AI Overviews Are Reshaping the Customer Journey: Can SEO Keep Up?

With AI Overviews and large language models now answering user queries directly on the search results page, we’re witnessing a major shift in how people search — and how websites are (or aren’t) found. In this article, I want to take a closer look at how LLMs in general — and AI Overviews in particular—are affecting user behavior and what that means to website owners and the way we approach SEO.

But before we dive into what’s changing, let’s take a step back and see what used to work best in SEO. What was the most effective strategy to drive traffic and conversions to a site?

In SEO we differentiate between navigational, transactional and informational search queries.

• Navigational: searching for a particular website or document – e.g. “youtube”
• Transactional: searching for a product or service with the intention of making a transaction – e.g. “brooks running shoes man size 43 on sale”
• Informational: searching for information, early in the decision-making process – e.g. “how many kilometers will running shoes last”.

One of the most effective strategies for SEO in the last 10 years has been creating as many touchpoints during the decision-making process as possible and getting people onto a particular website as early as possible. We did this by creating valuable and helpful content, often in form of blog posts, and getting it ranked on top of the search results.

Until now, most of our focus has been on informational search queries and our main KPI was clicks. We wanted to get as many of them as possible. Of course, our final goal was to get conversions, or monetizing traffic in some other way, but the foundation was clicks.

This strategy worked incredibly well for many years and withstood many changes and challenges.

The integration of featured snippets above the organic results on Google was introduced about 10 years ago. It marked the beginning of a new era, where Google started trying to answer many search requests directly, without necessarily sending people off to websites. The number of zero-click-searches has been on the rise ever since. People’s hunger for information was often satisfied directly on the SERP, without requiring any further click or action.

Some website owners suffered more than others. But overall, traffic from Google remained the prominent and most important traffic source by a huge margin.

SEO experts adjusted their strategies, we started optimizing for getting our content into those featured snippets, but overall, things stayed the same. For the most part, we continued doing SEO like we did 5, 10 or even 15 years ago. We built a keyword strategy based on search intent and volume, created great content around those keywords, made sure we checked all the technical boxes and finally strengthened the authority and trustworthiness of the domain by building some valuable backlinks.

In May 2024 Google introduced AI Overviews in the USA, followed by several other markets before rolling it out in German speaking countries in March 2025. This constituted Google’s response to ChatGPT and other LLMs, which had been increasingly luring users away from Google, and represented an effort to shift from being a search engine to being an answer engine. The resulting change to search behavior as well as the effects on organic traffic was immediate and huge.

Google was not holding back rolling this out. If anything, they initially showed AI Overviews for more searches and then rolled it back to some extent, only to increase the share again, once they had collected more user data. Currently, studies of large sample groups estimate that AI Overviews are triggered for around 13% of all search queries, 80% of those being informational searches.

Also, the appearance of AI Overviews is strongly dependent on the industry, or topic of the search queries. Sectors like Science, Health, Law, Travel, and Education are seeing particularly high AI Overview visibility, while categories such as Shopping or Real Estate are still much less affected.

AI Overviews tend to appear especially often for searches related to explanations, decision-making, or technical troubleshooting—e.g. “How does…work?”, “Is it dangerous when…?”, “What’s the best time to…?”, etc.

Those numbers certainly indicate that people are more than satisfied with this new feature. Here are some of my reflections on why AI Overviews are such a big hit with users:

• Instant gratification: answers without requiring a click.
• 10 blue links offer a list of choices while AI Overviews offer an overview as well as context.
• Many top-ranking pages offer near-identical, heavily optimized content/SERPs – AI Overviews cut through that noise and provide a “cleaner” summary.
• AI Overviews shift the dynamic from clicking and reading to asking and understanding.

From a personal point of view, I can say that within weeks of the roll-out, my perception of the 10 blue links on the SERPs changed from being the best entry point to the world’s knowledge, to being somewhat boring and outdated. AI Overviews offer very good user experience in many cases, much better than the good old list of 10 links.

