Tracking ChatGPT Referral Traffic in GA4

One of the biggest analytics blind spots right now is how companies track chatgpt referral traffic ga4 configurations.

That gap makes content investment hard to justify, because the leads it produces arrive looking like generic direct visits. ChatGPT is now part of ordinary monthly habit for a large share of Australians, so plenty of your buyers are vetting you in a conversation before they ever open a search engine.

We know that setting up proper measurement takes an hour of focused effort. Without it, six months of ChatGPT optimisation produces numbers with nothing to compare them to.

An accurate baseline is critical, and it is the first thing we establish on any AI SEO in Perth engagement.

We are going to break down the exact steps needed to isolate these sessions, explain why the numbers often look deflated, and outline a practical reporting structure.

Set this up before you start the work

Proper configuration gives you a clear baseline before you launch any AI search campaigns. Tracking this data accurately ensures you can measure the actual impact of your visibility efforts rather than guessing.

ChatGPT use in Australia has grown quickly, and most businesses still have no way to isolate the traffic it sends. The setup is straightforward once you know where to look.

Understanding what the numbers mean is the harder half, and it is where this guide spends most of its time.

We find that a clean setup prevents you from confusing an AI-driven visit with a generic direct session. This distinction determines whether you can confidently invest in your generative engine strategy.

The referrers to watch

AI assistants pass a referrer string when a user clicks an attribution link, which means you must monitor specific hostnames in your analytics. The hosts worth grouping include the major generative engines that explicitly cite sources.

The list changes as platforms evolve. Perplexity in particular has become a meaningful share of AI-driven visits in some B2B categories, so it is worth including from the start.

New hosts appear as products change and new models launch. We strongly advise reviewing this list quarterly rather than assuming it is complete. Staying updated ensures your data remains accurate as user behavior shifts.

Finding them in GA4

Start in Reports, then Acquisition, then Traffic acquisition to locate these sessions and track chatgpt referral traffic ga4 natively. Change the primary dimension to Session source and add a search filter for one of the hosts above.

Our recommendation is to check the new default “AI Assistants” channel that GA4 rolled out in 2026, which automatically groups traffic with an ‘ai-assistant’ medium. If you have had any AI-referred traffic, it appears here, usually classified under Referral or this new default grouping.

If nothing appears, check two things before concluding you have none.

We suggest ensuring your date range is wide enough first. These numbers are often small, and a seven-day window may genuinely be zero.

Second, verify that a consent banner or cookie configuration is not suppressing referrer data. Our experience shows thousands of Australian businesses lose tracking visibility due to strict Consent Mode v2 setups that block analytics until a user explicitly opts in.

Building a reusable AI channel

For ongoing reporting, you must build a permanent filter instead of relying on manual searches. You achieve this by creating a dedicated exploration segment and a custom channel group using a regex filter.

Standard GA4 properties limit you to just two custom channel groups per property. You must plan your taxonomy carefully to avoid hitting this limit.

• As an exploration segment. Explore, then a blank exploration. Create a segment scoped to Session. Condition: Session source matches regex chatgpt\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai. Name it “AI assistants”. Apply it to a free-form table with landing page as the row dimension and sessions, engaged sessions and conversions as values.
• As a custom channel group. In Admin, under Data display, open Channel groups and create one based on the default. Add a new channel named “AI assistants”, positioned above Referral so it captures those sessions first, using the same regex against Source. This makes the channel available in standard reports rather than only in explorations.

We prefer the custom channel group approach because it integrates smoothly into daily reporting workflows.

The landing-page dimension is the useful one here. It tells you which pages assistants actually send people to, which often contradicts expectations and provides directly actionable insights for your content strategy.

Why the numbers look small

AI referral numbers appear artificially low because mobile apps like the ChatGPT iOS client strip the referrer header before the user ever reaches your site. This technical limitation means a significant portion of AI-influenced traffic is completely invisible by default.

Our agency sees clients panic when they expect hundreds of visits but only log a handful. Expect single or double digits per month for most businesses, even where the AI-search work clearly has an effect.

Four distinct reasons explain this gap, and none of them are fixable by better tracking configuration.

• Most AI-influenced journeys do not include a click. The assistant names three providers. The user reads, forms an impression, and searches your brand later. That session arrives as organic branded or direct.
• Attribution links are one option among several. Users read the answer far more often than they follow the sources.
• Some sessions carry no referrer at all. App-based and desktop-client sessions frequently arrive as direct.
• The influence is often several steps back. An assistant conversation in week one shapes a shortlist. The enquiry arrives in week five through a branded search.

We advise treating the referral count as a firm floor, not a comprehensive measurement. Treat movement in it as directional evidence and pair it with the broader approach in measuring AI search visibility.

Things that quietly break the tracking

Tracking breaks silently when technical misconfigurations drop the necessary tracking parameters. These issues obscure valid sessions and distort your entire analytics picture.

