ChatGPT is already sending your med spa website visitors. GA4 just files them under generic “Referral” or, worse, “Direct,” where you’ll never notice them. In this guide I give you the finished copy-paste regex and the exact 10-minute setup I use for every clinic I work with, so you can see AI traffic as its own channel starting today. Regex last updated July 2026.
Why GA4 buries AI traffic by default
GA4 classifies sessions into default channels using a fixed list of rules. Search engines get “Organic Search,” social platforms get “Organic Social”; but chatgpt.com, perplexity.ai, and claude.ai aren’t on any of those lists, so they fall into the catch-all Referral channel, mixed in with every directory listing and blog backlink pointing at your site.
It gets messier. A large share of AI-assistant clicks arrive with no referrer header at all: the ChatGPT mobile app and some desktop interactions strip it when opening external links. GA4 has nothing to go on, so those sessions land in Direct, indistinguishable from someone typing your URL. Analysis published by Authority Tech in April 2026 found that 33.1% of measured AI-search sessions (867 of 2,619) arrived with a utm_source=chatgpt.com tag but no HTTP referrer, and other 2026 analyses put the total no-referrer share of AI clicks anywhere from a third to well over half depending on device mix.
One helpful quirk works in our favor: since mid-2025, ChatGPT appends utm_source=chatgpt.com to many of the links it cites. Because a UTM source overrides the missing referrer, those sessions still show chatgpt.com as the session source in GA4, which means a single source-based regex can catch both referrer-based and UTM-only visits. That’s the foundation of the setup below.
The complete AI referrer domain list and the copy-paste regex
These are the source values the major AI assistants currently pass to GA4 (verified July 2026):
| Assistant | Source value(s) in GA4 | Notes |
|---|---|---|
| ChatGPT | chatgpt.com, chat.openai.com | Also arrives as utm_source=chatgpt.com with no referrer |
| Perplexity | perplexity.ai | Strong referrer passing; rarely lands in Direct |
| Google Gemini | gemini.google.com, bard.google.com | Legacy bard.google.com still appears on old links |
| Microsoft Copilot | copilot.microsoft.com | Some Copilot clicks credit bing.com instead |
| Claude | claude.ai | Small volume but high engagement in my client data |
| Meta AI | meta.ai | Low volume for most local businesses |
| Grok | grok.com | Growing since standalone app launch |
| DeepSeek | deepseek.com | Mostly technical audiences |
| You.com | you.com | Niche but persistent |
And here is the single regex that covers all of them. Copy it exactly: GA4’s “matches regex” is what we’ll use, so the escaped dots matter:
chatgpt\.com|chat\.openai\.com|perplexity\.ai|gemini\.google\.com|bard\.google\.com|copilot\.microsoft\.com|claude\.ai|meta\.ai|grok\.com|deepseek\.com|you\.com
I revisit this list whenever a new assistant starts sending measurable clicks. If you’re reading this after a major new AI product launch, add its domain with a pipe (|newdomain\.com) and you’re current again.
Step-by-step: build the “AI Referrals” custom channel group
Total time: about 10 minutes. You need Editor or Administrator access to the GA4 property.
- In GA4, go to Admin (gear icon, bottom left) → under Data display, click Channel groups.
- Click Create new channel group. Don’t edit the default group directly; a copy keeps the standard definitions intact for comparison. Name it something like Channels + AI.
- Click Add new channel. Name it AI Referrals.
- Set the condition to: Source → matches regex → paste the regex above. That’s the only condition you need.
- Click Apply, then (this is the step most people miss) drag the AI Referrals channel above Referral in the channel list. I drag it to the very top. GA4 evaluates channel rules in order, and if Referral fires first, your AI sessions stay buried exactly where they were.
- Click Save group.
To view it: open Reports → Acquisition → Traffic acquisition, click the primary dimension dropdown (it says “Session default channel group”), and switch to your new group’s Session channel dimension. You’ll now see “AI Referrals” as its own row with sessions, engagement rate, and key events alongside every other channel.
One critical caveat: custom channel groups are not retroactive. GA4 assigns channel membership at processing time, so your AI Referrals channel only populates from the day you create it. For history, use the next section.
Alternative: an Exploration report for historical data
Because the channel group can’t backfill, I build a quick Exploration to answer “how long has this been happening?”
- Go to Explore → Blank.
- Import the dimension Session source / medium and the metrics Sessions, Engaged sessions, and Key events.
- Drag Session source / medium into Rows and the metrics into Values.
- Add a filter: Session source / medium → matches regex → paste the same regex.
- Set the date range back as far as your property allows (14 months on standard retention settings).
This works retroactively because you’re filtering raw source data rather than relying on channel assignment. Every med spa property I’ve run this on had ChatGPT sessions going back months before anyone noticed. If you want the broader measurement context first, my med spa GA4 tracking guide covers the campaign and key-event setup this builds on.
