GA4 Shows AI Traffic as Direct: Why & How to Fix It

Last updated: 18 October 2026

GA4 (Google Analytics 4) shows AI traffic as Direct because tools like ChatGPT and Perplexity often strip referrer data or open links in ways that leave no source information. An analysis found that 70.6% of AI-referred visits arrive without referrer headers, so GA4 defaults them to Direct rather than attributing them to the AI platform that sent them (Seresa, 2026).

Key Takeaways

An analysis of 446,000 AI-referred visits found that 70.6% of AI-referred visits land as Direct traffic in GA4 because they arrive without referrer headers, according to AuthorityTech (2026).
For every detected AI referral session, roughly 2.5 additional sessions arrive misattributed as Direct, a 2.5–3x multiplier identified in AuthorityTech's internal data (AuthorityTech, 2026).
ChatGPT drove 76% of all AI Search sessions across 371,847 sessions analysed on Clickport customer sites in April 2026, and 35.7% of that AI traffic arrived with no referrer at all (Clickport, 2026).
Google added a native "AI Assistant" channel to GA4's Default Channel Group on 13 May 2026, requiring no configuration from property owners (Google Analytics Help Center / TechWyse, 2026).
Brands appearing in ChatGPT recommendations were 2.5x more likely to receive a site visit within 7 days than brands that weren't recommended (Similarweb, cited by Seresa, 2026).

Why Does GA4 Classify ChatGPT and Other AI Tool Traffic as Direct?

GA4 is a web analytics platform built by Google that assigns a traffic source to every session based on signals such as the HTTP referrer header, UTM parameters and click IDs. When none of those signals exist, GA4's fallback channel is Direct, which Google's own documentation describes as "website traffic that doesn't have a clear referral source" (Google Analytics Documentation).

AI assistants frequently break the referrer chain in ways a normal browser link doesn't. Many mobile apps for ChatGPT, Claude and Gemini open links inside in-app browsers that don't pass a referrer at all. Some AI chat interfaces render links as plain text that users copy and paste into a new tab, which also strips any source signal. Across UK and global sites alike, this is why a business in Manchester or London can see a sudden spike in Direct sessions that actually originated from an AI recommendation, not a bookmark or typed URL.

This isn't a one-off glitch. AuthorityTech's analysis of 446,000 AI-referred visits found that 70.6% of AI traffic lands as "direct" in GA4 — effectively the default resting state for any session where the browser can't pass on where the click came from.

How Does GA4's Attribution Model Handle Traffic Without UTM Parameters or Referrer Data?

GA4's attribution model works through a strict waterfall: it checks for Google click IDs (gclid), then UTM parameters, then the HTTP referrer, and only falls back to Direct when none of these are present. A UTM parameter — a short code appended to a URL (such as ?utm_source=chatgpt) that tells GA4 where a visitor came from — is the most reliable override, because GA4 reads it ahead of referrer data. If a link has no UTM parameter and the referrer is blank or stripped, GA4 has no basis for attribution and assigns Direct by default.

This matters because AI platforms rarely add UTM parameters themselves. Unless a business has proactively tagged its own content (for PR placements, guest posts, or pages likely to be cited by an AI engine), any resulting click arrives attribution-blank. Clickport's analysis of 371,847 sessions found that 35.7% of AI traffic arrived with no referrer at all in April 2026 — meaning more than a third of that traffic had nothing for GA4's waterfall to work with.

What's the Technical Difference Between Referral and Direct Traffic Classification in GA4?

Referral traffic in GA4 is any session where the browser sends an HTTP referrer header identifying the originating domain, while Direct traffic is any session where that header is missing, blocked, or explicitly stripped. The referrer header is a small piece of metadata a browser sends to the destination server, naming the page the user clicked from — it's how GA4 knows a visitor came from bbc.co.uk or reddit.com, for example.

Several mechanisms strip this header before it reaches GA4:

HTTPS-to-HTTPS downgrades in referrer policy: modern browsers apply strict referrer policies by default, which can omit the header even between secure sites.
In-app browsers: apps like the ChatGPT mobile app often use embedded WebViews that don't forward referrer data the way a standard Chrome or Safari tab does.
Copy-paste behaviour: when a user copies a link from an AI chat response and pastes it into a new browser tab, there is no referring page at all — this is technically correct Direct traffic, not misattribution.
Referrer exclusion list settings: GA4 properties maintain a list of domains to exclude from referral attribution (commonly used for payment gateways). If an AI domain is wrongly added here, or if a redirect chain crosses an excluded domain, genuine referral data gets dropped and the session reverts to Direct.

Businesses auditing their GA4 referrer exclusion list should check it doesn't inadvertently include any AI or redirect domains relevant to their traffic.

Which AI Platforms Are Most Commonly Misattributed as Direct Traffic?

