How to Track Traffic From ChatGPT and Perplexity in GA4

You can track identifiable traffic from ChatGPT, Perplexity and other AI assistants in GA4 by reviewing session source and medium in the Traffic acquisition report, then grouping recognised referral domains into a custom AI-assistant channel. The result will be a useful minimum, not a complete count. Some AI clicks arrive without a referrer and appear as Direct.

GA4 measures visits that reach the website. It does not show every time an AI system cites the brand, mentions an article without a click or answers a question using information from another source.

That distinction matters. My guide to earning citations in ChatGPT and Google AI Overviews covers visibility. This guide covers the referral traffic that can be observed after somebody clicks.

Find AI referrals in Traffic acquisition

Open GA4 and go to Reports, Acquisition, then Traffic acquisition. Change the primary dimension from Session default channel group to Session source / medium or Session source.

Search for known assistant domains. Depending on the product and route used, identifiable sources may include ChatGPT, Perplexity, Gemini, Copilot, Claude or another assistant. Domain formats can change, so inspect the actual values in the property rather than relying forever on one static list.

Record:

• Sessions.
• Engaged sessions.
• Average engagement time.
• Key events.
• Landing page.
• Market or device where relevant.

Use session-scoped dimensions for traffic acquisition. First-user source answers a different question about how a person was originally acquired and may hide later AI-referred visits.

Create a custom AI-assistant channel

GA4’s default channel grouping may place these visits under Referral. A custom channel group lets you review several assistant sources together without changing the underlying data.

In Admin, open Channel groups and create a new custom group based on the existing structure. Add a channel called AI Assistants and define source conditions using the domains actually present in your reports.

A starting source pattern might cover domains containing terms such as:

chatgpt

perplexity

gemini

copilot

claude

Test the rule against real rows before relying on it. Avoid broad conditions such as “ai”, which could capture unrelated domains. Keep a dated list of included sources and review it monthly because products, redirects and referral behaviour evolve.

The custom group makes reporting easier. It does not recover visits that arrived without referral information.

Use an Exploration for deeper analysis

Build a free-form Exploration using:

• Session source / medium.
• Landing page plus query string.
• Device category.
• Country.
• New or established users.
• Key events relevant to the business.

Filter the exploration to recognised AI sources or the custom channel group. Compare the result with Organic Search, Referral and Direct, but do not assume the channels represent equivalent intent.

AI-referred traffic may be small and highly specific. A visitor could arrive on a detailed PR guide after asking a complex question, spend several minutes reading and leave without triggering a conversion. That session may still indicate strong content relevance.

Measure the landing page first

The landing page shows which content earned the click. Group pages into meaningful types:

• High-intent service and hiring guides.
• AI-search and digital PR articles.
• Media-relations guides.
• Campaign analysis.
• Sector pages.
• Case studies.

Look for patterns. Do visitors arrive on answer-led articles? Do they move from an insight to a service page? Do they read several related posts? Does one page generate enquiries or newsletter sign-ups?

This analysis helps distinguish traffic from visibility. A page may receive few clicks but still be repeatedly cited. Another may attract visits because the title closely matches a practical prompt.

Define meaningful key events

Do not report AI traffic only as sessions. Configure events that reflect a real next step, such as:

• Contact-form submission.
• Consultation booking.
• Click to an email address.
• Download of a useful guide.
• Meaningful scroll or engagement threshold.
• Movement from an insight to a relevant service page.

Check that the event fires correctly and does not double-count. A contact-page view is not the same as a completed enquiry.

For Fireflies, the most useful path is likely to be an insight that establishes authority, followed by a visit to a relevant service, sector, case study or contact page. That is why internal linking remains part of answer-engine optimisation and PR, not merely a navigation detail.

Understand what GA4 cannot tell you

GA4 normally cannot reveal the user’s prompt. It cannot show an impression count for a citation across all assistants. It cannot identify a source when the referring information is stripped. It cannot prove that an AI answer caused a later direct visit.

Some users will read an AI answer, remember the brand and return later through search or direct navigation. Others may see a citation but never click. Treat referral data as one part of the evidence.

Do not attempt to infer sensitive user information from a landing page or small sample. Report trends at a level appropriate to the volume.

Keep unwanted referrals separate

The unwanted-referrals setting in GA4 is intended for domains that are part of the website’s own journey, such as certain payment or account services. Do not add genuine AI assistants to that list simply to tidy the report. Doing so can remove valuable attribution.

Use custom channel grouping for classification. Use unwanted-referral rules only when the domain is not a genuine acquisition source.

Add citation monitoring outside GA4

Pair the traffic report with a repeatable citation check. Use a stable set of prompts related to the business, record whether Fireflies or its content appears and note the cited source.

Also review:

• Google Search Console queries and landing pages.
• Brand searches.
• Earned-media backlinks.
• Referral traffic from cited publications.
• Enquiry forms asking how the person found the business.
• Direct feedback from prospects.
• This combined view is stronger than claiming that one analytics channel captures “AI visibility”.

Annotate changes in the data

GA4 reports are only comparable when the definitions remain visible. Add an annotation in the monthly report when the website migrates, consent settings change, a key event is repaired or a new assistant domain is added to the channel.

Compare like with like. A rise after adding three sources may reflect better classification rather than more demand. A fall after a consent update may reflect measurement loss rather than weaker visibility.

Keep a simple change log containing the date, rule, reason and person who made it. That record will make quarterly analysis far more reliable than trying to reconstruct old settings from memory.

Build a monthly report

A useful monthly section can show:

• Identifiable AI-assistant sessions.
• Sources included in the channel definition.
• Leading landing pages.
• Engagement and key events.
• Movement to commercial pages.
• Citation checks and notable changes.
• Data limitations.

Keep the reporting period and definition consistent. If a new assistant domain is added, annotate the change so an apparent increase is not mistaken for organic growth.

What should you do with the findings?

Strengthen pages that attract relevant visitors. Improve the direct answer, author attribution, evidence, internal links and next step. Create a distinct follow-on article when the data reveals an unmet question. Do not produce several near-identical pages to chase a handful of referrals.

If high-quality articles earn citations but few visits, that may reflect zero-click behaviour. The response is to build stronger brand recognition and useful paths for people who do click, while continuing to earn independent coverage that AI systems can cite.

Across 15+ years in PR, I have measured visibility across changing media channels. AI referrals add another signal. They become meaningful when connected to the content people chose, the authority that brought them there and the action they took next.

If you need a PR and AI-search strategy that connects earned authority with measurable website behaviour, talk to Fireflies Management.