It often begins with a glance at analytics data: visitor numbers are declining. Sometimes slowly, sometimes abruptly. No obvious trigger is apparent. No major Google update, no new competition, no fundamental change in content. Yet the curve points downward.
Looking more closely, you often discover a second phenomenon: scattered traffic from ChatGPT, Claude, Perplexity, and other AI systems. But their volume rarely comes close to explaining the drop in search engine traffic in a way that suggests users are now taking the AI route instead. The real change is not that users have stopped using Google. Rather, it's that Google itself is becoming AI search.
Google increasingly delivers the answer itself
The data already clearly shows where this is heading: according to an analysis by Seer Interactive, organic click-through rates for search queries with AI Overviews are on average 61 percent lower than for comparable searches without AI summaries. At the same time, many websites show another pattern: visibility in search results remains stable or even increases, while the number of actual visitors declines.
At first glance, this seems to confirm the AI migration thesis. In fact, these numbers reveal something quite different: the user often doesn't leave Google at all. Because Google delivers the answer directly. What previously appeared as a list of links leading to answers increasingly becomes the answer itself. The search engine is evolving into an answer engine.
This distinction matters, because it shifts perspective. The central question is no longer:
Google versus ChatGPT.
But rather:
Link lists versus direct answers.
Meanwhile, a new, rapidly growing channel is emerging. According to Similarweb, AI platforms generated more than 1.1 billion referrals to the world's 1,000 largest websites in June 2025 – a 357 percent increase year-over-year. Measured against total search traffic, it's still a small share. Yet the momentum demonstrates that a second ecosystem is establishing itself alongside classical search.
For many companies, this creates a situation that initially seems contradictory:
- Visibility remains.
- Clicks disappear.
The middle of the customer journey vanishes
Here lies the real shift. Previously, a typical purchase decision consisted of many small intermediate steps:
- Search query
- multiple search results
- comparison articles
- reviews and test reports
- product pages
- ratings
- purchase decision
Each step generated traffic. Today, AI takes on a significant portion of this work. The user states their problem, receives a pre-filtered selection, and often gets an explanation for why certain providers are better suited than others. The actual research increasingly happens within the AI itself. This fundamentally changes the customer journey. Not its beginning. Not its end. But its middle. That's where many clicks used to come from. And that's exactly where they're disappearing now.
Fewer clicks don't automatically mean less revenue
One aspect is often misinterpreted: when AI handles the pre-selection, the quality of remaining visitors changes. Someone who lands on a website following an AI recommendation is often already much further along in the decision process than a typical search engine user. The visitor no longer comes to get oriented for the first time. They often come to confirm an already prepared decision. For companies, this means: traffic can decline while the purchase intent of remaining visitors increases. This is why one metric will likely lose importance in coming years:
How many visitors come to my website?
And another will become more important:
What's the likelihood that visitors become customers? And how can I as a company influence this?
Reach remains relevant. But reach alone loses its explanatory power.
The new task: providing arguments to AI
This is where the real strategic challenge begins. In classical digital marketing, it often sufficed to be found. The user then did the evaluation. They read product descriptions, compared providers, and formed their own opinion.
In the AI world, the process works differently. AI already creates a pre-selection and often provides a reason:
Provider A is particularly suitable for ...
Provider B is cheaper but has ...
Provider C is the best choice if ...
These exact explanations are the real value for the user. And here lies a new requirement for companies. AI can only work with the arguments it finds. In the future, it's no longer enough that AI knows that a company exists. It must also understand:
- What it's particularly suited for.
- What problems it solves.
- Where its strengths lie.
- What limitations it has.
- Who it's suitable for – and who it's not.
You might say: previously, a company had to be findable. In the future, it must be defensible to AI.
Why small companies can benefit from this
This development carries not only risks. It also opens opportunities. In the classical Google world, large brands, strong domains, and companies with substantial SEO budgets often dominated. AI systems follow at least partly a different logic.
A recent scientific study of over 55,000 search queries shows that nearly 30 percent of sources cited in Google AI Overviews don't even appear on the first search results page. Source selection therefore isn't exclusively determined by classical rankings. This doesn't mean authority disappears. Well-known brands and established sources retain considerable advantages. But the barriers to entry are changing.
A specialized provider with clear positioning can today appear in an AI answer even though they would never appear on Google's first search results page. For many small and medium-sized businesses, that's a rare opportunity. Not because size suddenly doesn't matter. But because specialization becomes more important.
Winners and losers
This development doesn't affect all market participants equally. Business models whose value creation lies between question and decision face particular pressure:
- Comparison portals
- affiliate sites
- pure information portals
- parts of classical publisher business
AI systems are increasingly taking over precisely these functions.
It's different for companies that sell products, services, or expertise. For them, less traffic can very well come with higher visitor quality. The same mechanisms therefore create both winners and losers simultaneously.
The real shift is trust and authority
Behind all technical developments stands ultimately a cultural change. Users once trusted search engines, test reports, and comparison portals. Today, many people increasingly trust an AI's recommendation. Not because AI is necessarily right. But because it provides a reason. It functions like a personal research assistant that gathers, evaluates, and synthesizes information. This creates a new authority alongside search engines, media, and experts. And authority isn't defined by reach. It's defined by trust.
What companies should understand now
The decisive question is no longer:
How do I prevent traffic loss?
But rather:
Why should an AI recommend my company?
Whoever can answer this question has truly understood the real change. Because clicks will very likely not return to previous year levels.
The direction of development now seems clear: more and more search queries will be answered without users needing to visit external websites. Pre-selection is shifting into search and AI systems. Actual research is increasingly outsourced. This fundamentally changes the task of marketing.
It's no longer enough to be visible. Companies must provide information from which compelling arguments can be derived. Those who understand this early can benefit from this development. Those who continue to optimize exclusively for reach are fighting a change that can no longer be reversed.
Sources
- Seer Interactive: Analysis of Google AI Overviews' impact on organic click-through rates (2025)
- Similarweb: Development of AI referral traffic to the world's largest websites (2025)
- Xu, Iqbal, Montgomery et al.: Measuring Google AI Overviews: A Large-Scale Study of Source Selection and Ranking (2026)
- Search Engine Land: Evaluation of current CTR data and AI Overview effects (2025)