TL;DR: GA4 first sorts visits into channels, then decides which channels receive credit for enquiries or sales. Different reports use different rules, so their channel totals can differ without either being wrong. Check the report’s scope, your attribution settings and the quality of your tracking before judging a channel’s performance.
GA4 shows channel numbers in several places, and they often disagree. Each report answers a slightly different question. This guide explains how GA4 sorts traffic into channels, how it then shares credit between them, and what to check before you trust a channel report. If you also need to present this to a client or manager, our guide to SEO reporting and attribution covers the reporting side.
How Does GA4 Sort Traffic Into Channels?
Every visit arrives with a source (where it came from), a medium (how it got there) and sometimes a campaign name. Our GA4 glossary explains source and medium in more detail. GA4 runs these values through the default channel group, which Google describes as rule-based definitions of your traffic sources. Traffic from Google Ads and other ad platforms is sorted using details such as the ad platform and campaign type instead. You cannot edit the default channel group, but you can build a custom channel group if you need different rules.
| Channel | What puts a visit there |
| Direct | The source is (direct), and the medium is (not set) or (none). |
| Organic Search | The source is a recognised search site, or the medium is exactly “organic”. |
| The source or medium matches “email” or a close variant such as “e-mail”. | |
| Referral | The medium is “referral”, “app” or “link”. |
| Unassigned | No channel rule matches the data. |
Direct and Unassigned cause most of the confusion. Google describes Direct as people arriving by a saved link or by typing your address. The rule itself only checks for (direct) as the source with no useful medium, so visits that reach GA4 without source information may also land there. A returning visitor who opens a bookmark can also be reported under an earlier source rather than as Direct: in Google’s own example, a bookmarked visit appears under the affiliate source from the visitor’s previous session. Unassigned is different: GA4 had data, but no rule matched it. Google lists several causes of missing source data: a Google tag that does not fire on the Initialisation trigger, so the session_start event is missing; incomplete or incorrect UTM tagging on links; and consent mode set up in a way that creates sessions without session_start. When session source and medium are (not set), the traffic shows as Unassigned in the default channel group.
Top Tip: Open the Traffic acquisition report and check the Unassigned row before you debate attribution models. If it is more than a small slice of your sessions, investigate the source data and tracking setup first. Check campaign tags, whether the Google tag fires on Initialisation, and any consent setup issues that could leave sessions without a session_start event. Changing the attribution model will not fix missing source information.
Where Does AI Traffic Appear in GA4?
AI traffic does not all appear in one channel. Google includes clicks from AI Overviews and AI Mode in Organic Search. Visits from recognised assistants, such as ChatGPT, Gemini and Claude, have a separate AI Assistant channel. Google assigns this channel when the referrer matches a recognised AI assistant.
The AI Assistant total therefore does not represent all visits from AI experiences. Google’s AI search clicks sit within Organic Search, and the Organic Search channel total does not separate them from clicks on traditional search results.
There is another limit: GA4 records activity on your website. If someone reads an AI answer that recommends your business but never visits your site, that exposure will not appear as a website visit. A low AI Assistant total does not, on its own, prove that your business is rarely mentioned in AI answers.
To check the traffic GA4 has recorded:
- Open the Traffic acquisition report and select Session default channel group.
- Find AI Assistant and add Session source / medium as a secondary dimension to inspect the sources behind those visits.
- Compare sessions with recorded key events, such as enquiries or purchases. Check which pages these visitors land on in a landing-page report or Exploration filtered to the same channel.
This helps you assess what identifiable AI visits contribute to your website. It does not measure every AI mention or recommendation, and small numbers of visits or enquiries are a limited basis for judging performance.
AI Assistant is a channel, not an attribution model. It describes how GA4 classified the traffic. Which channel gets credit for a later enquiry still depends on the dimension’s scope and attribution rules explained below.
How Does GA4 Decide Which Channel Gets the Credit?
