GA4 has a channel grouping problem that most PPC teams have not noticed yet. The default AI Assistant channel - designed to capture traffic arriving from AI tools - does not capture it cleanly. Instead, it splits what is effectively a single referral source across multiple channels, so your reports undercount the total volume. If you are using GA4 data to evaluate where your converting traffic comes from, and to feed that back into paid media decisions, you are working with numbers that are quietly off.
This is not a minor rounding error. When a single source is fragmented across three separate channel buckets in GA4's default setup, you lose the ability to see its true contribution to conversions, assisted sessions, or any downstream metric you care about. For an agency or in-house team trying to justify budget allocation or assess whether paid and organic sources are pulling their weight, inaccurate channel data is a genuine problem.
How the Fragmentation Actually Happens
GA4 classifies sessions into channel groups based on a set of default rules that evaluate the medium and source attached to each session. The issue with AI referral traffic is that it does not always arrive with a clean, consistent UTM signature. Different AI tools pass referral data in different ways - sometimes as a direct visit, sometimes with a recognisable referral source, sometimes with parameters that GA4's default rules assign to unrelated channels.
The result is that traffic from what is functionally one source ends up spread across the default channel grouping as three distinct entries. GA4's AI Assistant channel captures some of it, but the rest falls into Direct or Referral or another bucket entirely, depending on how the session was tagged. You are not seeing one number for that traffic source - you are seeing fragments of it in three places, none of which are labelled consistently.
This is a structural limitation of how GA4 applies its default channel grouping logic, not a data collection failure. The data is there - it is just being classified incorrectly because the default rules were not built to handle the variation in how AI tools pass referral information.
Why This Matters for Paid Media Attribution
Channel groupings in GA4 directly affect how you read conversion paths, assisted conversions, and multi-touch attribution. If a user arrives via an AI tool, browses without converting, then returns later via a paid search click and converts, the assisted session from that first touchpoint is being miscategorised. Your paid channel looks like the only contributor. Your attribution model - whether last click, data-driven, or anything in between - is working from a flawed input.
For lead generation specifically, where the consideration cycle is often longer and multi-session journeys are common, misattributed first touches can skew your understanding of which channels are warming up your audience before paid search closes the conversion. That matters when you are making decisions about budget, bid strategy, and where to invest in top-of-funnel activity.
There is also a reporting credibility issue. If you are presenting channel performance data to a client or senior stakeholder using GA4's default groupings, and those groupings are silently undercounting a growing referral source, the story you are telling is incomplete. Custom channel groupings are not an optional refinement - they are increasingly a baseline accuracy requirement.
Building a Custom Channel Grouping That Works
The fix is to create a custom channel grouping in GA4 that consolidates the fragmented traffic under a single, clearly defined rule set. GA4 allows you to build custom channel groups in the Admin section, and these sit alongside - rather than replacing - the default grouping, so you are not breaking anything for anyone else in the account.
The approach is to define a new channel group that uses source-based rules broad enough to capture all the variations in how AI tool traffic arrives. That means looking at the referral sources and session medium combinations that your GA4 data is currently splitting across multiple channels, and writing rules that group them together. You may need to check the raw session source data - available in the Traffic Acquisition report filtered by source - to identify which sources are being miscategorised before you can write accurate rules.
Once built, apply the custom channel grouping to your key GA4 reports and to any Looker Studio dashboards where you are reporting on traffic source performance. The custom grouping will give you a consolidated, accurate view of that traffic's volume and conversion contribution, without the fragmentation that the default grouping creates.
What to Check Before You Trust the Output
Custom channel groupings are only as accurate as the rules you write. Before treating the output as reliable, cross-reference the consolidated channel total against the sum of what was previously spread across the three default buckets. If the numbers do not add up, your rules have gaps - either missing a source variant or overlapping with another channel definition.
Also check your conversion tracking setup. If GA4 is relying on page-based conversion events without solid UTM preservation across the session, you may find that the channel grouping is now accurate but the conversion attribution is still broken at the parameter level. Channel grouping fixes the classification problem - it does not fix UTM stripping or session stitching issues that might exist further upstream in your tracking.
If you are using Google Tag Manager to manage your GA4 configuration, this is also a good point to audit how sessions from referral sources are being handled - specifically whether any redirect or landing page behaviour is stripping the referral information before GA4 can capture it. A clean channel grouping built on dirty session data is still going to give you wrong numbers.
The Broader Point About Default GA4 Configurations
GA4's default settings are a starting point, not a finished measurement setup. The default channel grouping was built around a world of fairly predictable traffic sources - organic search, paid search, direct, email, social. As the referral ecosystem gets more varied and less consistent in how it passes session data, the default rules increasingly fail to keep up.
This particular issue with AI referral traffic is a clear example of that pattern. The default grouping was not designed with the variety of AI tool referral behaviour in mind, so it produces fragmented output. The fix requires deliberate configuration - which means someone in your team or agency needs to own the GA4 setup and treat it as a living configuration, not a one-time installation.
For PPC teams, the stakes are practical. Channel data informs budget decisions, bid strategy inputs, and how you explain performance to clients. If the underlying attribution data is fragmented by a default setting nobody has revisited, the decisions built on top of it are compromised. Reviewing and customising your GA4 channel groupings is unglamorous work - but it is the kind of measurement hygiene that separates accounts with reliable data from those that are flying partially blind.