ChatGPT, Perplexity, Google Gemini and Microsoft Copilot are the platforms most frequently misattributed as Direct traffic in GA4, with ChatGPT responsible for the largest single share of this problem. Clickport's April 2026 data found ChatGPT drove 76% of all AI Search sessions across 371,847 analysed sessions, making it the dominant source of AI-driven visits for most UK sites — and therefore the biggest contributor to the Direct-traffic distortion.

"Google Analytics now provides a dedicated way to measure and analyze traffic originating from popular AI assistants" (Google Analytics Release Notes) — a reflection of how significant this misattribution problem had become across platforms including ChatGPT, Gemini, Deepseek, Copilot and Grok.

AI platform	Typical referrer behaviour	Native GA4 AI Assistant channel coverage
ChatGPT (OpenAI)	Frequently strips referrer, especially via mobile app	Yes
Google Gemini	Mixed — web version often passes referrer	Yes
Microsoft Copilot	Often strips referrer in integrated/app contexts	Yes
Perplexity	Inconsistent reporting across sources	Disputed — coverage unclear
Google AI Overviews	Appears within standard Google Search results	Explicitly excluded
Google AI Mode	Appears within Search; distinct from AI Assistant channel	Explicitly excluded

Google's own documentation confirms the native channel covers ChatGPT, Gemini, Deepseek, Copilot and Grok, while explicitly excluding AI Overviews and AI Mode traffic, which remain classified under Organic Search (Google Analytics Help: Default channel group). Whether Perplexity sits inside that native channel is inconsistently reported across current sources, so businesses relying on Perplexity referrals should verify with a custom regex rather than assume native coverage.

How Can I Identify How Much of My Direct Traffic Is Actually From AI Assistants?

Businesses can estimate their hidden AI traffic by cross-referencing Direct traffic spikes against known AI citation activity, rather than taking the Direct figure at face value. AuthorityTech's internal data shows a 2.5–3x multiplier: for every detected AI referral session, approximately 2.5 additional sessions arrive as misattributed Direct. Applying that multiplier to a known AI-referral baseline gives a rough estimate of total AI-driven traffic hiding in the Direct bucket.

A practical, low-cost check:

Filter GA4's Traffic Acquisition report to isolate sessions already tagged with an AI source (via the native AI Assistant channel or a custom channel group).
Compare the trend line of that AI-tagged traffic against the trend line of unexplained Direct traffic growth over the same period.
Check Google Search Console for a parallel rise in branded search queries — a common side effect when AI tools recommend a brand by name and users later search for it directly.
Where possible, run a short on-site survey asking new visitors how they found the business.

This last method isn't theoretical. One site owner surveyed nearly 400 users and found over 40% discovered the site through ChatGPT, while GA4 showed almost all of them as Direct (HackerNoon, cited by Seresa, 2026). It's worth remembering, too, that not every AI interaction produces a click to measure in the first place: a 2026 Pew Research study (cited by Yasser Soliman) of nearly 69,000 searches found users clicked a result on just 8% of searches showing an AI summary, versus 15% without one, and clicked a link inside the summary on only 1% of visits — meaning even perfect attribution would never capture the full scale of AI influence.

"Every article on our own site was researched, written and published by the platform — the byline says so. We don't show prospects a demo corpus; we show them our production one. If the engine couldn't do this for us, we'd have no business selling it to you." — Lauren Dawkins, Head of Content, Aether AI

How Do I Set Up Custom Channel Groupings in GA4 to Separate AI Traffic?

A custom channel group in GA4 is a user-defined set of rules that overrides the Default Channel Group to classify traffic matching specific conditions, such as a referrer domain containing "chatgpt.com" or "perplexity.ai". Google's own documentation confirms that "custom channel groups can be applied to your reports retroactively" (Google Custom Channel Groups Documentation), meaning a business can set this up today and see it reflected in historical data, not just from the date of creation forward.

To build one in GA4:

Navigate to Admin > Data display > Channel groups and select "Create custom channel group".
Add a new channel named "AI Assistants" and define a condition using regex matching against Session source, e.g. matching chatgpt\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com.
Order the AI Assistants channel above Direct in the rule hierarchy, so matching sessions are claimed before they fall through to Direct.
Apply the custom channel group to existing reports to retroactively reclassify historical sessions that do carry referrer data.
Cross-check results weekly against the native AI Assistant channel Google introduced in May 2026, since the two may capture slightly different session sets.

The ceiling on this fix is important to state plainly: a custom channel group can only reclassify sessions that still carry some referrer or UTM signal. It cannot recover the 70.6% of AI-referred visits that arrive genuinely referrer-blank (Seresa, 2026) — those sessions were Direct the moment they hit the browser, and no GA4 configuration can retrieve information that was never sent.

Has Google Fixed This With a Native GA4 AI Assistant Channel?

