---
title: "How to track client referral traffic coming from ChatGPT and Perplexity in GA4"
description: "Track ChatGPT and Perplexity traffic in GA4 by building custom channel groups that capture referrer domains, API source parameters, and automated UTM tags."
url: https://vizzybl.ai/blog/how-to-track-client-referral-traffic-coming-from-chatgpt-and-perplexity-in-ga4
category: "GEO Know-How"
author: "Jez Lloyd"
date_published: 2026-10-26T22:48:04.523Z
date_modified: 2026-10-26T22:48:04.523Z
reading_time_minutes: 8
source: "Vizzybl"
---
# How to track client referral traffic coming from ChatGPT and Perplexity in GA4
## Key takeaways
- GA4 defaults ChatGPT and Perplexity traffic to Direct or Referral channels because tagging parameters shift and answer engines often lack consistent UTM parameters.
- Perplexity never attaches UTM parameters to cited links, whereas ChatGPT alternates between automated UTM tags and standard HTTP referrer headers depending on the interface.
- Custom channel groups using regular expressions capture visits from web interfaces, mobile app deep links, and API sources that GA4 would otherwise fragment or misclassify.
- GA4 cannot identify the specific prompts that drove AI referrals because answer engines omit query strings and prompt text from destination links to preserve user privacy.
- Capturing AI referral metadata in hidden form fields allows marketing teams to pass source attribution from GA4 directly into CRM lead records.
*Last updated: 2026-10-26*
"We see AI referral traffic in Google Analytics, but we cannot tell which prompts or pages are driving those visitors." Marketing teams, agencies, and SEO professionals routinely run into this exact wall when inspecting GA4 referral sources. Tracking referral traffic coming from ChatGPT and Perplexity in Google Analytics 4 requires a combination of custom channel definitions, careful source parameter mapping, and custom landing page tracking.
GA4 does not separate generative answer engines into a dedicated traffic channel out of the box. Instead, it lumps these visits into standard referral domains or misclassifies them entirely as direct traffic whenever browser HTTP referrer headers are dropped. Isolating this traffic demands a clear understanding of how each platform structures outbound URLs and how GA4 processes session parameters.
## Why does GA4 classify ChatGPT and Perplexity traffic as Direct or Referral?
GA4 classifies incoming sessions based on the presence of HTTP referrer headers and inbound UTM parameters. If a visitor clicks a link inside an answer engine and the browser passes a recognizable HTTP referrer domain, GA4 assigns the session source to that referral domain. If no referrer header is sent and no UTM parameters exist on the destination link, GA4 defaults the session to Direct.
In our tests analyzing outbound citation links, we have seen that Perplexity and ChatGPT handle referral metadata in completely different ways. Of 15,306 links Perplexity cited in answers we tracked (14,775 through its API between August 20 and October 22, 2026, and 531 in its web app between August 27 and September 18), not one link carried a UTM parameter. Perplexity relies entirely on the HTTP referrer header passed by the browser. When a user's browser, extension, or network security protocol strips that referrer header, GA4 has no source data and classifies the visit as Direct.
ChatGPT uses a mix of referral headers and automated UTM tags. When answers come through OpenAI's API, every cited link is tagged with `utm_source=openai`. In our tests across 2,139 API-generated links between August 20 and October 22, 2026, 100% carried this tag. ChatGPT's web application behaves differently, applying `utm_source=chatgpt.com`.
However, web app tagging can be inconsistent. From August 27 to October 8, 2026, ChatGPT's web app appended `utm_source=chatgpt.com` to 2,706 out of 2,711 cited links. Between October 9 and October 15, that tag dropped off almost completely, appearing on only 6 of 680 cited links, before returning on October 16 and appearing on every link by October 17. When automated UTM tags disappear, GA4 falls back on the raw HTTP referrer header, such as `chatgpt.com` or `android-app://com.openai.chatgpt`. If that header is absent, the session converts to Direct.
| AI Platform | Interface | Primary Tagging Behavior | GA4 Default Source |
| --- | --- | --- | --- |
| ChatGPT | Web Application | `utm_source=chatgpt.com` (subject to API/web shifts) | `chatgpt.com` or `Direct` |
| ChatGPT | API / Enterprise Integrations | `utm_source=openai` | `openai` |
| Perplexity | Web & API | No UTM tags applied | `perplexity.ai` or `Direct` |
## How to set up a custom channel group for AI referrals in GA4?
To isolate AI answer engine visits in GA4, you must build a custom channel group that evaluates both the `Source` parameter and `First user source` dimension against standard regex expressions. This ensures visits from chat interfaces, mobile applications, and platform APIs consolidate into a single dedicated channel rather than scattering across generic Referral or Direct categories.
