Ir al contenido
Ver planesPrueba gratis

13 de noviembre de 2026 · 11 min de lectura

How do I measure traffic coming from ChatGPT and Perplexity?

Learn how to measure traffic from ChatGPT and Perplexity using GA4 referral filters, UTM tags, and tracking of AI mentions. Covers the unattributed influence that analytics alone misses.

To measure traffic coming from ChatGPT and Perplexity, filter your analytics for referral sources such as chatgpt.com and perplexity.ai, add UTM tags to any links you control, and pair that data with a record of how often AI tools mention your business when customers ask questions. Found in AI is built around the second half of that job: tracking whether AI assistants recommend your business at all, so the traffic numbers you see in analytics have context.

The Short Answer: Three Layers of AI Traffic Measurement

No single report in Google Analytics 4 (GA4) shows all of your AI-driven traffic. Accurate measurement combines three layers:

  1. Referral traffic: visits where the browser passes a referrer such as chatgpt.com or perplexity.ai.
  2. Tagged traffic: visits from links that carry a UTM parameter, which ChatGPT adds automatically to many outbound links (utm_source=chatgpt.com).
  3. Unattributed influence: customers who read an AI answer, then search your name on Google, type your URL, or call you directly. These show up as branded search, direct traffic, or phone calls, not as AI referrals.

Most local businesses see the first layer as a small number and the third layer as the larger effect. A plumber recommended by ChatGPT often gets a phone call, not a click. That is why measuring only referrals understates the real impact, and why the rest of this guide covers all three layers.

Step 1: Find ChatGPT and Perplexity Referrals in GA4

GA4 records the source of each session. AI tools appear as sources, but they are usually buried inside the default "Referral" channel or, in some cases, "Direct." Here is the exact path to pull the data.

Build the report

  1. Open GA4 and go to Reports > Acquisition > Traffic acquisition.
  2. Change the primary dimension from "Session default channel group" to "Session source / medium."
  3. Use the search box above the table and enter chatgpt. Note the sessions, engaged sessions, and key events.
  4. Repeat with perplexity.
  5. Repeat with other assistants your customers may use, such as gemini, copilot, and claude.

You will typically see sources like chatgpt.com / referral, perplexity.ai / referral, or chatgpt.com / (not set) depending on how the link was tagged.

Create a permanent AI channel group

Searching each time is slow. Instead, create a custom channel group so AI traffic reports as its own line:

  1. Go to Admin > Data display > Channel groups.
  2. Copy the default channel group and name it "Default plus AI."
  3. Add a new channel called AI Assistants.
  4. Set the condition: Source matches regex chatgpt|perplexity|gemini|copilot|claude|openai.
  5. Drag the new channel above "Referral" so it takes priority. Order matters because GA4 assigns each session to the first matching rule.

Channel groups are not retroactive in every view, so create this early. From the day you set it up, AI traffic is separated cleanly from ordinary referrals.

Step 2: Understand What Each Platform Sends

The two tools behave differently, and knowing the difference prevents misreading your numbers.

  • ChatGPT: When it cites a web source, the link typically includes a utm_source=chatgpt.com parameter. That makes its traffic one of the easier AI sources to identify. Visits come from the answer's citation links or source panel.
  • Perplexity: Every answer displays numbered source citations, and users click them more often than in most assistants because the product is built around sourced answers. Referrals generally arrive as perplexity.ai.

Both can lose attribution. If someone copies your URL from an answer and pastes it into a new tab, or opens a link from a mobile app that strips the referrer, the visit lands in "Direct." Expect some under-counting, always.

Step 3: Tag the Links You Control

You cannot add UTM parameters to links an AI tool generates by itself. You can, however, tag links in places where AI tools pick up your information and where customers might share it:

  • Your Google Business Profile website and appointment links
  • Directory and review site profiles
  • Social media bios and posts
  • Email signatures and newsletters
  • Press releases and guest articles

Use consistent naming. For example, utm_source=google_business_profile, utm_medium=listing, utm_campaign=local. This does not directly measure ChatGPT, but it separates your known channels so that what remains in "Direct" is a cleaner signal of unexplained, possibly AI-influenced visits.