Reading the numbers: what “normal” looks like in 2026
Set expectations before you open the report. Conductor’s 2026 cross-industry analysis puts AI referral traffic at roughly 1.08% of total website traffic on average, with ChatGPT driving about 87.4% of that AI slice. Previsible’s July 2026 report, based on 6.77 million LLM-driven sessions, measured ChatGPT’s share even higher at 92.4%, while SE Ranking’s 2026 study found a more distributed split: ChatGPT 74.8%, Gemini 11.6%, Perplexity 7.2%, Copilot 3.5%, Claude 2.6%. Methodologies differ, but the shape is consistent: AI is a small-but-growing single-digit share, and ChatGPT dominates it.
For a local med spa, I’d call anything from a handful of sessions to ~2% of traffic normal right now (est.; local-business data is thinner than the B2B datasets these studies skew toward). What matters more is quality: Demand Local’s 2026 statistics roundup reports ChatGPT referral traffic converting at 7.1%, second only to paid search at 7.8%. That matches what I see with clinics: someone who asked an assistant “best Botox provider near me who’s actually medical-director supervised” and clicked through is much closer to booking than a cold Instagram scroller. Getting recommended in those answers in the first place is the job of answer engine optimization, and it’s becoming a standard line item in the med spa marketing plans I build.
The dark-traffic caveat: what GA4 will never show you
Even a perfect channel group only sees people who click through with a measurable source. Three things stay invisible:
- Answer-bot fetches. When ChatGPT or Perplexity reads your pricing page to compose an answer, that’s a server request from GPTBot, OAI-SearchBot, PerplexityBot, or ClaudeBot; no JavaScript runs, so GA4 records nothing. The user may act on your information without ever visiting.
- In-chat answers. If the assistant fully answers the question (“they charge around $12/unit and open Saturdays”), there’s no click at all.
- Residual no-referrer clicks. Visits with no referrer and no UTM still land in Direct, uncounted by any regex.
The first gap is measurable, just not in GA4. Your server access logs record every bot fetch, and I built a free AI bot log analyzer that parses a raw log export and shows you exactly which AI crawlers are reading which pages, how often. Pair it with the GA4 channel and you see both halves: what the machines read, and which humans clicked.
The monthly AI-visibility dashboard I build for clients
For clinics on my generative engine optimization retainers, the monthly report has five rows; you can replicate it in an hour:
- AI Referrals sessions (from the channel group), month over month, split by source.
- Key events from AI Referrals (bookings, calls, form fills) versus the site average. I tie this to revenue with the same math as my missed-call calculator: sessions × conversion rate × average treatment value.
- Bot fetch volume from server logs, a leading indicator; fetches rise weeks before clicks do.
- GSC Generative AI impressions (see FAQ below) for the Google side of the picture.
- Prompt spot-checks: running the clinic’s money queries through ChatGPT and Perplexity and logging who gets recommended. My GEO prompt generator builds that query list from your services and city.
Trend beats snapshot here. A clinic at 40 AI sessions a month isn’t impressive until you notice it was 6 sessions five months ago.
Want a second set of eyes on this for your clinic? Book a free strategy call or call/text me at +91 97297 12388.
Frequently asked questions
Does Google AI Overviews traffic show up in this AI Referrals channel?
No. Clicks from AI Overviews and AI Mode carry a google.com referrer, so GA4 files them under Organic Search with no way to separate them client-side. That’s what Google Search Console’s new report is for; see the next question.
What is the Search Console Generative AI report?
On June 3, 2026, Google announced dedicated Generative AI performance reports in Search Console (per the Google Search Central blog), separating impressions from AI Overviews, AI Mode, and generative features in Discover from classic organic results. At launch it shows impressions, pages, countries, and devices (no clicks, CTR, or queries yet), with Google saying more metrics are coming. It rolled out first to a limited group of UK properties with a global rollout to follow, so don’t panic if your property doesn’t have it yet.
Didn’t GA4 add its own AI channel in 2026?
Google has been rolling out native recognition of some AI assistants in default channel grouping; Digital Applied reported a native AI channel appearing in properties from May 2026. It’s a good start, but it only classifies sessions with intact referrers, coverage of newer assistants lags, and rollout has been uneven. The custom channel group takes 10 minutes, covers UTM-only sessions, and puts the domain list under your control, so I still build it on every property.
Why did my ChatGPT traffic suddenly drop to zero?
Usually it didn’t, the attribution moved. Check Direct traffic for the same period: if it jumped, referrer or UTM behavior changed on ChatGPT’s side (it has changed several times since 2025). Also re-check that your AI Referrals channel still sits above Referral in the channel group order; edits by another admin can reset it.
Should I treat AI traffic differently from organic traffic in my reporting?
Yes, for two reasons. First, intent: AI visitors arrive pre-qualified because the assistant already matched them to you, Demand Local’s 2026 data shows ChatGPT referrals converting near paid-search rates. Second, it’s your early-warning metric for AI visibility. Organic rankings tell you about last year’s playbook; AI referral trend tells you whether assistants recommend you at all.
How often should I update the regex?
Quarterly is enough for most clinics. Skim your Referral channel for unfamiliar AI-looking domains, and add any newcomer that sends real sessions. I update the regex in this article whenever the landscape shifts, so bookmarking this page works too.
