Referral traffic is the visits that arrive because someone else put a link to your site on theirs: a blog post citing you, a news story, a forum thread, a directory that actually gets read. In Google Analytics 4 it is one channel among several in the default channel group, and GA4 decides membership with explicit rules rather than judgement.
Knowing those rules matters because the channel report is only as honest as its inputs. If you know why a session was classed as Referral rather than Direct or Organic Search, you can tell a real link from a reporting artefact. This guide works from Google’s own channel definitions and stays clear of numbers it cannot support.
Referral traffic definition in GA4
Google’s help page on the GA4 default channel group defines it plainly: “Referral is the channel by which users arrive at your site via non-ad links on other sites/apps.” The examples Google gives are blogs and news sites. The rule behind the definition is short: the session’s medium is one of “referral”, “app” or “link”.
Two words in that definition do the work. “Non-ad” means paid placements belong in the paid channels, not here. “Other sites/apps” means the link lives somewhere you do not control, which is exactly why the channel is worth watching. A referral is a link someone else chose to place, and it is the part of your backlink profile that sends people as well as signals.
Medium values come from the referring request or from tags you add yourself. If you tag a link with a medium of “referral”, GA4 will class it as Referral regardless of where it sits, which is one reason consistent UTM parameters matter so much for clean reporting.
Direct vs referral traffic, and the other channels
Referral is easiest to understand next to its neighbours. The table below sets out the rules GA4 uses for the channels people most often confuse with it.
| Channel | Google’s description | Rule (simplified) |
|---|---|---|
| Referral | Non-ad links on other sites or apps | Medium is “referral”, “app” or “link” |
| Direct | A saved link or a typed URL | Source is exactly “(direct)” and medium is “(not set)” or “(none)” |
| Organic Search | Non-ad links in organic search results | Source matches Google’s search-site list, or medium is “organic” |
| Organic Social | Non-ad links on social sites | Source matches a social-site list, or medium is social, social-network, social-media, sm, “social network” or “social media” |
| AI Assistant | Assistants such as ChatGPT, Gemini, Deepseek, Copilot and Grok | Medium “ai-assistant” |
| Unassigned | Traffic that matches no rule | Falls through every other rule |
The direct vs referral traffic distinction is the one that trips people most. Direct, in Google’s words, is “the channel by which users arrive at your site/app via a saved link or by entering your URL.” When referrer information is lost on the way, a visit that began with a click on someone else’s link can end up as Direct instead. Our guide to direct traffic and SEO covers why that channel is often a catch-all.
One detail worth knowing: Google’s AI Overviews and AI Mode are not in the AI Assistant channel. They fall under Organic Search. So a drop in Referral is not explained by searchers moving to Google’s AI features; those visits were never Referral in the first place.
How to find referral traffic in GA4
The standard route, in general GA4 use, is the traffic acquisition report:
- Open Reports, then Acquisition, then Traffic acquisition.
- Set the primary dimension to Session default channel group and find the Referral row.
- Filter to Referral, or add Session source as a secondary dimension to see which sites sent the visits.
Session source is where the report becomes useful. It turns a single channel total into a list of referring domains, and that list is often a better prospecting sheet than a backlink export, because every row is a link that actually sends people.
Pick the scope deliberately. GA4 offers three versions of the dimension: Default channel group is event-scoped, Session default channel group is session-scoped, and First user default channel group is user-scoped. For “which sites send me visits”, the session-scoped version is the natural fit. For “which channel first brought these users”, use the first-user version, and expect the totals to differ.
Set the date range wide enough to be useful. A single week of referral data is mostly noise for a small site, because one shared link can dominate it. A quarter or more shows which referrers send visits steadily and which were one-off spikes. Sort by sessions, then look at engagement alongside them: a referrer that sends fewer visits but holds attention is often worth more than a large one that bounces.
It also helps to note the landing page for each referrer. Adding the landing page dimension shows which of your pages earned the link, and that is usually more actionable than the referring domain alone.
Self-referrals, spam and other distortions
This section is judgement rather than Google’s wording. Referral is the channel most easily polluted by sources that are not real recommendations.
- Payment gateways. A buyer who leaves to pay and returns can arrive back as a referral from the payment provider, breaking the original session’s attribution.
- Your own subdomains. If a help centre, app or shop on a subdomain is not configured as part of the same property, visits between them can look like referrals from yourself.
- Referral spam. Fake hits that carry a referrer purely to appear in your reports. They inflate the channel without sending a single real reader.
The fix in each case is to review Session source regularly and treat unfamiliar or self-owned domains with suspicion. If a domain sends many sessions with no engagement, it is probably not a reader.
Using referral data for link work
Once the noise is removed, the referral list is a map of who already values your work. Three uses stand out. First, compare it with your backlink data: links that appear in a backlink check but never send a session are signals only, while links in both lists are doing double duty.
Second, watch for referrers that disappear. A site that sent steady sessions and then stops may have removed or broken the link, which is a candidate for link reclamation. Third, look at which pages attract referrals and ask what made them citable; that tells you what to publish next.
A stated limitation: GA4 can only classify what it receives. Lost referrers, blocked tracking and consent choices all remove sessions or move them between channels, so referral totals are a floor, not a census. For how analytics compares with Search Console, see Search Console vs Google Analytics, and for the full picture of channels, our guide to website traffic sources.
Frequently asked questions
What counts as referral traffic in GA4?
Sessions whose medium is “referral”, “app” or “link”. Google describes it as users arriving via non-ad links on other sites or apps, such as blogs and news sites.
What is the difference between direct and referral traffic?
Direct requires the source “(direct)” with medium “(not set)” or “(none)”, meaning a saved link or typed URL. Referral is a click on a non-ad link on another site.
Does traffic from ChatGPT count as referral?
GA4 has a separate AI Assistant channel, with medium “ai-assistant”, covering assistants such as ChatGPT, Gemini, Deepseek, Copilot and Grok.
Is traffic from Google AI Overviews referral traffic?
No. Google states that AI Overviews and AI Mode fall under Organic Search, not the AI Assistant or Referral channels.
The takeaway Referral traffic is defined by medium, not by intuition. Read it by Session source, strip out payment gateways, subdomains and spam, and what remains is a list of sites that chose to send you readers.