Our technicians constantly fix setups where a single checkbox update wipes out entire traffic channels. Four specific configuration problems account for most cases where the numbers look wrong.

• Consent mode suppression. If your consent banner blocks analytics until acceptance, sessions from users who never accept are missing entirely. That affects every channel, and it hits small-volume channels hardest because a handful of lost sessions is a large proportion of the total.
• Internal traffic filters. A filter excluding your office IP range is fine. One excluding a broad range, or a stale range from a previous office, silently removes real sessions.
• Cross-domain gaps. If your enquiry form lives on a booking subdomain that is not configured for cross-domain measurement, the session restarts at the handover. The required _gl URL parameter drops, and the original AI referrer is lost right before the conversion.
• Referral exclusion lists. Adding a host to the referral exclusion list makes those sessions direct. Check that nobody has added an assistant domain there in an attempt to clean up the report.

We always check all four areas before concluding a site has no AI traffic.

In practice, the most common cause of a zero is a configuration issue rather than a genuine absence of visitors.

What assisted conversions can and cannot prove

Assisted-conversion data proves that an AI-sourced session participated somewhere in a user’s journey to conversion, rather than just acting as the final click. This is genuinely useful, and it represents the strongest analytics evidence available for this channel right now.

This is the report that shows the hidden value of top-of-funnel content. It cannot prove that the AI interaction directly caused the conversion.

GA4 uses a data-driven attribution model that redistributes credit among the touchpoints it can see, but it cannot measure impressions.

The largest part of AI influence involves a user reading a brand name without clicking, an action that is invisible to analytics by construction.

We find the honest framing for a board report is that referral and assisted-conversion figures are a visible fraction of a larger effect, with the actual fraction unknown. Anyone quoting a precise multiplier to correct for it is inventing the multiplier.

Filling the gap with proxy signals

Because the direct referral number acts only as a floor, you must pair it with proxy indicators that move for the same underlying reasons. Monitoring these secondary metrics provides a much clearer picture of your actual brand visibility.

Our strategy uses a triangulation method to evaluate campaign success. Pair your referral data with these three specific indicators to build a stronger case for your generative search strategy.

• Branded organic search volume. Pull impressions and clicks for queries containing your business name from Search Console. An assistant that named you produces exactly this behaviour a few days later. It is noisy and affected by everything else you do, so read direction over months rather than weeks.
• Direct traffic to deep pages. Direct sessions landing on a specific service or guide page, rather than the homepage, usually means someone was given that URL directly. Some of that is email and word of mouth. A steady rise alongside AI-search work is worth noting.
• Enquiry source questions. The single cheapest measurement upgrade available is a “how did you hear about us” field on your enquiry form. Include an explicit option for an AI assistant. Self-reported attribution is imperfect, yet it remains better data than anything the analytics platform can natively provide for this channel.

We know that none of these prove absolute causation. Together with the referral figure they give you three imperfect views of the same effect, which is the realistic standard for this channel right now.

Setting up the same view in Looker Studio

You can recreate this exact reporting structure in Looker Studio by connecting your GA4 property and applying the identical regex filter to the session source. If you report outside GA4, the configuration transfers directly and easily.

Build the dashboard the same way. Connect your GA4 property, add a filter on Session source using the regex chatgpt\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai, and build a time series of sessions and conversions.

Use the landing page as a breakdown dimension.

We advise adding a scorecard for branded organic clicks from a Google Search Console connector alongside your GA4 data, so both critical signals appear on one unified page.

Add a text block stating the known under-count explicitly. A dashboard without that caveat will be read as a complete measurement, and the person reading it in six months will not remember the conversation you had when you built it.

What to report, and how often

Report AI sessions, conversions, and top landing pages on a monthly basis to track the ongoing trend rather than obsessing over absolute numbers. Keeping the reporting cadence consistent prevents overreactions to minor daily fluctuations.

Our standard recommendation is a strict, four-line format for monthly updates. Include sessions and conversions from the AI channel, noting the trend over time. List the top landing pages from that channel, provide branded organic search volume as a supporting indicator, and document work completed against the roadmap.

We suggest adding prompt coverage movement to your quarterly review, since AI models update their indexes periodically.

Day-to-day prompt visibility moves too slowly to be worth reporting monthly. The mechanics behind those slower algorithmic movements are covered in how ChatGPT finds business information.

If you want the equivalent question for Google’s surface, read our breakdown on AI Overviews and organic traffic.

Conclusion

Taking the time to track chatgpt referral traffic ga4 metrics is no longer optional for serious businesses. It is the only way to prove that your Generative Engine Optimisation efforts are actually driving revenue.

We highly recommend auditing your current cross-domain tracking setup this week.

Start by verifying your custom channel groups and checking your consent mode configurations.

If you need a second pair of eyes on your analytics architecture, get in touch with WestAI Search for a full audit.

Accurate data changes everything.