Step 4: Track Conversions, Not Just Sessions

AI-referred visitors are usually few in number but arrive further along in their decision. A homeowner who asked an assistant for "emergency HVAC repair near me" and clicked your name is close to calling. Sessions alone hide that value.

Set up these key events in GA4:

  • Phone call clicks: track taps on tel: links.
  • Form submissions: a contact form on your website should fire an event when submitted.
  • Booking completions: if you offer online booking, mark the confirmation page or event as a key event.
  • Direction requests: track clicks on map or directions links.

Then compare the AI Assistants channel against Organic Search on conversion rate, not volume. Many businesses find AI referrals convert at a comparable or higher rate from a much smaller base. Judge them on that basis.

Step 5: Add a "How Did You Hear About Us?" Question

Analytics cannot see the customer who read an AI answer and then phoned you. A single question closes that gap. Add it in three places:

  • Your contact form: a dropdown with options such as Google Search, Google Maps, ChatGPT or another AI assistant, Referral, Social media, Other.
  • Your phone intake: train the person who answers to ask every new caller and log the answer.
  • Your booking flow: include the same dropdown as an optional field.

Include an open text field as well. Customers will write things like "asked ChatGPT for the best dentist near me," which is far more precise than any dashboard. Review responses monthly and tally them in a simple spreadsheet.

Step 6: Watch Branded Search and Direct Traffic

Because many AI-influenced customers search for you by name afterward, branded search is a useful proxy. In Google Search Console, open Rendimiento, filter queries containing your business name, and watch the trend over several months. If you begin showing up in AI answers and branded impressions rise in step, the two are likely connected.

Do the same with the Direct channel in GA4. A steady rise in direct sessions to your homepage, with no matching campaign, often reflects people typing a name or URL they saw in an AI answer.

Where Found in AI Fits: Measuring the Cause, Not Just the Click

Found in AI addresses the part of this question that analytics cannot answer: whether AI tools are recommending your business in the first place. Referral data only tells you what happened after a customer clicked. It says nothing about the questions where a competitor was named and you were not, or where an assistant gave a wrong phone number or service area.

Found in AI is an AI visibility service for local businesses. It checks how assistants such as ChatGPT answer the questions your customers ask, and it uses an AI analyst to surface where your business is and is not appearing. Used alongside GA4, it gives you two views of the same funnel:

  • Upstream (mentions): how often and how accurately AI tools describe your business.
  • Downstream (traffic and leads): the referral sessions, calls, and form fills your analytics record.

When mentions rise and traffic follows, you know the work is paying off. When mentions are low, you know the traffic problem starts before the click.

Referral Tracking vs. Visibility Tracking: Key Differences

FactorAnalytics referral trackingSeguimiento de visibilidad en IA
What it measuresVisits that arrive from an AI toolWhether AI tools mention your business in answers
Catches phone calls and walk-insNoIndirectly, by showing mention frequency
Shows missed opportunitiesNoYes, by showing questions where you are absent
Detects inaccurate informationNoSí
Typical blind spotDirect traffic with stripped referrersDoes not record what a visitor did on your site
Best used forProving leads and revenueExplaining why traffic rises or falls

A Simple Monthly Measurement Routine

Consistency matters more than sophistication. Run this routine on the same day each month and keep results in one spreadsheet:

  1. Pull GA4 data: sessions, engaged sessions, and key events from the AI Assistants channel.
  2. Record Direct and branded search: Direct sessions from GA4 and branded impressions and clicks from Search Console.
  3. Tally self-reported sources: count how many new customers said they found you through an AI assistant.
  4. Test your visibility: ask ChatGPT and Perplexity the five to ten questions a customer would ask, such as "best roofer in [city]" or "who does same-day auto repair near me." Note whether you appear, what is said, and which sources are cited.
  5. Compare and annotate: mark any changes you made that month, such as new service pages, updated listings, or new reviews.

