How to Track Website Traffic from ChatGPT and Other AI Platforms

A few years ago, “where does my traffic come from?” had a short list of answers: Google, social media, email, direct visits and links on other websites. Today there is a new entry that many analytics reports still hide: people who ask an AI assistant a question, see your brand in the answer, and click through to your website.

If you have noticed a vague bump in direct traffic, or a strange new source like chatgpt.com in your referral report, you have already seen it. The challenge is that AI traffic is messy to measure. Some clicks carry clear labels, some carry none, and the rules have changed several times in the past year.

This guide explains what you can track, what you cannot, and how to set it up properly, so you can answer the question your boss or client will soon ask: “Is AI sending us customers?”

Why AI Traffic Deserves Its Own Line in Your Reports

AI assistants behave differently from search engines. Someone who clicks a link inside a chatbot answer has usually already read a summary, compared options and narrowed down their choice. By the time they reach your site, they are often further along the buying journey than a casual search visitor.

There are three practical reasons to measure it:

  1. It shows whether your content is being cited. If AI platforms send visitors to a page, that page is being treated as a trusted source.
  2. It helps you judge quality. AI visitors may convert, bounce or browse differently from organic visitors, and you can only learn this if they are separated in your data.
  3. It protects your reporting. Without tracking, this traffic is quietly mixed into Direct or generic Referral, and your real channel performance gets distorted.

The Hard Truth: You Will Never See Everything

Before touching any settings, set expectations. Analytics tools can only report what the visitor’s browser tells them. Many AI visits arrive with incomplete information:

  • Missing referrer. Some AI apps, especially mobile apps and in-app browsers, open links without telling your site where the visitor came from. These visits land in Direct.
  • Copy and paste. A person may read your brand name in an answer and type your address into a new tab. That visit looks like ordinary direct traffic.
  • Brand searches. Another person may see your name in an AI answer, then Google it later. Your report credits Google, not the assistant.
  • Changing behaviour. Platforms adjust how they tag links. In early October 2026, for example, ChatGPT links briefly stopped carrying their usual tag and referrer for roughly a week, so many sites saw their ChatGPT numbers drop and their Direct numbers rise.

The lesson is simple: treat your measured AI traffic as a minimum, not a total. Industry estimates of how much AI traffic goes unlabelled vary widely, so rely on your own trends rather than anyone’s headline percentage.

Step 1: Check What Google Analytics 4 Already Shows

Before building anything, look at what you already have.

  1. Open Google Analytics 4 and go to Reports → Acquisition → Traffic acquisition.
  2. Change the primary dimension to Session source / medium.
  3. Search for terms such as chatgpt, openai, perplexity, claude, gemini and copilot.

GA4 introduced a built-in AI Assistant channel in May 2026, which automatically groups traffic from several major assistants. Check whether it appears in your default channel report. Two cautions apply:

  • It only works going forward from its launch date; it does not rewrite your history.
  • It does not necessarily cover every platform. Some sources, such as Perplexity, may still show up under generic Referral on certain properties.

If you can see all the platforms you care about, you already have a good start. For most sites, it is worth going one step further.

Step 2: Build a Custom AI Channel Group

A custom channel group lets you define exactly which sources count as “AI traffic” and see them as one row beside Organic Search, Direct and Social. A custom group can also be applied to past data, which the native channel cannot.

Here is how to set it up:

  1. In GA4, click Admin (bottom left).
  2. Under Data display, choose Channel groups.
  3. Click Create new channel group and name it something clear, such as Default + AI Traffic.
  4. Click Add new channel and name it AI Referrals.
  5. Set the condition to Session source → matches regex, then paste a pattern like this:

chatgpt\.com|chat\.openai\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|copilot\.microsoft\.com|grok\.com|deepseek\.com

  1. Drag the new channel above Referral. This step is the one most people miss. GA4 reads the rules from top to bottom and assigns each session to the first channel that matches, so if Referral sits higher, your AI rule never fires.
  2. Save the group.

Two practical notes. First, standard GA4 properties limit how many custom channel groups you can create, so plan your structure before using up a slot. Second, review your pattern every few months, because new AI tools and domains appear regularly.

