Google AI Overviews: What They Mean for SEO

Search has been disrupted more in the last 12-18 months than it has in the preceding 10 years combined. And that isn’t even hyperbole.

It’s highly likely you’ve seen changes in your own organic search performance. And, if somehow not, you must have seen people from all corners of the known universe talking about one or all of ‘AI search’, ‘AI visibility’, and GEO (generative engine optimisation).

ChatGPT and Claude aside, one of the changes to come out of the burgeoning AI trend of today is Google AI Overviews. You’ll certainly have seen them; they’re those informative AI summaries that appear at the top of an ever-increasing share of organic search results. 

And while they haven’t completely rewritten the rules of organic search and information discovery, they have augmented it. Zero-click search is now very much a real thing; searchers can find answers to their questions without ever leaving Google.

You can see how and why that might be a problem for publishers who rely on their traffic for revenue and discovery. Will anyone even land on this article?

Google, of course, says everything is fine and not to worry. Whereas, a growing number of studies and businesses are showing and finding otherwise. No doubt, the answer is nuanced and lies somewhere between the two perspectives.

For what it’s worth, below is our at-the-time-of-writing perspective on Google AI Overviews and the state of SEO. So, what’s going on, and what can you do about it?

What are Google AI Overviews?

Google AI Overviews are the prominent AI-generated responses you see at the top of search results when asking Google certain questions and queries.

Though they don’t appear at the top of every search query - and when they do appear, they occasionally feature below paid results - they are becoming more and more common as time rolls by.

If you want to see an example in the wild, AI Overviews most commonly appear above informational and problem-solving queries: ‘how’, ‘who’, ‘what is’, etc…

AI Overviews vs. AI Mode

At this point, it’s worth highlighting that while similarly named, there is a world of difference between Google AI Overviews and AI Mode.

Whereas the former is an in-SERP feature with the aim of - in Google’s words - enhancing the user experience. AI Mode is essentially opt-in and more akin to the experience you would get with ChatGPT and other LLMs.

AI Overviews are part of the Google search package. You don’t really have control over whether or not they appear. They’re baked into the experience.

💡 Further reading: GEO and Web Design: What is Generative Engine Optimisation?

How Do Google AI Overviews Work?

While its AI Overviews were officially launched in the US in May 2024, Google had originally introduced the functionality as part of its Search Generative Experience back in 2023.

At a basic level, AI Overviews work by pulling together information from a variety of sources using a technique called retrieval-augmented generation (RAG), before turning it over to Gemini - Google’s own LLM - to synthesise a summary.

As you’d expect, the process is somewhat more complex than that. And, broadly speaking, there are three areas worth getting into here:

Query Fan-Out

Query fan-out is more or less exactly what it sounds like. Instead of Google running just one search when an AI Overview is triggered, it also runs multiple related searches.

For instance, if you were to search ‘what’s the best CMS platform?’ that search would actually become potentially dozens of related sub-queries. These subqueries might cover searches related to CMS comparisons, ecommerce capability, multi-language support, and pricing.

Source Retrieval

The system will then pull content from across the web - primarily those highly ranked in organic search, though not always - and evaluate each source for quality: relevance, authority, structure, and topic alignment.

We’ll get on to this topic more closely shortly, but it’s worth noting that schema markup, semantic HTML, and a clean and sensible site architecture all influence how likely it is that your content will be referenced and cited in AI responses generally.

Synthesis and Citation

After all the research is done and dusted, the final step is for Gemini to gather it all together in a cohesive, coherent and useful summary answer.

It’s important to remember, as mentioned above, that the top cited sources aren’t necessarily those that rank at the very top of the traditional SERPs.

A page ranking at the lowly position of seven that gives a clearer, more direct answer to a sub-question is liable to be cited ahead of the page at position one.

How Are AI Overviews Impacting Organic Traffic?

It goes without saying that AI summaries appearing in the most prominent position on the SERPs is having a measurable impact on organic traffic. And while Google has been a little coy on the topic, we’ll dig into some more independent data to paint, hopefully, a truer picture.

