Measuring AI Search Visibility

Google’s AI Overviews are currently one of the most visible changes in search behavior for the general public. They appear for selected search queries and provide an AI-generated summary directly in the Google interface. This usually shows links to the sources, but often users have already obtained their information and do not click further.

According to Sistrix, an AI Overview was already displayed for 9.3 percent of all Google search queries in Germany in June 2025. In the UK, the figure was almost 24 percent at that time. The trend is exponentially increasing for all countries. Crucially, around 75 percent of the linked content in AI Overviews comes from the top 12 organic hits. This means that classic SEO rankings remain relevant for the time being, but are no longer sufficient on their own.

Large language models (LLMs) such as ChatGPT and Perplexity are increasingly becoming alternative search engines. They process search queries based on context, provide individualized answers, and sometimes integrate live web data. More and more users are starting their research directly in these interfaces without taking a “detour” via Google. AI browsers such as Comet could soon dominate the internet, especially among younger target groups.

The number of monthly sessions with AI assistants rose by 180 percent between April 2024 and April 2025. While ChatGPT continues to dominate, tools such as Gemini (Google), Claude, and DeepSeek are also gaining market share. Gartner predicts that companies could lose up to 25 percent of their organic search traffic by 2026 if they fail to respond to this development.

AI is not only changing access to information, but also the purchasing decision process. The classic customer journey—consisting of awareness, consideration, and conversion—can be shortened through generative responses.

Instead of clicking through many pages, users often receive a specific product recommendation or problem solution directly in the chat. This means that many classic touchpoints are eliminated, especially in the upper and middle funnel (TOFU and MOFU). Affiliate models, ad-supported comparison portals, advice sites, and other websites in particular are losing visibility and reach.

At the same time, there are currently many signs that the quality of remaining website visits is increasing: those who visit a page after an AI recommendation usually have a clearer intention and are further along in the decision-making process. Initial data indicates that the conversion efficiency of AI traffic is higher than that of traditional channels.

Classic metrics such as “clicks on search results” alone are no longer sufficient. They must be enriched with other, new metrics in order to obtain a complete picture of brand perception. This is the only way to determine online attribution to purchasing decisions in the best possible way. Today, for example, it is crucial whether your content appears in AI responses, even if no one clicks on it.

Our assumption about the development is as follows:

The following KPIs can be a good starting point for getting familiar with the topic:

Generally, KPIs should be divided into two categories:

• AI visibility metrics: before visiting the website
• AI website performance metrics: from the moment the website is visited, triggered by AI

More detailed KPI definitions and further details on this topic can be found in our blog post “New KPIs for website measurement.”

During the webinar, we received various questions about AI Search Visibility. We were able to answer some of these questions directly during the live event, while others were clarified individually in personal conversations afterwards. Since many of the topics are highly relevant for marketing decision-makers, we are answering them again here in a collective format. This way, everyone who was unable to attend live can also benefit from the insights and recommendations.

Our prediction is that websites will continue to be relevant in the short and medium term. Although they will receive fewer direct visits due to the changes described above, they will remain important: both for AI crawlers and for content that builds trust. Without high-quality content, there can be no meaningful AI responses. Content remains the basis of all AI-driven visibility. In addition, LLMs can already be linked to online shops, so that users can find the information they need via Chat GPT & Co., but then make the actual purchase via the website. This could mean that in some cases, the user interface of websites is no longer as relevant, but the backend systems with the content still remain very important.

The best way is to use a personal example: How has your own search behavior changed? Many decision-makers have already noticed that they are using tools such as ChatGPT or Perplexity more often instead of Google. After that, the slides and recording of our webinar can help to delve deeper into the topic in a concise manner and anchor it strategically within the company. A short internal workshop can also be helpful.

Create your own AI marketing channel. In GA4, you can use regular expressions (Regex) to capture referrers such as chatgpt.com, perplexity.ai, or gemini.google.com.

In Adobe Analytics, create a new channel rule in the Marketing Channel Manager. It is important to place the rule before the general “Referral” category. This ensures that AI traffic is assigned correctly and can be analyzed separately.

Adobe Analytics offers the option of defining your own marketing channels based on referrer logic. For AI traffic, it is recommended to create a new channel rule that recognizes domains such as chatgpt.com, perplexity.ai, or gemini.google.com. It is important to place this rule in the Channel Manager above the general “Referral” rule. This ensures that visits from AI responses are recorded separately. In Workspace Reports, you can then see how this traffic behaves, which pages are visited, and how well this content performs.

That depends on the volume of traffic and the role of search for your brand, as well as your individual business and marketing goals. If you rely heavily on organic traffic or already appear in AI overviews, you should update daily. In other cases, weekly updates are sufficient. It is important to establish a fixed process for identifying and evaluating changes in a timely manner.

AI sources such as ChatGPT or Perplexity usually leave a specific referrer that is visible in the web analytics tool. ChatGPT, for example, appends utm_source=chatgpt.com to every link. However, these referrers must be actively tracked and categorized, for example in a separate channel called “AI Traffic.” With Google AI Overviews, the distinction is more difficult because they are displayed within Google Search. Here, a change in click behavior with the same ranking can be a signal, e.g., more impressions but fewer clicks than before.

Yes, there are now tools that specialize in exactly that. They analyze whether and how often your domain or brand name appears in AI responses, e.g., in ChatGPT, Perplexity, or Gemini. This so-called “zero-click presence” is important because it shows whether you are generating visibility despite a lack of clicks. Tools such as Peec or Otterly help to systematically track such mentions. This is particularly relevant for brand building, thought leadership, and awareness campaigns.

AEO stands for Answer Engine Optimization and aims to prepare content in such a way that it can be easily processed and adopted by classic AI overviews (such as Google). GEO, on the other hand, stands for Generative Engine Optimization and goes one step further. This involves optimizing content for LLMs such as ChatGPT or Perplexity. This means that the content is semantically clear, prompt-compatible, and thematically relevant. Both approaches pursue similar goals, but with different technical and content requirements. For many companies, it is worthwhile to consider both, especially if both organic search traffic and AI-supported discovery are relevant.

A good example is users who arrive at your website via AI recommendations. These people often come with a clear interest in information. If you notice that they arrive at the page but hardly scroll or bounce immediately, this can be a sign that the content does not meet their expectations. Conversely, a high scroll depth can show that your content is relevant and that people are continuing to read. This allows you to measure which content is really effective in AI-driven sessions and which you should optimize.

Many companies are currently faced with the question of how to respond to changes in search behavior. Our clear recommendation is: don’t wait, but start now with initial measures to systematically record and improve the visibility and impact of your content in the AI context. This will allow you to learn and optimize step by step.

We suggest the following steps to get started with the topic:

• Analyze impressions and clicks for your top keywords.
• Manually check whether your content appears in AI overviews.
• Set up an AI traffic channel in GA4 or Adobe Analytics –> You can obtain an Adobe Analytics referrer list and the GA4 RegEx list (as of July 2025) from us free of charge on this page.
• Use new tools to measure your AI visibility
• Define clear KPIs for GEO and AEO initiatives
• Optimize your content for user intent, structure, and timeliness

Our webinar “New KPIs for AI Search” by our Senior Digital Analytics & AI Expert Constanze Bogner was aimed specifically at marketing managers who are concerned with how AI Search changes the visibility and measurability of their website. The focus was on new KPI logic, practical examples, and specific recommendations for GA4 and Adobe Analytics.

You can download the recording of the webinar, the slides, the Adobe Analytics referrer list, and the GA4 RegEx list free of charge here: