Does AI recommend your products?
Perform a free audit.
Generative search changes the rules of the game - AI Search (e.g. ChatGPT, Gemini, Perplexity) generates response results instead of a list of links, which means that traditional metrics such as last-click conversions no longer fully capture the business impact of this channel. In previous articles, we showed how visibility in AI differs from classic SEO - now it's time to go a step further and show how to measure the effectiveness of such visibility in GA4 and other tools.
Introduction
For many companies, the first months of activity around AI Search come as a surprise: traffic is there, but sales are few. In practice, AI Search often assists a purchase decision earlier than it closes it in GA4. Therefore, reporting on effectiveness requires properly configured KPIs and analytical segments.
In this article, we will discuss:
- how to accurately track traffic with AI Search
- what KPIs matter at each stage of the funnel
- how to set GA4 to better understand the impact of AI
Standard GA4 reports are not enough
GA4 shows traffic sources, sessions and transactions. However:
- AI Search often initiates interest, and purchase decisions are made in another channel
- traffic from AI models can be classified as "Direct" or "Referral" without clear designation
- many AI-related conversions are not included in the traditional attribution report
In practice, the lack of sales in the first few months does not mean there is no impact of AI - rather, it indicates that we do not analyze the relevant metrics.
How to identify traffic with AI Search in GA4
The key is to create dedicated AI traffic definitions. Here are the practical steps:
1. Make sure AI sources are properly labeled
In GA4, it's a good idea to set up custom channels where you assign traffic from addresses such as:
- chat.openai.com
- chatgpt.com
- openai.com
- gemini.google.com
- perplexity.ai
- grok.com
With this:
- aI traffic will not be confused with other channels
- more easily determine AI's contribution to user paths
2. Mark events specific to Ai
Create events such as:
gtag('event', 'ai_session_start', {source: 'AI Search'});
gtag('event', 'ai_engaged_session', {engagement_time_msec: ...});This allows you to analyze sessions in more detail than just by source.
3. Set segments
The "AI users" segment should include:
- users who have entered at least once from an assigned AI source
- users who had interactions specific to AI Search (e.g., sessions from chatbot responses)
These segments enable:
- comparisons of AI vs. other user behavior
- analysis of their conversion paths
KPIs that really count
In the context of AI Search, it is useful to think about KPIs in a multi-level way:
Level 1 KPI - visibility
- number of sessions from AI tools
- aI unique users
- increase in the share of AI in traffic sources
The goal: to confirm that the brand is recognizable and appears in the models' responses.
Level 2 KPI - traffic quality
- average session time
- views per session
- returns on the site
The goal: to assess whether visitors from AI Search are interested in content.
Level 3 KPI - purchase intention
Before transactions occur, the intent should be apparent:
- entries on offer pages
- add to basket
- procurement processes initiated
- newsletter signups / contact forms
The goal: to assess AI users' willingness to buy.
Level 4 KPIs - sales and attribution
After configuring attribution according to models (e.g., data-driven), analyze:
- aI-assisted transactions
- conversion paths in which AI was one of the steps
Goal: Assess the real contribution of AI to revenue.
How to configure GA4 step by step
A. Segments and explorations
Create explorations like "Conversion paths" where you analyze:
- users starting with AI sessions
- time to conversion
- number of interactions along the path
B. Definitions of auxiliary conversions
Do not limit yourself to transactions, set as conversions, for example:
- enter product page
- CTA "contact"
- pDF download
- newsletter signup
This makes it possible to extract signals of intent earlier than at the arrival.
C. Multichannel attribution
Compare results:
- in the last click model
- in a data-driven or time decay model
This will show how often AI "assists" in shopping paths.
In the era of generative search engines, traffic is just the beginning. What matters is whether we can convert interest into intent and then into a transaction. Proper configuration of metrics and attribution in GA4 allows us to see what has been hidden until now.
Tomasz Cincio - CEO of Semly.ai
Example event in GA4
Below is an example of an event you can add to better track AI sessions:
<script async src="https://www.googletagmanager.com/gtag/js?id=G-XXXXXXX"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
// Standardowe GA4
gtag('config', 'G-XXXXXXX');
// Zdarzenie sesji z AI
gtag('event', 'ai_session_start', {
event_category: 'AI Search',
event_label: 'AI Session'
});
</script>Such an event allows for later segmentation and analysis.
FAQ
Is it possible to uniquely attribute sales to ChatGPT or another AI model?
Not in a fully deterministic way. AI Search very often acts as a channel to initiate or support a purchase decision. Therefore, attribution models, path analysis and assisted conversions, not just last click, are key.
Why does traffic from AI often have high quality but low direct conversion?