Ahrefs found that the presence of an AI Overview correlates with a 58% lower click-through rate for the top-ranking page (December 2025 data, updated from a 34.5% drop in April 2025).

Seer Interactive's most recent study on the impact of AI Overviews on organic CTR across 5.47 million tracked queries and 2.43 billion organic impressions. It found that while CTRs on organic search results when an overview is shown have recovered vs. 2025, it still lags significantly behind the control.

A randomised field experiment by researchers at the Indian School of Business and Carnegie Mellon University found AI Overviews reduced organic clicks by 38% on triggered queries, with no measurable improvement in user satisfaction.

Of course, these studies use differing methodologies, and there are some surprising results, but it seems that the direction of travel is unanimous. Organic clicks where AI Overviews also appear are down.

What Google Says (And Doesn't Say) About AI Search

On 15 May 2026, Google published its first official guide to optimising for generative AI search.

Unsurprisingly, this was considered quite a big moment in an industry that has been making - usually educated - guesses about what influences AI Overviews. And now there’s an official position - from Google at least. 

AI Search is Still Search 

The big headline from what Google has said is that AI search is still search. 

This is important because much of the discourse around AI search has been that it sits somewhat apart from traditional search. As it turns out, that isn’t quite the case. Though there are a couple of asterisks that we’ll get to later. 

So, depending on your own initial position, Google’s update could be one of reassuring or frustrating. Reassuring because the fundamentals of SEO still matter. Frustrating because plenty of the GEO advice going around over the past 18 months turns out to be redundant. 

What Doesn’t Help 

One of the most useful parts of Google’s guide is what it rules out: 

Schema markup isn’t required specifically for AI search visibility. That said, it’s still absolutely worth implementing, especially for rich results in regular search. But it’s just not as important as many GEO articles have made it out to be. 

LLMs.txt files (similar to robots.txt, but for AI) currently do little to nothing for AI search. In fact, Google ignores them. So, until something changes, you can remove them from your to-do list. 

"Chunking" content to make it easier for AI to understand isn’t a thing. Google understands nuance and context across full-length pages. That said, good content structure remains vital. 

Inauthentic mention-building doesn’t help. In fact, it probably hinders. Google has spam filters for SEO, and those policies also apply to its AI features. 

What Does Help 

As you’ve guessed by now, a lot of what helps in AI search matches SEO best practice: 

Unique, first-hand experience content. Google understands the difference between commodity content and content that offers different perspectives based on experience. Lean into the latter. 

Technical hygiene. Indexable, crawlable, semantic HTML. Everything needs to be accessible and readable. It goes without saying, really. 

Multimodal content. Images and videos get surfaced in AI responses, so visual assets matter more than they used to. 

E-E-A-T fundamentals. Experience, expertise, authoritativeness, trustworthiness. These signals haven’t changed and remain absolutely essential. 

Caution: This is Google-Centric 

There’s a reason everyone looks to Google on topics like SEO and AI search. We needn’t go into why. But that also makes a lot of circulating content Google-centric. 

This article is specifically about the Google ecosystem. It is not about the likes of ChatGPT, Claude and Perplexity (etc…). The departure is unlikely to be huge, but there could be differences. 

Either way, just because an action doesn’t move the needle in Google AI Overviews, it’s our position that current best practice and delivering the highest possible standards is always the best path. 

Schema, as an example, may not be as important as it has been made out to be. But it’s still worthwhile deploying it as accurately and thoroughly throughout your site as possible.

Why Being Cited in AI Overviews Matters

The first thing to say here is that there’s really no choice. There’s no point raging against the impact of AI on organic search performance. It’s you who needs to adapt.

It's demonstrably true that AI Overviews have an impact on clicks. But they can still provide visibility for your brand. In fact, it’s pretty easy to argue that being mentioned prominently in an AI Overview is a better outcome than just appearing in the top blue link.

5 Reasons Your Brand Isn't Being Cited by AI

So, what do you do about all of this?

While optimising for AI visibility is, in many ways, still a bit of a black box, we can say with certainty that it’s not time to throw out the SEO rulebook just yet.

