In September 2025, a rumour went round the SEO world that Google had quietly added an AI Overviews filter to Search Console’s Performance report. John Mueller shot it down publicly. Fabricated, he said. No such filter existed.
Eight months later, on June 3 2026, Google shipped one for real. A dedicated Generative AI performance report, built to show exactly what everyone thought they’d been promised the year before. That’s the whole story of this topic in one beat. Something everyone assumed was already happening, wasn’t, until it suddenly was, and even Google’s own comms team couldn’t keep the timeline straight.
I get asked about this constantly right now, by business owners who’ve watched a competitor show up in a Google answer box and want to know how, and by junior marketers trying to build a strategy around a feature that’s still being built underneath them. So here’s what I actually know, what I’ve tested myself, and what I’d tell you if you asked me over coffee instead of in a client deck.
Table of Contents
What AI Overviews Actually Are (and How They’re Different From AI Mode)
An AI Overview is the AI-generated summary that shows up at the top of a Google results page, pulling from a handful of web sources to answer a query directly. AI Mode is a separate, more conversational surface inside Search – built for the multi-step, comparison-heavy questions that used to take someone three or four searches to answer properly.
Google’s own documentation says both features can use something called query fan-out – issuing a batch of related searches behind the scenes to build a fuller answer before showing you anything. If that sounds familiar, it’s because it’s the exact mechanism I wrote about in Query Fan Out in GEO – clusters of related content don’t just help you rank for more terms, they’re structurally what these AI systems are built to consume.
The two features aren’t the same thing wearing different names, and treating them that way is one of the more common mistakes I see. AI Overviews behave like an enhanced snippet. AI Mode behaves like a research assistant. Different triggers, different models, and increasingly, different content wins in each.
There’s a third front nobody’s really talking about yet. Google started blending AI-generated answers into the People Also Ask boxes in November 2025, and by early 2026 roughly 12.6% of PAA answers were AI-written rather than pulled from a snippet. Eighty-one percent of AI Overviews now ship with a PAA section attached. AI in search isn’t one box anymore. It’s leaking into the furniture around it.
What’s Actually Happening to Your Traffic
Here’s the part that matters more than the mechanics. Ahrefs ran a study across 300,000 keywords using Search Console data, comparing December 2023 against December 2025. When an AI Overview appears above the number one result, that top result now keeps roughly 42 of the 100 clicks it used to get. Position two loses about half its clicks. Even position ten, buried at the bottom of the page, drops close to 20%.
That 58% figure had nearly doubled in eight months. Whatever pace you think this is moving at, it’s probably faster.
Seer Interactive ran a parallel study across 53 brands and 5.47 million queries and found organic click-through on AI Overview queries falling 61% by September 2025 – then recovering to 2.4% by February 2026, up from a low of 1.3%. Zero-click searches, where nobody clicks anything at all, now sit at 60% across all queries and climb to 80-83% once an AI Overview shows up on the page. Informational content gets hit hardest – how-to guides, health information, recipes, anything explaining rather than selling – down 30-40%.
I want to be straight with you here. If your traffic dropped and you’re staring at an AI Overview on your target keyword, it’s tempting to point the finger and move on. Don’t. Before you decide AI Overviews cost you the traffic, go check your actual rankings and whether you lost keywords outright. Nine times out of ten there’s a more boring, more fixable explanation sitting in your Search Console data, and blaming the AI Overview means you skip the diagnosis and never fix the real problem.
What Actually Gets a Page Cited
Most of what’s written about this topic is guesswork dressed up as strategy. So instead of repeating someone else’s seven-point listicle, I ran the numbers myself.
Working across six B2B SaaS domains and more than 80,000 keywords – Usercentrics, SEON, Aircall, Nextiva, Rezi, and RingCentral – I pulled every keyword that triggers an AI Overview, matched it against whether each domain’s own page got cited, and ran correlation and logistic regression on the result. Here’s what actually moves the needle.
Rank one is close to a hard gate. Across five of the six domains, citations came almost exclusively from the number one organic position. Drop to position two or three and the citation rate falls to zero, or close to it. Only SEON broke the pattern meaningfully, pulling 8% of its citations from outside the top slot. Every rank position you lose costs you roughly a quarter of your citation odds.
Once you’re ranked first, backlinks stop mattering. Not “matter less.” Stop. Across all six domains, backlink count and domain authority showed no statistically significant relationship with citation once rank was held constant, and in a couple of cases the relationship ran slightly negative. Links get you to the top spot. They do nothing once you’re standing on it.
Informational intent beats commercial intent by a wide margin. Pages tagged informational were two to eight times more likely to get cited than pages carrying any commercial framing, and a commercial flag on a page cut its odds by roughly half. Keyword difficulty between 40 and 60 correlated positively with citation on four of the six domains – Google’s AI system seems to lean harder on a summary when a topic is genuinely contested, not when it’s simple.
Page type followed a clear order almost everywhere: resource and glossary pages got cited around 9% of the time, plain blog posts 5.5%, listicles 2.7%, and product pages usually bottomed out near 1.8%. RingCentral broke that pattern entirely, with product pages hitting a 15.4% citation rate. I went and looked at why. Their product URLs are stuffed with genuine FAQ content and how-to detail – pages that read like a help centre article, not a pitch. The lesson isn’t “avoid product pages.” It’s that Google’s system is reading tone and structure, not the folder your URL sits in.