Sorting tells GA4 what each visit was. Attribution decides which channel gets credit for a key event, which is an action you have marked as important to the business, such as an enquiry form. GA4 does not use one rule everywhere. The rule depends on the scope of the dimension, and the dimension’s name shows which scope it is.
| Dimension name | Where you will see it | Attribution rule |
| Starts with “First user” | User acquisition report | Paid and organic last click |
| Starts with “Session” | Traffic acquisition report | Paid and organic last click |
| No prefix, such as Default channel group | Model comparison and Key event paths reports, and Explorations | The model chosen in your property’s attribution settings |
The User acquisition report looks at where new users came from. The Traffic acquisition report looks at where new and returning users came from each time they started a session. Google says both use the paid and organic last click rule. User acquisition shows how each user first arrived, but it does not use a first-click model. Event-scoped traffic-source dimensions, such as Default channel group without a prefix, use the reporting attribution model you choose. This also applies when you use those dimensions in Explorations.
What Do the Attribution Models Do?
In its attribution reports, Google currently offers three models:
- Data-driven is the default. It uses machine learning on paths that did and did not end in a key event to share credit across the touchpoints, and it is built separately for each key event.
- Paid and organic last click gives all the credit to the last channel the person clicked (or, for YouTube, engaged with) before the key event.
- Google paid channels last click gives all the credit to the last Google Ads channel clicked. If the path has no Google Ads click, it falls back to paid and organic last click.
First click, linear, time decay and position-based models were removed as of November 2023, so you may still see them described in older guides. Google also says every model leaves Direct visits out of the credit, unless the whole path was Direct.
A Worked Example
Say someone finds you through Google search on Monday, clicks a paid search ad on Wednesday and types your web address on Friday to send an enquiry, all within the lookback window. This is an illustration, not a client result.
- Paid and organic last click: Paid Search gets all the credit, because Friday’s Direct visit is ignored.
- Data-driven: credit is shared between the touchpoints in proportions the model estimates, so Organic Search and Paid Search can each receive a part. Direct receives none.
When reviewing one of our client’s year-on-year performance, we noticed that the sales and revenue figures in GA4 weren’t lining up with the client’s actual sales. We went back to WooCommerce and compared the purchase records against the transactions reported in GA4 to understand the gap. This gave us a more reliable view of how sales were performing year on year, without overstating or understating results based on GA4 alone. It’s why I always think it’s important to validate ecommerce reporting against actual sales data, particularly when making decisions based on revenue performance. – Fiorela Imerai, SEO Account Director
What Should You Check Before Trusting a Channel Report?
- Compare like with like. Check whether each dimension starts with “First user”, “Session” or nothing. Reports with different scopes will not match, and that is expected.
- Check the reporting attribution model. It sits in the attribution settings, and you need at least the Marketer role to change it. Changing it applies to historical as well as future data in key event reports. Google says it does not affect the user-scoped and session-scoped dimensions.
- Know your lookback window. By default, GA4 looks back 30 days for acquisition key events (first_open and first_visit) and 90 days for all other key events. The other key events can be set to 30 or 60 days, and changes apply going forward only.
- Tag campaign links consistently. The channel rules look for particular source and medium values, such as an email medium for email campaigns. Our GA4 for SEO guide covers setup and checking your data.
- Judge SEO on key events as well as sessions. Under last click, a paid or email visit just before the enquiry takes the credit. The Model comparison report shows how different models change the value of each channel, so you can see how much credit organic search gets under each.
“When a client’s GA4 numbers don’t match what they’re seeing in the business, the first thing I check is what GA4 is actually counting. In the Traffic acquisition report, a key event is an action, such as clicking an email address, tapping a phone number or submitting a form, credited to the channel that started that session. It can’t tell you whether a real enquiry followed, whether it was relevant, or whether it turned into a booking. It also misses enquiries that never trigger an event, such as someone who calls or emails without clicking a link, or someone who doesn’t accept cookies. For a first aid training client, we log every organic key event in a shared lead sheet. The client confirms the outcome of each one and adds any website enquiries GA4 didn’t capture, so we get the full picture. Over 14 months, 35 logged enquiries weren’t relevant, 83 were genuine leads, and 19 became bookings. That’s what we report on, not the raw GA4 total.” – Kezia Humphries, SEO Account Manager.
How Wildcat Handles This
Wildcat’s SEO consultancy service includes reporting and tracking set-up and regular analysis and reporting, aimed at more than vanity metrics. If your GA4 channel numbers do not match the enquiries and sales you can see, you can arrange a free consultation to talk it through.