Getting cited by AI - including in AI Overviews - and appearing at the top of traditional SERPs are in many ways two sides of the same coin.

This is actually a topic we’ve gone into some detail in our article about GEO and web design. So, instead of rehashing what’s already there, we’ll instead look at the commonest factors standing in the way of you being cited in AI Overviews.

1. Your website isn't crawlable or indexable by AI

This one is a surprisingly common failing. Many websites simply aren’t able to be properly crawled, indexed and understood by AI.

It doesn’t matter how good your content is if Google (and other LLMs) can’t access it, then you’ve got a big issue. The usual culprits are:

Robots.txt blocking access. This one’s a bit of an own goal. A misconfigured robots.txt file can block the likes of Googlebot, GPTBot, ClaudeBot, PerplexityBot, or Google-Extended from your site. Sometimes it's deliberate (a staging directive). Sometimes it's inherited from a previous build. Either way, the result is the same: invisible content.

Hosting blocking access. This isn’t a problem with all hosts, but it is something worth checking as it’s surprisingly common. For example, Cloudflare flipped its default to block AI crawlers for new domains, with one-click blocking available to existing customers.

Now, it’s worth saying that if you’re worried about your content being scraped for model training, it’s a pretty handy feature. But it’s a fairly unhelpful one if your goal is to be cited in AI Overviews.

Accidental noindex tags. It’s fairly common for pages to get launched with noindex tags left over from development. It’s human nature to forget stuff, so it’s worth checking. A quick crawl of your site in Screaming Frog will surface any pages telling search engines to look away.

Poor site structure. Just because there’s content on your website and you’re not blocking any crawlers, doesn’t mean Google or AI will find, understand, and index it.

If your site is three layers deep, has poor internal linking, broken links here and there, and orphaned pages that can’t be found, you’re just making everyone’s life difficult.

Client-side rendering. Bots are not humans. If your content only appears after JavaScript executes, you're kind of hoping that it’s rendered correctly by every bot that visits. Googlebot will probably do a pretty good job, but AI crawlers are much younger and less reliable.

Luckily, fixing all of this isn’t hard. You need to audit your crawlability. Check your indexing in Google Search Console. Validate your robots.txt and sitemap. Check your hosting/CDN settings. Make sure you’re not hiding your on-page content.

2. Your content is hard to extract

Let’s expand on one of the points from above: AI Overviews favour content that gives clear, direct answers. If your blog posts read like meandering essays that finally get to the point in paragraph six, you're making it harder for the system to find what it needs.

The fix isn't to dumb things down. It's to be more disciplined. Lead with the answer, then expand. Use headings that mirror human search queries. Add an FAQ section. Define things clearly and early. The content can still be substantial and in-depth, but it needs structure.

This is also where schema markup is worth a mention. Schema tells machines what your content is, not just what it says - that this is a product, this is an FAQ, this is an author, this is a review. 

Google has now clarified that schema isn't strictly required for AI Overview visibility, which is useful to know. 

But it's still worth implementing: it helps with rich results in traditional Search, and other AI platforms - ChatGPT, Claude, Perplexity - may rely on it to parse your content reliably.

💡 Further reading: The Role of Schema Markup in SEO & AI Search (2026 Update)

3. You're not mentioned enough

Ahrefs has found that unlinked brand mentions correlate more strongly with AI Overview inclusion than backlinks do. That's one of the more significant shifts in an arena where backlinks have been, more or less, king.

If your brand isn't being talked about in reviews, podcasts, listicles, and credible third-party publications, AI systems have no signal that you're a recognised name in your industry. Backlinks alone are no longer enough; you need to be playing a credible part in conversations across the web.

The important word there is 'credible'. Don't chase inauthentic mentions, as this can have the opposite effect. Google has been open and clear about how its core ranking systems focus on high-quality content (see below). 

Google's recently published guidance explicitly calls out inauthentic mention-building as something that doesn't help. 

Coordinated campaigns, paid placements, and any of the slightly desperate/grey-hat tactics that have started circulating in GEO circles are filtered out by the same spam systems that govern traditional Search.