Because users use AI mainly at the research, comparison and education stage. This is the upper and middle stage of the funnel, not the moment of finalizing a transaction.
How to distinguish valuable traffic from AI from random clicks?
By analyzing: engagement time, number of page views per session, returning users, transitions to listing pages and search engine usage. Sessions alone are not a sufficient indicator.
Is GA4 enough to measure AI Search?
GA4 is a good base, but in practice it is worth supplementing it with: response audits of AI models, monitoring of brand visibility in responses, query testing and AI visibility tracking tools.
How long do you have to measure data to draw business conclusions?
The minimum reasonable period is 3-4 months. More stable conclusions about the impact on sales usually occur at a horizon of 4-6 months.
Does the lack of sales after 3 months mean that the channel is not working?
No. If the number of sessions is increasing, the quality of traffic is improving, and micro-conversions are occurring, the channel is in the process of building influence. The problem is the lack of these signals, rather than the lack of the transaction itself.
Which micro-conversions are best for traffic analysis with AI?
Most common: entries to product and offer pages, add to cart, start checkout, fill out form, sign up for newsletter, download materials.
Can AI Search cannibalize SEO or paid campaigns?
It can change their share of conversion paths, but most often works complementarily. In multichannel reports, you'll often see that AI initiates visits and SEO or direct closes them.
How do you report the impact of AI to management or the customer if there are no sales yet?
Through four blocks:
- visibility
- traffic quality
- intention
- impact on pathways
Such reporting shows the trend and business potential, not just the bottom line.
Is it worth creating separate dashboards for AI Search?
Yes. A separate dashboard allows you to track channel dynamics without the "noise" of other sources and makes it easier to communicate effects within the organization.
How do you know if offline or B2B sales are supported by AI?
Through: questions in forms ("how did you reach us?"), analysis of first visits to GA4, correlation of visibility in AI with number of inquiries.
Do advertising campaigns make sense with the growing role of AI Search?
Yes, but their role is changing. They often take over the function of closing the demand that was previously generated by SEO and AI Search.
Glossary
AI Search - An AI-based search model in which the user receives direct responses generated by language models (e.g., ChatGPT, Gemini, Perplexity) instead of a classic list of links.
GEO (Generative Engine Optimization) - A set of actions to increase brand, product and content visibility in responses generated by AI models.
AEO (Answer Engine Optimization) - Optimizing content for systems that answer user questions (search engines, voice assistants, chatbots) with the goal of providing clear, easily quotable answers.
AI visibility (AI visibility) -The level of brand, product, or content presence in AI models responses, regardless of whether the user clicks on the link.
AI citation / citation in AI - A situation in which an AI model invokes a brand, domain or piece of content as a source of information in its response.
Zero-click search - Queries in which the user receives an answer without having to go to the website. In AI Search, this is the dominant interaction model.
Micro-conversions - User actions that signal interest in an offer but are not yet a sale, such as going to a product page, signing up for a newsletter, downloading material, starting a checkout.
Assisted conversions (assisted conversions) - Transactions in which a particular channel (e.g., AI Search) appeared in the user's path, but was not the last click.
Multichannel attribution - A model for attributing sales value to various user-brand touch points (e.g., AI Search, SEO, advertising, direct), rather than attributing the total to a single source.
Data-driven attribution - Attribution model in GA4, which uses data and algorithms to determine the real contribution of each channel to conversion.
User intent (user intent) - Stage and purpose of the user's query, e.g., informational (research), comparative, transactional. AI Search is dominated by informational and advisory queries.
Share of voice in AI - Brand share of AI-generated responses compared to competitors for a specific set of queries.
Conversion path - The sequence of user interactions with the brand before the goal is executed (e.g., AI → SEO → direct → purchase).
Last click - Attribution model attributing the entire conversion value to the last input source, which in the case of AI Search very often underestimates its real impact.
Sources
AI Search Metrics That Actually Matter - analysis of key KPIs for visibility in generative results and their impact on conversions.
GEO Metrics: KPIs for Competitive Visibility - a description of generative visibility metrics that differ from classic SEO metrics.
Tracking AI Overview Success - analysis of metrics related to the presence of AI Overviews and their impact on subsequent user behavior.
Measuring Success in AI Search: Metrics That Matter - review of metrics such as share of voice, citation frequency and AI snapshot presence.
Answer Engine Optimization (AEO) - an encyclopedic discussion of the AEO concept, which is closely related to optimization under AI Search.
Google Analytics - general documentation of the analytical tool we use as a basis for measuring KPIs.
Semly.ai Blog - aI Search and visibility materials - other articles on visibility strategies in AI models.
Share:
