How to measure traffic from ChatGPT, Perplexity and AI Overviews

Measure ChatGPT and Perplexity traffic in your analytics by grouping visits whose referrer or UTM source is an AI tool, such as chatgpt.com or perplexity.ai. AI Overviews clicks are counted inside Google Search Console's normal web search totals and cannot be split out cleanly. Add regular prompt checks to capture mentions that never produce a click.

Why AI traffic is hard to see

Classic organic search is easy to measure. A visit arrives from Google, analytics labels it organic search, and Search Console tells you the query.

AI search breaks that in three ways:

1. Visits from AI assistants often arrive as referrals, not organic search, so they get mixed in with every other website that links to you.
2. Some visits arrive with no referrer at all and land in direct traffic. This happens with links opened from some desktop and mobile apps, or when someone copies a link from an answer and pastes it into a browser.
3. Many AI answers mention your brand without linking, or link without being clicked. That influence is real but leaves no trace in analytics.

So you need three layers: referral tracking for clicks, Search Console for Google's AI features, and manual or tool-based checks for mentions.

Layer 1. Track AI referrals in your analytics

Know the sources

These are the domains that commonly show up as referrers when someone clicks a link inside an AI tool. Referrer behaviour can change, so treat this as a starting list and add any new AI domains you see in your own reports.

Because OpenAI documents the UTM tag, ChatGPT is the easiest of these to track. In GA4 it often appears as source "chatgpt.com" with medium "referral" or "(not set)". Check both the session source and the UTM source dimensions so you do not miss it.

Build an AI traffic segment in GA4

1. In GA4, go to Admin, then Data display, then Channel groups.
2. Copy the default channel group so you can edit it.
3. Add a new channel called AI Search.
4. Set the rule so the session source matches a regular expression covering the AI domains, for example chatgpt|perplexity|copilot|gemini|claude.
5. Move the AI Search channel above Referral in the list. Channel rules are evaluated in order, and if Referral comes first it will capture these visits.
6. Save, then use the new channel group in your traffic acquisition reports.

The new group applies to data going forward. For history, use an exploration with a filter on session source matching the same pattern.

Report on outcomes, not just sessions

Sessions alone undersell AI traffic, because visitors from AI tools have often already read a summary and arrive closer to a decision. For each AI source, track:

• Sessions and engaged sessions
• Key events such as form submits, demo requests and audit bookings
• Conversion rate compared to organic search and paid
• Landing pages, to see which content AI tools send people to

Landing pages are the most useful part. They show which of your pages the AI tools trust enough to link.

Layer 2. Read AI Overviews through Search Console

Google's AI Overviews and AI Mode live inside Google Search. Clicks from them arrive as normal Google organic traffic in analytics, with no separate referrer.

Google's documentation on AI features says that sites appearing in them are included in the overall search traffic in Search Console's Performance report under the Web search type. A click on a link in an AI Overview counts as a click, and appearing in one counts as an impression. In other words, the data is there but blended with regular results.

What you can still do:

• Track query groups. Pick the questions where you know an AI Overview appears. Watch impressions, clicks and click-through rate for those queries over time.
• Watch for impression growth without click growth. Rising impressions with flat clicks on question queries is a common sign that an AI Overview is answering above you.
• Check the SERPs by hand. Search your top questions and record whether an AI Overview appears and whether you are cited in it.
• Check your Search Console account for newer reports. Google has been adding reporting for AI features, and what is available depends on your account. Use it if you have it, and keep the query-group method as a backstop.

Layer 3. Measure mentions that produce no click

A large share of AI influence is invisible in analytics. A buyer asks ChatGPT for agencies that do automation, sees your name, and later types your URL directly or searches your brand on Google. Analytics credits that visit to direct or branded search.

To capture this:

Run a prompt panel

1. Write 20 to 40 prompts that match how your buyers ask questions. Mix category questions ("best agencies for marketing automation"), problem questions ("how to automate invoice approvals") and comparison questions.
2. Run each prompt in ChatGPT, Perplexity, Gemini, Claude and Google with AI Overviews, at a set cadence such as monthly.
3. Record whether your brand is mentioned, whether you are linked, your position in any list, and which sources were cited.
4. Track the share of prompts where you appear over time.

Answers vary from run to run, so look at trends across many prompts rather than reacting to a single answer. Several commercial tools now automate this kind of tracking if you want to scale it.

Watch branded search and direct traffic

If AI tools start recommending you, you would expect more people searching your brand name. In Search Console, filter queries containing your brand and watch the trend. In analytics, watch direct traffic to your homepage and key service pages.

These are indirect signals with other possible causes, so read them alongside your prompt panel rather than on their own.

Ask new leads

Add an optional "How did you hear about us" field to your contact or booking form, with ChatGPT, Perplexity and other AI tools as explicit choices next to Google search. Self-reported attribution is imperfect, but it is often the only way to see AI influence that leaves no trace. Our guide to lead form best practices covers how to add this without hurting conversion.

Putting it into one report

A monthly AI visibility report can be one page:

If you already send weekly reports to leadership, add this as a section rather than a separate document. Our post on automated reporting shows how to pull these numbers together without doing it by hand.

Common mistakes

• Leaving AI visits inside Referral. Unless you create a channel, AI traffic is mixed with partner sites and directories.
• Ordering the channel rule below Referral. GA4 evaluates channel rules in order, so the AI rule has to come first.
• Judging AI visibility on sessions alone. Low volume can still mean high intent. Look at conversion rate.
• Reacting to one prompt result. AI answers change between runs. Use many prompts over time.
• Ignoring direct traffic. Some AI clicks arrive without a referrer, so an unexplained rise in direct traffic to deep pages is worth investigating.

How we help

Many companies already get some AI traffic and have no way to see it. Setting up the channel, the prompt panel and the reporting takes a few hours and turns AI search from a guess into a number. If you want help putting it in place, book a business audit and we will include AI search measurement in your plan.