4. You're not prioritising long-tail keywords

We’ve already covered the fact that AI Overviews most commonly trigger for long-tail, conversational, comparison-style queries. The sort of keywords you never would have prioritised in search 5 years ago.

But it’s not just about when AI Overviews appear. This change also reflects how people are searching both within AI and starting to search within Google itself.

Search queries are getting longer and more specific, as well as more conversational. Long gone are the days when everyone was searching in 1-3 keyword-heavy phrases like ‘best CMS’. These days, you’d be much more likely to search for ‘what’s the best CMS for a mid-size business looking to grow quickly?’

The implication here is kind of simple to explain. Basically, don’t write for the searches of 15 years ago. Prioritise the long-tail. This is the content most likely to be cited.

5. You're not creating quality content

This one feels a bit basic and obvious. And it’s pretty much always been the case. But creating quality content is arguably more important now than it has ever been.

What has changed is the bar for what constitutes quality content. It’s much, much higher than it used to be. And it’s important to be aware of what’s required to perform well across traditional SEO and AI-generated answers.

Avoid producing thin content that contains a lot of AI filler. You want your website to be considered a trusted, authentic and, as much as possible, original source.

It isn’t about word count, though it helps, given the importance of topical depth. What’s far more important is first-hand experience, real analysis and research, original data (if feasible), and quotations from real people.

Now, we’re not saying don’t use AI in your content workflow (Claude was involved in this article). That seems silly, and there’s nothing wrong at all with using AI tools to research, draft, or even edit. But avoid at all costs publishing AI slop without effort, checks or balances.

💡 Further reading: Web Design & SEO: Building a Website That Ranks in 2026

How to Track AI Overview Performance

Unfortunately, Google doesn’t distinguish between AI Overviews and plain old organic search in Google Search Console. So, we have to go about this in a different way…

Our SEO tool of choice is Ahrefs, which lets you filter your organic keywords by SERP feature, including AI Overviews. For our money, that’s the easiest way to see which of your keywords are triggering AIOs and how that correlates with traffic changes.

Semrush has a similar feature if that’s more your cup of tea. And if you don’t have access to these tools directly, you can just do some manual query searching yourself.

What AI Visibility Means for Web Design

Again, this is a topic we’ve covered in greater depth in our GEO article. However, it’s absolutely worth repeating as a lot of agencies and businesses aren’t getting it quite right (yet).

A huge proportion of GEO and AI search advice focuses on content. Write better content. Give your content more structure. Build as many mentions of your brand as possible. All of that is, of course, true and useful.

But the foundations beneath that content matter just as much, if not more. And as web designers and developers, it’s where we’re most focussed. All that is to say that build matters, specifically:

Performance. Slow pages annoy everything and everyone: Google, LLMs, and your users. To be avoided at all costs!

Accessibility. The same things that make your pages accessible to screen readers (clear hierarchy, semantic structure, alt text, descriptive labels) make it easier for AI to parse.

Crawlability. As we’ve covered in-depth above, if AI crawlers can't access your content reliably, it can't be cited, or its citability will drop.

Structured data. Schema is more essential than it has ever been. Learn about it, understand it, and deploy it to give crawlers more contextual information about your content.

💡 Further reading: How to Build an Accessible Website in 2026

💡 Further reading: How AI is Changing Web Design and Development

SEO Isn’t Dead. But It Is Different.

SEO fundamentals aren’t being eroded or replaced. They remain fundamental. But optimising for citations in AI Overviews should be seen as a layer on top of them. You can’t just sit back and assume your organic rankings will turn into AI visibility.

Put simply, rankings still matter; they just no longer guarantee a click (if they ever did). Content still matters; it just needs to be better, more structured, more machine-readable. Similarly, authority still matters, but it isn’t all about backlinks any more.

If they weren’t already, thin sites and those that underperform technically are now being deprioritised across the internet. Ultimately, if you want your website to be visible this year and in the future, start working with the best practices mentioned above.

More than ever, you can’t cheat your way to success. It must be earned.