You cannot control how an AI platform labels its links. You can control the links you place in your own material that AI tools may pick up, such as:

  • Your website’s product and resource pages shared in newsletters, PDFs and social posts
  • Links inside documents or knowledge bases that feed AI-powered tools
  • Links you submit in profiles, directories and listings

For these, use a consistent tagging scheme. For example:

  • utm_source=chatgpt (or perplexity, gemini, claude)
  • utm_medium=ai_referral
  • utm_campaign= a short label for the purpose

One point from experience: always include utm_medium along with utm_source. When a visit has a source but no medium, and no referrer, GA4 can drop it into an Unassigned bucket, which hides it from your channel reports.

Step 4: Create a Dedicated Report

A channel group is useful, but a focused report makes the data easy to share.

  1. In GA4, open Explore and create a Free form exploration.
  2. Add Session source / medium and Landing page as rows.
  3. Add metrics such as Sessions, Engaged sessions, Engagement rate, Conversions and Revenue.
  4. Apply a filter where Session source matches the same AI regex used above.

This view answers three important questions: which AI platforms send visitors, which pages they land on, and whether those visitors take action. The landing page list is especially valuable, because it shows exactly which of your pages AI tools consider worth citing.

Step 5: Look at Your Server Logs

Google Analytics only sees visitors who load your page and run its tracking code. AI systems also visit your site in other ways: crawling pages to learn from them, or fetching a page live when a user asks a question. These bots appear in your server or CDN logs, not in GA4.

In your hosting dashboard, Cloudflare analytics or raw log files, look for user agents such as:

  • GPTBot, OAI-SearchBot and ChatGPT-User (OpenAI)
  • ClaudeBot and Claude-User (Anthropic)
  • PerplexityBot (Perplexity)
  • Google-Extended (Google’s AI training control)

This data does not tell you about human visitors, but it shows how often AI systems read your content and which pages they are most interested in. Pages that bots fetch frequently are strong candidates for being cited in answers. Check each company’s official documentation for the current list of user agents, since names change.

Step 6: Capture the Source Before the Visitor Is Lost

Analytics tells you who visited. For a business that cares about leads and sales, the next question is who bought. To connect AI traffic to revenue:

  • Add a hidden field to your enquiry and contact forms that captures the session source and UTM values.
  • Pass that field into your CRM, so the original source stays attached to the lead through to closed deal.
  • Add a simple question on forms: “How did you hear about us?” with options such as Google, social media, a friend, and AI assistant (ChatGPT, Gemini, Claude, etc.).

The self-reported answer is surprisingly powerful. It catches exactly the visitors that technical tracking misses, such as the person who saw your name in a chatbot and typed your address in manually.

Step 7: Estimate the Invisible Traffic

Since some AI traffic is always hidden in Direct, you can build a rough picture by watching patterns instead of single numbers:

  • Spikes in Direct traffic to deep pages (a specific blog post or product page, rather than your homepage) are a classic sign of hidden AI visits, since few people type long URLs by hand.
  • Growth in branded search after your content starts appearing in AI answers suggests assistants are building awareness.
  • Before-and-after comparisons are useful when you publish a major guide or update a key page.

Treat these as signals to investigate, not proof.

Common Mistakes to Avoid

Building the channel but forgetting the order. If AI Referrals sits below Referral, it will show zero sessions.

Trusting one number. Platform behaviour changes, and a sudden drop can be a tracking change on the platform’s side, not a real loss of visitors. Check Direct traffic before reacting.

Ignoring quality. Ten AI visitors who enquire may be worth more than a hundred who bounce. Look at engagement and conversions, not just sessions.

Using only a regex. Combine technical tracking with form questions and CRM fields for a fuller picture.

Not documenting your setup. Write down your regex, report filters and date of changes so your team and future analysts can interpret trends correctly.

A Simple Monthly Checklist

  1. Review sessions, engagement and conversions from your AI channel.
  2. Check which landing pages receive AI visitors.
  3. Compare with Direct traffic to deep pages.
  4. Skim server logs for AI crawler activity.
  5. Read the “How did you hear about us?” answers and CRM leads.
  6. Update your regex if new platforms appear.

AI assistants are becoming a real discovery channel, and measuring them well is mostly a matter of setup and honesty about limits. Start with what GA4 already offers, build a custom channel group with the right order, tag the links you control, and back it up with server logs and a simple “how did you hear about us?” question.

You will never capture every visit, but you do not need to. What matters is spotting the trend early: which AI platforms notice your content, which pages they favour, and whether those visitors turn into customers. The brands that start measuring now will understand this channel long before their competitors do.