Google has partially addressed this problem: it added a native AI Assistant channel to GA4's Default Channel Group on 13 May 2026, requiring no configuration from property owners (Google Analytics Help Center / TechWyse, 2026). Google's help documentation states that "you can now identify how users are discover[ing your site through chatbots like ChatGPT, Gemini, and Claude]" (Google Analytics Help Center).

This update is useful but not a complete fix. It only classifies sessions where a referrer or identifying signal already exists — it does nothing for the 70.6% of AI visits that arrive with no referrer at all. SEO commentator Lily Ray has described the broader trend of AI tools obscuring traffic sources as "Not Provided 2.0" (Lily Ray, Amsive), a reference to the historic loss of keyword-level data when Google encrypted organic search in 2013. The parallel is apt: the underlying activity is happening, but the reporting layer can't always see it.

The native channel also explicitly excludes Google AI Overviews and Google AI Mode traffic, both of which continue to appear under Organic Search (Google Analytics Help: Default channel group) — a distinction worth flagging to any UK marketing team reporting AI visibility up to a board or client.

Your GA4 AI Traffic Attribution Checklist

Check whether the native AI Assistant channel (added 13 May 2026) is already populating data in your GA4 property.
Build a custom channel group with regex matching for chatgpt.com, gemini.google.com, copilot.microsoft.com and perplexity.ai.
Apply the custom channel group retroactively to historical reports, not just from today forward.
Audit your referrer exclusion list to confirm no AI or redirect domains are wrongly excluded.
Add UTM parameters to any content specifically created for PR placements, guest posts or AI-citation bait before publication.
Cross-reference unexplained Direct traffic growth against Google Search Console branded query trends.
Track AI citation activity directly, rather than relying solely on GA4 session data, since many AI interactions never generate a click at all.

FAQ

Why does ChatGPT traffic show up as Direct in GA4?

ChatGPT traffic shows up as Direct in GA4 because its mobile app and in-app browsers frequently fail to pass a referrer header, and GA4 defaults any session without a referrer, UTM parameter or click ID to the Direct channel. Clickport's April 2026 analysis found ChatGPT drove 76% of all AI Search sessions, making it the single biggest contributor to this misattribution.

What is the difference between Direct and Unassigned traffic in GA4?

Direct traffic is a specific GA4 channel assigned when no source signal exists at all, while Unassigned traffic appears when GA4 has partial data that doesn't fit any defined channel rule. Both can hide AI-driven visits, but they're triggered by different gaps in the data GA4 receives.

Does GA4's new AI Assistant channel fix the Direct traffic problem?

No, it only partially fixes it. The native AI Assistant channel, added on 13 May 2026, classifies AI sessions that still carry some identifying signal, but it cannot recover the portion of AI traffic — found to be 70.6% in one analysis — that arrives with no referrer data whatsoever (Seresa, 2026).

How do I set up a custom channel group to track AI traffic in GA4?

Go to Admin > Data display > Channel groups in GA4, create a new channel group, and add a regex condition matching known AI domains such as chatgpt.com, perplexity.ai and gemini.google.com against Session source. Google confirms custom channel groups can be applied retroactively to existing reports (Google Custom Channel Groups Documentation).

Does the GA4 AI Assistant channel include Perplexity?

Coverage of Perplexity within the native AI Assistant channel is inconsistently reported across current sources, so it shouldn't be assumed. Businesses that want guaranteed Perplexity tracking should add it explicitly to a custom channel group regardless of native coverage.

How do Google AI Overviews and AI Mode traffic get classified in GA4?

Google AI Overviews and AI Mode traffic are classified under Organic Search in GA4, not under the native AI Assistant channel, because they appear within standard Google Search results rather than through a separate chatbot interface (Google Analytics Help: Default channel group).

Can I retroactively recover AI traffic that already shows as Direct in GA4?

Partially — a custom channel group can retroactively reclassify historical sessions that still carry a referrer or UTM signal GA4 previously filed under Direct. Sessions that arrived completely referrer-blank cannot be recovered retroactively, since the identifying data was never captured in the first place.

Measuring AI Visibility Beyond GA4 With Aether AI

This GA4 attribution problem is exactly why relying on web analytics alone to prove AI-driven visibility falls short for UK businesses trying to understand their return from AI search engines. Aether AI addresses this by tracking citation activity directly across six AI engines — ChatGPT, Perplexity, Google AI Overviews, Claude, Gemini and Copilot — rather than waiting for a click that GA4 may never correctly attribute in the first place.

Aether AI's own production site demonstrates this approach in practice: every article published across its four brands is tracked for citation performance from the moment it goes live, with 281 articles published across those brands in the last 30 days. That same engine — covering automated article generation, citation tracking, keyword and competitor tracking, and Google Search Console integration — is what Aether AI runs for its own content, not a separate demo version shown only to prospects.

If GA4's Direct-traffic bucket is leaving you guessing about how much of your traffic is really coming from AI engines, run a free AI-visibility audit with Aether AI to see where your content is already being cited — and where it's being missed.