven through 2024. If you're comparing a pre-rollout period against post-rollout, your impression and CTR curves will look alarming simply because of the product launch, not because you did anything wrong.
FAQ
Does appearing in AI Overviews help or hurt my SEO?
Neither, directly. It's not a ranking signal in the traditional sense. It reflects that your content was already trustworthy and well-structured enough for Google to cite it. The SEO impact is indirect: if AI Overviews reduce clicks to your page, engagement signals may weaken slightly over time, but this is speculative at this point. There's no confirmed mechanism.
Can I see which specific pages are cited in AI Overviews?
Yes. In the Search Console AI Overviews filtered report, switch the primary dimension from "Queries" to "Pages." You'll see which URLs are accumulating impressions from AI Overview appearances. Sort by impressions descending and start there.
Why does my CTR look so low in the AI Overviews filter compared to normal web search?
Because the user experience is fundamentally different. In a standard search, a user sees your title and description and has to click to get information. In an AI Overview, the information is served to them before they decide whether to click. Lower CTR is expected and doesn't mean your page is performing poorly. It means the product is working as Google designed it.
Should I try to optimise specifically for AI Overview citations?
Write accurate, well-structured, authoritative content. That's the same answer it's always been. There's no separate "AI Overview optimisation" technique that I'd stake my reputation on right now. The sites I see cited most consistently in AI Overviews are the ones that have spent years building genuinely useful content, not the ones that reverse-engineered the algorithm six months ago.
The data is there. Most people haven't looked at it properly yet, which is fine because it gives you a real edge if you do. Spend an hour with the export, run the comparison I described, and you'll understand your AI Overview exposure better than 90% of your competitors. That's not a guarantee of more clicks. But it is a guarantee of better decisions.