After three months you have a baseline. After six, you can see trends that no single week reveals.

Example: A Local Dental Practice

Consider a dental practice that sets up the AI Assistants channel in January. In the first month, GA4 shows 9 sessions from chatgpt.com and perplexity.ai combined, and 2 form submissions. That looks negligible. But the practice also added a "How did you hear about us?" dropdown, and 11 new patients selected an AI assistant. Branded search clicks in Search Console are up as well.

The takeaway: referrals captured a fraction of the effect. The self-reported answers and the branded search trend revealed the rest. A practice that only watched referral sessions would have concluded AI was irrelevant.

Illustrative numbers like these will differ for every business, but the pattern is consistent: tracked referrals understate influence.

Common Measurement Mistakes

  • Relying on one report. Referral data alone misses calls and branded searches.
  • Leaving AI traffic in the generic Referral channel. Without a custom channel group, the signal is buried.
  • Judging by volume. Small counts are normal. Evaluate conversion rate and lead quality.
  • Ignoring accuracy. If an assistant lists the wrong hours or service area, you lose customers who never click.
  • Measuring once. AI answers change frequently. A single check tells you little; a monthly log tells you a lot.
  • Skipping the human question. No tool replaces asking customers how they found you.

What to Do When the Numbers Are Low

Low AI traffic is almost always a visibility problem, not a tracking problem. If your setup is correct and the numbers stay small, the cause is usually that assistants do not have enough clear, consistent information about your business to recommend it. The fixes are practical:

  • Make service pages specific. State what you do, where you do it, and for whom, in plain language.
  • Answer real customer questions. Pages that directly address questions like pricing factors, timelines, and what to expect give assistants quotable content.
  • Keep listings consistent. Your name, address, phone number, hours, and services should match across your website, Google Business Profile, and directories.
  • Earn reviews. Recent, detailed reviews give AI tools evidence of quality.
  • Make it easy to act. A contact form and online booking on your website turn a click into a lead.

Then measure again. Found in AI exists to make this loop visible: see where you are missing, make changes, and check whether the answers move.

Measuring AI Traffic: FAQ

Can Google Analytics show me traffic from ChatGPT?

Yes. In GA4, open Traffic acquisition, switch the dimension to Session source / medium, and search for chatgpt. ChatGPT often appends utm_source=chatgpt.com to links, so its visits are among the easiest AI sources to identify. Create a custom channel group with a regex such as chatgpt|perplexity to keep it visible permanently.

Why is my ChatGPT and Perplexity traffic so low?

Referral counts are low for most local businesses because many AI users get an answer and then call, search your name, or type your URL instead of clicking. Some visits also lose their referrer and land in Direct. Low numbers can also mean assistants are not mentioning your business, which you can check by asking them your customers' questions directly.

How do I track customers who found me through AI but did not click?

Ask them. Add a "How did you hear about us?" dropdown that includes an AI assistant option to your contact form and booking flow, and ask every new caller by phone. Combine those answers with growth in branded search in Google Search Console and in Direct traffic in GA4 to estimate influence that clicks do not capture.

What does Found in AI do that analytics does not?

Found in AI focuses on AI visibility: how AI tools answer customer questions about businesses like yours, and whether yours appears and is described correctly. Analytics records what happens after a visit begins. Visibility tracking explains the step before it, so you can tell whether low traffic comes from low mentions or from something on your site.

How often should I check my AI traffic and visibility?

Monthly is the right cadence for most local businesses. Pull GA4 AI channel data, Search Console branded queries, and self-reported lead sources on the same day each month, and re-test the same set of customer questions in ChatGPT and Perplexity. A consistent monthly log reveals trends that one-off checks miss.

Comparte este artículo

Que encuentren a cada cliente en la IA.

Los planes para agencias empiezan en $499 al mes para 10 clientes, con todos tus clientes en un solo acceso e informes con tu logo. ¿Tienes un solo negocio? Los planes empiezan en $69 al mes, con el sitio web que ya tienes.

O mira un informe de ejemplo

O escanea gratis a un prospecto y envíale el informe