Setting up a custom channel group involves navigating to Admin Settings in GA4, selecting Data Display, and editing Channel Groups. You create a new rule named "AI Referrals" or similar, using OR logic to catch both standard source domains and platform-specific UTM tags.
```text
Channel Name: AI Referrals
Rule Condition:
Source matches regex: (.*chatgpt.*|.*openai.*|.*perplexity.*|.*claude.*|.*gemini.*)
OR
Source platform matches regex: (.*chatgpt.*|.*perplexity.*)
```
Building this rule requires combining all domain variants. When we checked custom channel rules against the links answer engines actually output, we found a critical coverage gap. A standard rule that only looks for `chatgpt.com` or `openai.com` misses every API-driven interaction tagged strictly as `openai`. Conversely, a rule looking only for `openai` misses web app interactions using `chatgpt.com`.
A comprehensive custom channel group rule must account for:
1. Web referral domains (`chatgpt.com`, `perplexity.ai`, `news.perplexity.ai`).
2. Mobile app deep links (`android-app://com.openai.chatgpt`, `android-app://ai.perplexity.app`).
3. Appended UTM source tags (`openai`, `chatgpt.com`).
Failing to include API source strings means losing visibility into third-party integrations, enterprise tools, and developer platforms powered by LLMs. Configuring this channel costs zero additional budget in GA4, but it fixes reporting fragmentation and prevents AI-driven traffic from polluting your general referral reporting.
## Can GA4 report which prompts generated referral traffic from ChatGPT and Perplexity?
GA4 cannot report the specific user prompts or queries that generated referral traffic from ChatGPT or Perplexity. Answer engines intentionally omit prompt text, conversation identifiers, and query string parameters from outbound links to maintain user privacy and prevent data leakage.
When a user asks an AI engine a question and clicks a cited link, the web browser receives a clean target URL. Neither ChatGPT nor Perplexity appends the original user prompt to the query string of the destination URL. Because GA4 only records details present in the landing page URL, the referrer header, or custom event parameters passed during page load, the exact prompt remains invisible within standard web analytics.
While GA4 cannot display the query text, you can determine what content answers specific prompts by mapping traffic down to the specific landing page level in GA4. If an AI engine repeatedly cites a specific product guide or technical specification page, you can infer the topical intent of the underlying prompts.
For organizations that need deeper query-level monitoring, dedicated platforms fill this visibility gap. An [AI visibility platform](https://vizzybl.ai) like [Vizzybl](https://vizzybl.ai) tracks how answer engines cite websites across target prompts, bridging the gap between what users ask inside LLMs and the landing page hits recorded in GA4.
Where standard GA4 tracking is not the right fit is full-funnel keyword attribution for generative engines. Standard web analytics tools simply lack the architecture to reverse-engineer private chat prompts from external HTTP requests.
## How to track conversion rates for AI referral traffic in GA4?
Tracking conversion rates for AI referral traffic in GA4 requires pairing your custom channel group with standard key event (formerly conversion) reporting across landing pages. Once your custom channel group is active, you can isolate AI session conversion rates against paid search, organic search, and social channels.
To view conversion metrics, open the **Traffic acquisition** report in GA4, set the primary dimension to your new custom channel group, and select your primary key event (such as lead form submission, demo request, or completed transaction) from the metric dropdown.
To analyze engagement quality alongside conversions, compare the following metrics across channels:
- **Engagement Rate:** High engagement rates indicate that the AI engine cited relevant content matching user intent.
- **Average Engagement Time:** Shows whether visitors coming from generative citations consume long-form content or exit immediately.
- **Key Event Conversion Rate:** Demonstrates commercial intent from users who clicked through AI citations compared to classic organic search traffic.
In our platform tracking, visitors arriving from generative answer engines often demonstrate higher engagement rates and intent than traditional search engine visitors. Because answer engines pre-summarize context and only present links as factual citations, users who click through are actively seeking deep documentation, pricing, or product verification. Measuring this performance provides marketing teams with clear data to justify investing in Generative Engine Optimization (GEO).
## How to link GA4 AI referral traffic data to CRM lead attribution?
Linking GA4 AI referral traffic to your CRM requires capturing session source metadata on your website forms and passing those parameters directly into custom CRM fields upon submission. Because GA4 operates as an aggregated analytics store, it does not pass individual user identifiers or full session histories directly to your CRM without custom implementation.