The most apparent development – and from a publisher’s perspective the most concerning – is that click-through rates drop significantly when AI Overviews appear in the search results. Some studies report drops of over 50%. This is in line with what major media publishers report. Some of them see a decline in organic traffic of 40–55%.

This development is also reflected in Google Search Console data for domains heavily affected by AI Overviews. Impressions often remain stable – or even increase – while clicks continue to decline. This growing gap between impressions and actual clicks is referred to as the Crocodile Mouth Effect.

The impact becomes clear when we look at real search behavior. For instance, when performing a search on a mobile device, users are presented with a concise AI Overview summarizing key facts at the top of the SERPs. Studies show that around 86% of users only skim these AI Overviews when searching for quick facts. On average, they spend 30 to 45 seconds engaging with the AI element, often without ever scrolling further or clicking on external links.

Even when the AI Overview is truncated, 88% of users click on “show more”. Despite that, the median scroll depth remains at just 30%, meaning that most users do not even reach the classic organic search results.

Interestingly, the domains most frequently included in AI Overviews are not always identical to those dominating classic organic search results. Health-related websites (e.g. netdoktor.de) appear particularly often in AI Overviews, as do general information portals and media platforms such as utopia.de, t-online.de, br.de, or focus.de. Many of these brands are also well-established in organic rankings – yet their performance within AI Overviews follows its own dynamic. As the effects of AI Overviews vary significantly between different domains, every domain needs to be analyzed individually, and averages are of limited value.

However, a strong presence in AI Overviews is a double-edged sword. On the one hand, it reflects high trust – these domains are considered reliable, authoritative sources. On the other hand, it often means a significant amount of their content is summarized directly within the AI Overview, resulting in fewer reasons for users to click through to the actual website. Visibility remains – but traffic does not necessarily follow.

Everybody is clear on the fact that click-through-rates are dropping. At the same time Google reports an increase in users and search queries on their platform when AI Overviews are present. While this is certainly good for Google, it is little consolation for publishers who have lost half of their traffic.

This is where a fine balance has been disturbed: publishers created great content and let Google crawl it, index it and rank it for free, in return for clicks. Most publishers continue to make their content available – often free of charge. AI uses this content for training, real-time information retrieval, summarization and showing that summary in the search results.

However, in this new world, publishers are not getting any more clicks in return. As a result, some are asking why they should continue creating valuable content, if it is not doing anything for them in terms of traffic.

This is certainly an issue that needs to be resolved sooner rather than later. Personally, I doubt that the market will sort this one out by itself. I think some legislative intervention may be called for in this case.

One hypothesis (that has not yet been backed with enough data to prove it) is that while traffic is down, the quality of that traffic is considerably higher and converts better. In many cases, when users arrive on websites, they are better informed than they have ever been. They already know what they want. They have had lengthy conversations with ChatGPT, discussing the pros and cons of various products and services, divulging their most personal information, interests, eating habits, training plans, room layouts, financial situation, health issues and all kinds of likes and dislikes.

In the last few years, a good part of the customer journey has already moved away from Google, or better: from websites. People watch tutorials and inspirational content on TikTok. They read what other users say on Reddit. But LLMs take away even more of it.

The learning for SEO experts and website owners is that if we only focus on being present in the top search results on Google, we are definitely missing out. Organic traffic from Google may still be the channel that brings most of the conversions. However, a conversion can only happen if our target audience has already built a relationship with our brand, being exposed to it on many other platforms during the inspiration and research phase.

We have been getting personalized search results for many years, depending on our language settings, physical location, logged-in status, interests and recent browsing behavior. The local pack is based on location. Ads are targeted via clever AI in Google Ads. The ratio of information sites vs. shops is tilted in favor of our preference. Sites that we have recently visited – like our own website – will rank on top, even if it ranks on page 2 for other users.

But all this was quite limited from today’s perspective. It was still possible to look at neutral, un-personalized results, either directly (VPN, incognito browser window) or through SEO tools. Tracking those rankings still made sense. It gave us a good idea of what most people out there would see, when they searched for a particular keyword.

