AI assistants are sending visitors to your site right now, and GA4 is almost certainly logging them as Direct traffic. The fix is a custom channel group — specifically an AI Referrals channel — built inside GA4’s channel settings. It takes under an hour, costs nothing, and immediately separates ChatGPT, Perplexity, Gemini and Copilot visits from your organic and direct baselines. Here’s exactly how.
Quotable definition: A GA4 AI Referrals channel is a custom channel group rule that captures sessions referred from generative AI platforms — including chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and similar — and groups them as a distinct acquisition source. Without it, most AI-referred sessions are misclassified as Direct because the referrer header is stripped when a user clicks a link inside a chat interface, making the traffic invisible to standard channel reports.
Why Standard GA4 Misses AI Traffic
When someone copies a URL from a ChatGPT answer and opens it in a new tab, there’s no HTTP referrer. GA4 sees a session with no source, no medium, and assigns it to the Direct bucket. The same happens with some Perplexity clicks on mobile. Gemini on Android occasionally passes a referrer, but inconsistently.
The result: your Direct channel is quietly inflating. You may already have more AI-referred visits than you think — you just can’t see them yet. Building a dedicated AI Referrals channel won’t recover historical sessions, but it starts clean attribution from today forward, which is the right baseline to build strategy on.
One important caveat up front: Google Search Console added a ‘Search Generative AI’ performance report on 3 June 2026, covering impressions from AI Overviews. It records impressions only — no clicks. GSC and GA4 solve different problems. You need both.
What to Include in Your AI Referrals Channel
Before you open GA4, decide which domains belong in the channel. The core list for a Singapore SME in 2026:
| Platform | Domain(s) to match | Notes |
|---|---|---|
| ChatGPT | chatgpt.com, chat.openai.com | Both still active; match both |
| Perplexity | perplexity.ai | Passes referrer on most desktop clicks |
| Google Gemini | gemini.google.com | Referrer inconsistent on mobile |
| Microsoft Copilot | copilot.microsoft.com, bing.com/chat | Match both legacy and new URLs |
| Claude | claude.ai | Low volume SG currently; include anyway |
| You.com | you.com | Niche but used in SG B2B research |
This list will grow. Build the habit of checking your Source/Medium report monthly for unfamiliar referrers — new AI platforms surface quickly.
Step-by-Step: Building the GA4 AI Referrals Channel
- Open GA4 → Admin → Data Display → Channel Groups. You’ll see the default channel group Google provides. Do not edit the default — create a new one by clicking Create new channel group. Editing the default causes problems if Google updates it.
- Name the group. Call it something clear: Custom Channels 2026 or [YourBrand] Channel Groups. You’ll add AI Referrals as one channel inside this group, alongside copies of your existing channels (Organic Search, Paid, Direct, etc.).
- Add a new channel: AI Referrals. Click Add channel. Name it AI Referrals. Under conditions, set: Session source contains — then add each domain from the table above as a separate OR condition. Use “contains” rather than “exactly matches” to catch subdomain variants.
- Rebuild your other channels first. GA4 evaluates channels top-to-bottom and assigns a session to the first match. Place AI Referrals before Direct in your ordering, otherwise stripped-referrer sessions still fall into Direct. Rebuild Organic Search, Paid Search, Email and Social above it.
- Set AI Referrals position above Direct. Drag it to sit just above Direct in the list. This is the single most common mistake — people build the channel correctly but leave it at the bottom, where Direct catches everything first.
- Save and publish. GA4 applies the new group to data from this point forward. Historical data is not reprocessed. Set a calendar reminder to check the channel in 14 days with a real-time filter to confirm it’s firing.
- Apply the custom channel group to your reports. In Acquisition → Traffic Acquisition, use the channel group selector (top-right of the table) to switch from the default to your custom group. It won’t appear automatically — you must select it each time, or pin it as the default in the report settings.
Add UTM Parameters Where You Can
The referrer-stripping problem is structural — you can’t fix it for organic AI citations. But if you’re running any paid placements or sponsored content that links to your site from AI-adjacent platforms, UTM-tag every URL. Use utm_source=chatgpt and utm_medium=ai-referral consistently so those sessions land cleanly in your channel even when the referrer drops.
Some AEO practitioners also tag their own press releases and third-party articles with UTMs, then watch which of those sources drive AI referrals back. It’s a useful signal for understanding which content placements are genuinely being cited. Our AEO/GEO/SEO service includes a citation-source audit as part of the monthly retainer — useful once you have six weeks of clean GA4 data to work from.
What the Data Will (and Won’t) Tell You
Once the channel is live, you’ll see sessions, engaged sessions, engagement rate, conversions and revenue attributed to AI Referrals — the same metrics as any other channel. That’s genuinely useful. You can compare conversion rate between AI-referred visitors and organic visitors, spot which landing pages AI is sending people to, and track whether AI referral volume is growing month-on-month.
What it won’t tell you: how often your brand was mentioned without a click, how many AI responses included you but the user didn’t follow the link, or your share of citations versus competitors. GA4 only sees the session that arrives. For impression-level visibility, you need the Search Console AI report for Google’s products, and separate monitoring tools for other platforms.