To capture this data effectively, set up a script on your website that reads the active session's referrer and `utm_source` parameters. When a visitor lands on your site from `chatgpt.com`, `openai`, or `perplexity.ai`, store these source values in the browser's local storage or session cookies.
When the visitor fills out a contact or lead form, populate hidden fields inside the form with those stored values:
```html
```
Mapping these hidden fields directly into standard CRM contacts (in systems like HubSpot or Salesforce) allows [marketing teams](https://vizzybl.ai/solutions/brands) and [agencies](https://vizzybl.ai/solutions/agencies) to attribute offline pipeline and won deals directly back to AI referral channels. Additionally, passing the GA4 Client ID into your CRM lets you leverage GA4 Measurement Protocol events to send offline conversion values back into Google Analytics, completing the attribution loop.
## Where Vizzybl AI is not the right fit
Vizzybl AI is strictly an AI visibility and GEO platform designed to monitor brand citations, prompt impressions, and engine recommendations across generative platforms. It is not a traditional SEO platform.
If your organization requires traditional keyword rank tracking for standard Google SERPs, backlink profile analysis, on-page site crawling, or technical desktop crawl auditing, traditional SEO platforms are the correct tool. Vizzybl AI focuses specifically on how generative answer engines synthesize brand information, render citations, and direct traffic across AI platforms.
## The bottom line
Tracking ChatGPT and Perplexity referral traffic in GA4 requires adapting to inconsistent source parameters, missing UTM tags, and automated platform shifts. By setting up custom channel groups that account for both web app referrers and API-specific source tags, you can isolate AI session activity, measure real conversion impact, and pass clean attribution data into your sales pipeline. To discover how your brand appears across generative search engines before visitors ever reach your site, join[ Vizzybl](https://vizzybl.ai).
## Frequently asked questions
### Can GA4 show the exact prompts users typed in ChatGPT or Perplexity?
GA4 cannot report the specific user prompts or queries that generated referral traffic from ChatGPT or Perplexity. Both answer engines intentionally omit user prompt text, conversation identifiers, and query string parameters from outbound citation URLs to protect user privacy. Analytics teams can only infer underlying topical intent by evaluating the specific landing pages that receive AI referral visits in GA4.
### What regex should you use for an AI referral channel group in GA4?
A custom channel group rule in GA4 should use regex matching patterns like (.*chatgpt.*|.*openai.*|.*perplexity.*|.*claude.*|.*gemini.*) for the Source parameter, or (.*chatgpt.*|.*perplexity.*) for the Source platform parameter. This regex catches web referral domains, mobile deep links such as android-app://com.openai.chatgpt, and API-generated UTM tags like utm_source=openai.
### What UTM tags does ChatGPT add to cited outbound links?
ChatGPT applies utm_source=openai to all outbound links generated through OpenAI's API. For outbound citations in the ChatGPT web application, ChatGPT applies utm_source=chatgpt.com, although this web application tagging can shift or drop intermittently. If these automated UTM tags are absent, GA4 relies on standard HTTP referrer headers like chatgpt.com to identify the session.
### How do you pass GA4 AI referral traffic data into a CRM?
Passing AI referral data to a CRM requires a website script that captures the active session referrer and utm_source values, saving them into browser cookies or local storage. When a visitor submits a lead form, the site populates hidden form fields with those stored AI referral values, sending first-touch source data and the GA4 Client ID directly into CRM contact records.
### Why did my ChatGPT referral traffic drop suddenly in GA4?
ChatGPT referral traffic drops in GA4 are often caused by changes in how OpenAI tags outbound links. Between October 9 and October 15, 2026, ChatGPT's web app stopped appending utm_source=chatgpt.com across nearly all cited links. If your tracking relies strictly on custom UTM tags rather than referrer domain regex rules, sessions temporarly shift into the Direct traffic channel.
### Does Perplexity add UTM parameters to cited website links?
No, Perplexity does not add UTM parameters to cited website links. Across thousands of tracked citations across both its API and web application interfaces, Perplexity's links carry no appended query tags. GA4 relies entirely on browser HTTP referrer headers to identify Perplexity traffic.
### How do I separate Claude and Gemini traffic from ChatGPT and Perplexity in GA4?
To separate Claude and Gemini traffic from ChatGPT and Perplexity, update your custom channel group in GA4 with distinct sub-rules. Use regular expressions matching anthropic, claude, gemini, and google source parameters to split these answer engines into dedicated channels alongside ChatGPT and Perplexity.