I’ll say the honest part out loud, because I don’t trust content that doesn’t. This is six domains in one adjacent B2B SaaS niche. It’s not gospel for every industry, and I’d want to see the same test run against ecommerce or local service businesses before I’d call any of this universal. But the sample is real, the numbers are real, and nobody else writing about this topic has shown their working the way I just did.
Entities and the Relationship Google Needs to See
There’s a piece my own data doesn’t fully explain, and it’s worth being honest about the gap. Backlinks stop moving citation odds once you’ve secured the top spot. Entities are a different mechanism entirely, and I think they matter on both sides of that equation.
An entity, in Google’s terms, is a clearly defined thing – a person, a brand, a service – that the system can identify with confidence and connect to other things it already knows. When I build anchor text around a brand name paired with a specific service, I’m not chasing link volume. I’m building the relationship Google needs to understand who does what. That relationship work helps cement the ranking that gets you to position one in the first place, and separately, it appears to raise Google’s confidence in citing you once you’re there – because citation isn’t just about which page ranks first, it’s about which source the system trusts enough to quote by name.
This is where E-E-A-T actually earns its reputation instead of just its acronym. Schema markup, author bios, and structured data are the technical scaffolding, and they’re worth doing properly. But scaffolding around nothing is still nothing. If the person named in that bio can’t demonstrate they know what they’re talking about, the bio is decoration. I’ll hold myself to that standard here too – this site still shows blog posts under “admin” instead of my name, which is exactly the kind of gap that undercuts everything I’ve just told you to fix on your own site. That’s getting corrected, and it should have been done already.
How to Actually Check If You’re Showing Up
Right now, the honest answer is: with real difficulty, and no single clean dashboard.
Google’s new Generative AI performance report gives you impressions – how often your pages showed up in AI Overviews, AI Mode, and AI-powered Discover, broken down by page, country, and device. What it doesn’t give you is clicks, or CTR, or anything you can use to make a spend decision. It’s currently rolling out to a subset of sites rather than everyone, so plenty of people reading this won’t have access to it yet. It’s a start. It is not a decision-making tool.
Outside of that, you’re working with Ahrefs data and manual spot-checks – searching your target terms yourself and watching whether an Overview shows up and whether you’re in it. It’s slow, it’s inconsistent because Overviews don’t trigger on every search every time, and it doesn’t scale past a handful of priority keywords. Nobody has built the clean answer yet. Anyone telling you otherwise is selling you something.
Don’t Let AI Overviews Drive Your Content Calendar
Here’s where I’ll push back on the entire premise a lot of AI Overview content operates from. The question isn’t “will this piece get cited by an AI Overview.” The question is “does the topic cluster need this piece to be complete.”
I look at what a cluster is missing before I look at what would win a citation. Build the content the cluster actually needs, sequence it properly, link it together, and the citations follow from having genuinely useful, well-structured, complete coverage of a topic – not from reverse-engineering a page to satisfy a citation algorithm that changes every few months anyway. Chase the algorithm and you’re rebuilding the strategy every time Google ships an update. Build the cluster and the algorithm has to come to you.
AI Overviews aren’t replacing search. They’re a layer sitting on top of it, built from the same organic results that were already there, pulling harder on the pages that already do the fundamentals right. Get the fundamentals right first. Everything else in this piece is detail work on top of that foundation, not a substitute for it.
Where this goes next, nobody fully knows yet. The tools are still catching up to the feature. That’s not a reason to wait. It’s the reason to start now, while most of the field is still guessing.
FAQ
Do I need special schema markup to appear in AI Overviews? No. Google’s own documentation states there are no additional technical requirements or special schema needed beyond standard SEO fundamentals. Schema helps your content get parsed cleanly, but it isn’t a citation shortcut on its own.
What’s the real difference between AI Overviews and AI Mode? AI Overviews are a summary that appears above standard results on a normal search. AI Mode is a separate, more conversational search experience built for multi-step or comparison-heavy questions. Both can use query fan-out, but they run on different models with different trigger conditions.
Can I opt my site out of AI Overviews? Yes. Google added a Search Console control that lets site owners block their content from appearing in AI Overviews, AI Mode, and AI features in Discover, separate from standard robots.txt directives.
Is AI Overview prevalence still growing in 2026?
The exact number depends entirely on who’s measuring. Trackers range from roughly 16% to over 60% of tracked queries showing an AI Overview, and Google itself puts the figure at “roughly 50%.” The methodology varies too much to trust any single number, but the trend across every tracker points the same direction.
What does zero-click search actually mean for my business?
It means a growing share of people who search for something never click through to any website, because the answer appeared directly on the results page. Zero-click sits around 60% of all searches and climbs to 80-83% when an AI Overview is present, which changes what “success” should look like for informational content specifically.
How do I track whether I’m being cited in AI Overviews today?
Google’s new Generative AI performance report in Search Console shows impressions if your site has access to it, though it doesn’t yet show clicks. Beyond that, Ahrefs data and manual searches on your priority keywords are the realistic options right now – slow, but the only tools that currently exist.
No comment yet, add your voice below!