Here’s the inconvenient truth: AI citation drives a small fraction of total clicks today. If your primary goal is driving meaningful traffic volume this quarter, AI referral optimisation is not the lever to pull first. Fix your existing SEO foundation, then build AI visibility on top of it.
Reading the Numbers Honestly
Before you get excited or alarmed by early AI Referrals data, hold these two facts in mind. First: the same prompt returns the same brand list less than 1-in-100 times, according to SparkToro research — there is no stable “rank” in AI answers. Your presence in an AI response is probabilistic, not positional. Second: according to BrightEdge analysis, brands with fewer than roughly 50 AI citations see their visibility swing by 50% or more week-to-week. Below that threshold, week-on-week GA4 comparisons are noise, not signal.
The practical implication: don’t optimise for a single week’s data. Look at four-week rolling averages minimum. A spike in AI Referrals one week followed by a dip the next is normal variance — it doesn’t mean your content fell out of favour.
Keeping the Channel Accurate Over Time
AI platforms launch and rebrand faster than any channel list can keep up with. Build a quarterly audit into your GA4 maintenance routine: pull the Source/Medium report, filter for sessions not matching any of your defined channels, and scan for new AI referrer domains. Add them to your AI Referrals channel rule promptly. Failing to do this means your Direct channel gradually re-inflates as new AI platforms emerge — exactly the problem you built the channel to solve in the first place.
Also worth noting: if you use a consent management platform (CMP) under PDPA compliance requirements, cookieless sessions from users who decline tracking will still appear as Direct regardless of referrer. That’s a separate attribution gap — not a GA4 channel problem, but worth knowing when you’re interpreting the numbers.
Connecting GA4 Data to Your AEO Strategy
A properly configured AI Referrals channel doesn’t just satisfy curiosity — it gives you the measurement foundation to make AEO and GEO investment decisions with actual data. If AI-referred visitors convert at 2× the rate of organic visitors, that’s an argument for prioritising structured content and FAQ schema. If they bounce immediately, that’s an argument for fixing the landing pages AI is linking to.
Measurement without strategy is just numbers on a screen. Strategy without measurement is guesswork. The two need to run together, which is why we build GA4 AI Referrals configuration into every engagement from week one.
If you’re not sure whether your site is being cited by AI platforms at all — or whether your current GA4 setup is hiding that traffic — the logical starting point is a quick audit. Our free AI-Visibility Check reviews your current citation footprint, flags GA4 attribution gaps, and tells you honestly whether AI search is generating measurable opportunity for your business right now. No sales call required to get the results.
Frequently Asked Questions
Will this channel capture all my ChatGPT traffic?
Not all of it. When a user copies a URL from ChatGPT and opens it in a new tab, there’s no referrer header — GA4 logs it as Direct, and there’s no technical fix for that. The AI Referrals channel captures sessions where the referrer header is passed correctly, which happens more reliably on desktop browsers and on platforms like Perplexity that are built with referrer passing in mind. Think of it as a floor, not a ceiling.
Does editing the default GA4 channel group cause problems?
Yes, potentially. Google occasionally updates the default channel group’s logic, and if you’ve edited it, your changes conflict with the update — sometimes silently. The safe approach is to leave the default untouched and build a new custom channel group that you fully control. Apply your custom group to reports manually via the channel group selector.
How long before I have enough data to act on?
For most Singapore SMEs, expect four to eight weeks before AI Referrals volumes are large enough to be statistically meaningful. BrightEdge data suggests brands with under ~50 AI citations see 50%-plus weekly swings in visibility — your GA4 traffic will reflect that variance. Set a 28-day comparison window as your minimum analytical unit, and resist drawing conclusions from a single good or bad week.
Will Google Search Console tell me the same thing?
Different things, not the same thing. Google added a ‘Search Generative AI’ performance report on 3 June 2026, which shows impression data from AI Overviews — but no click data. GA4 shows you clicks that arrived. GSC shows you impressions that may or may not have led to clicks. You need both: GSC for visibility, GA4 for behaviour and conversion.
Should I track AI referrals separately for each platform?
Only once you have volume. For most SMEs, a single AI Referrals channel is sufficient to start. Once you’re consistently seeing 50-plus sessions per month from the channel, it’s worth creating a secondary dimension breakdown by source to see which platforms (Perplexity vs ChatGPT vs Gemini) are driving traffic — and whether conversion rates differ. Before that threshold, the data is too thin to segment reliably.
Is there an easier way to do this — a GA4 template?
GA4 doesn’t support importable channel group templates as of mid-2026 — you build the rules manually each time. Some third-party analytics tools offer pre-built AI source groupings, but they sit outside GA4 and create a separate data silo. The manual build inside GA4 is slightly tedious but gives you one clean source of truth. It takes about 45 minutes the first time.
What if my site gets almost no AI referral traffic right now?
That’s useful information, not a failure. It means your content isn’t being cited — yet — and points to a gap in structured content, topical authority, or off-site citation signals. The GA4 channel gives you a live baseline. Improving citation probability is what AEO and GEO work does. Set up the measurement first; then you can track whether optimisation efforts are moving the number.