However, with AI overviews, and searches on ChatGPT and other LLMs, this is not the case anymore. Of course, people still use keywords, or prompts, to search for information. The question is will the AI retrieve information based on those keywords?
The answer is kind of no, because of the so-called fan-out technique. Here is what Google says about it:

If I conduct a simple search for “running shoes”, the AI may fan out quite a bit, based on what it knows about me.

• Best running shoes for hard concrete (I live in the city)
• What running shoes are best to mitigate knee problems (I’m in my mid-forties and my knees are bad)
• Comparison of the top-rated running shoes regardless of price (I’m an agency owner, using the latest iPhone, so I can probably afford more expensive shoes.)
• Most suitable running shoes for 10 km runs at moderate speed (yes, my phone knows that too)
• And the list goes on

After that, Google sources relevant content from its index based on these sub-queries, selects the most meaningful text-segments based on relevance and generates the answer with cited sources.

To improve the results for future searches, user interactions like clicks, scrolls, or follow-up questions are tracked. AI Overviews with high user acceptance influence which sources and answer paths will be favored for future searches.

As a result, I get a very different reply from my search for “running shoes” than other people.

So how are we supposed to track a particular brand’s rankings for this query? What does “neutral ranking” even mean – if no one gets neutral results anymore?

Will the AI divulge the information of the fan-out process to us, so we can at least do a proximation of keyword research?

This may be the end of keywords as we know them.

AI is fundamentally changing how people search, and how information is presented to them. With that, a good part of our skill-set and tool-stack will become obsolete in the very near future and the strategies and methods we have used to optimize documents to rank well in the past will also have to change.

There are already different strategic approaches for how companies can respond to the rise of AI Overviews – ranging from passive acceptance, active cooperation and open opposition to completely avoiding AIO-triggered search fields and building alternative traffic sources. The most effective ways to maintain visibility, brand presence, and control are active participation or strategic avoidance. Public opposition may increase pressure on Google but won’t bring back lost traffic in the short term. The riskiest approach is passive acceptance – yet, most websites are likely to follow that path.

Because this is such a new market, there is very little data on what works and what does not. It is a bit like the very early days of SEO, where testing was the only road to success. Like with any big challenge, it is also a huge opportunity. SEO experts and website owners who see this opportunity, who test, learn and adapt their methods, will most likely come out on top.

Here is an overview of how we are adapting.

• What keywords trigger AI Overviews and how often does a brand appear in them?
• Log file analysis: what content do the AI bots crawl?
• Traffic analysis
• Impact analysis: what pages / content are most affected by a loss of traffic?

• Definition of a set of keywords and prompts that are important along the customer journey
• Set up tracking for relevant AI agents
• Monitoring of brand mentions and citations of our own brand and of those of competitors
• Use the data to build a strategy and clear action items

• Simulate AI Overviews with a custom Gemini 2.5 setup to better understand content visibility and source selection
• Analyze how AI breaks down queries (Fan-out) and which sub-topics trigger citations
• Use simulation insights to systematically improve the likelihood of being cited within AI Overviews

• Content optimization: FAQs, abstracts, structure (clear answers, W-questions, visible authorship, and strong formatting), citation of sources
• Structure: internal navigation, Schema.org
• Technical: crawlability, LLMS.txt

• Strengthen E-E-A-T factors: social proof, trust elements (author profiles, etc.)
• Off-page SEO: citations, mentions, lists and directories, Wikipedia, PR
• Social media: engagement in relevant communities

• Offline
• TikTok
• Social Media
• Reddit
• Owned channels (e.g. newsletter)

In this new world of AI, our focus is moving away from creating content solely for the purpose of getting clicks. Instead, we need to focus on content that delivers real value and builds visibility, authority, and relevance across all platforms. After all, SEO remains the foundation – it is still the ticket to enter AI Overviews. However, if we want to stay visible and competitive in the long run, we must expand our strategies to stay present in this new AI-driven world.