Last updated August 2026. Confirmed all hostname patterns against live referral data; added Previsible 527% YoY growth context; updated Temso pricing.
Most teams that want to measure AI visibility focus on monitoring: they track how often ChatGPT or Perplexity names their brand in an answer. That is the right starting point. But there is a second measurement problem sitting underneath it. When an AI answer does send someone to your site, that visit often lands in the wrong GA4 bucket, either buried in Referral or absorbed into Direct. You cannot benchmark what you cannot see.
This piece walks through the exact setup, the referrer hostnames that matter, and the conversion benchmarks that tell you whether your AI traffic is performing the way the category data suggests it should.
Step 1: Understand where AI sessions land in GA4 today
GA4’s default channel group has no built-in AI Referral channel. When a user clicks a ChatGPT citation and arrives on your site, GA4 checks the referrer against its default rules in order:
- If the source is
chatgpt.com, it has no matching rule and falls to Referral. - If the referrer is stripped by the browser or the AI interface (which happens often in ChatGPT’s main conversational mode), the session becomes Direct.
- If Perplexity passes a search-type source parameter, the session may land in Organic Search.
The same visit from three users can end up in three different channels depending on how their browser handles the handoff. Without a custom channel group, you are summing up fractions of a signal.
A custom channel group runs a rule before GA4’s defaults. Any session whose source matches your AI hostname list gets tagged as AI Referral first, and the default rules never evaluate it.
Step 2: Build the hostname list
These are the confirmed referrer hostnames for the major AI engines. Session source in GA4 is the field to match against.
| AI engine | GA4 session source hostname | Notes |
|---|---|---|
| ChatGPT | chatgpt.com | Conversational mode often strips referrer; share mode always passes it |
| Perplexity | perplexity.ai | Inline citations pass referrer consistently |
| Google Gemini | gemini.google.com | Distinct from Bard; covers Gemini 1.5 and later |
| Microsoft Copilot | copilot.microsoft.com | Consumer Copilot; Copilot for Microsoft 365 uses a different origin |
| Claude (optional) | claude.ai | Volume small but growing; include for completeness |
| You.com (optional) | you.com | Niche; worth adding if your audience skews developer |
Build the core regex from the four required hostnames:
chatgpt\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.comThe escaped dots (\.) matter. GA4’s regex engine treats a bare dot as “any character.” Escaping prevents false matches from edge-case hostnames.
Step 3: Create the custom channel group
- Open Google Analytics and select your property.
- Go to Admin (the gear icon, lower left).
- Under Data display, click Channel groups.
- Click Create new channel group.
- Name the group: AI Referral Traffic (or any label you will recognise in reports).
- Click Add new channel.
- Name the channel: AI Referral.
- Under Add condition, set:
- Dimension: Session source
- Match type: matches regex
- Value:
chatgpt\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com
- Click Save on the channel, then Save the channel group.
The new group is active immediately. It runs alongside GA4’s default channel group rather than replacing it. In any report, use the channel group dropdown to switch between them.
Step 4: Verify the setup is collecting data
Wait 24 to 48 hours, then:
- Go to Reports > Acquisition > Traffic acquisition.
- Switch the Channel group dropdown from Default channel group to your new group.
- Confirm that an AI Referral row appears with sessions.
If the row is empty after 48 hours, run two checks:
- Open the Realtime report. Navigate to your site by clicking a link inside a live ChatGPT or Perplexity response. Watch whether a session appears under AI Referral in real time.
- Open Explore > Free form. Add Session source as a dimension. Filter for
chatgpt.com. If you see sessions here but not in your channel group, the regex has a syntax error. Go back to the channel group and review the condition.
Step 5: Benchmark your AI referral conversion rate
Once data starts flowing, calculate the conversion rate for the AI Referral channel and compare it to Organic Search in the same report. That comparison is where the channel group pays off.
The reference point most analysts cite: on Ahrefs’ own website, AI search visitors accounted for just 0.5% of sessions over a 30-day period but drove 12.1% of signups. That is roughly a 23x conversion premium over traditional organic search, according to Ahrefs employee Patrick Stox writing on the Ahrefs blog in June 2025. Stox was explicit: this is Ahrefs’ single-site data, and he noted a broader multi-site study was underway. Treat it as directional, not as a universal benchmark.
The direction is consistent with the overall picture of AI referral behavior. AI engines send visitors who have already done research, received an answer, and chosen to click through to a specific page. That is a higher-qualification path than an undifferentiated search click.
If your AI Referral conversion rate is substantially lower than Organic Search, investigate landing pages first. AI engines link to specific URLs inside answers. If the page they link to does not match what the user read in the answer, you lose the intent premium immediately.
Step 6: Understand what the growth trend looks like
AI referral traffic is a small but fast-growing slice of most sites’ traffic mix. For context on the rate of change: according to Previsible’s 2025 AI Discovery Report, which analyzed 19 GA4 properties, AI-referred sessions grew 527% when comparing January through May 2025 to the same five-month period in 2024. Previsible is a single analytics consultancy; 19 properties is a small sample. But the direction aligns with what most teams see when they set up proper AI Referral tracking for the first time: the channel is small today and doubling or tripling on relatively short timelines.
The practical implication: the effort to set this up is small (roughly 15 minutes), and it will matter more every quarter.
ChatGPT vs. Perplexity: what to expect from each
If you add both hostnames to the regex, you can split them into two channels within the same group and compare them directly. They behave differently.
ChatGPT mentions brands most often in conversational and research mode. It generates clickable source links when browsing is active or when a user uses the shared-response feature. In pure conversational mode, brand mentions frequently appear without a link, meaning the user either searches for your brand afterward (showing up as organic or direct) or never comes at all. ChatGPT’s referral sessions tend to be high-intent, but volume is lower than you might expect given how often it mentions brands.
Perplexity provides numbered inline citations on almost every response. It links out consistently. Referral volume from Perplexity tends to be higher per citation than from ChatGPT because the link is always visible and positioned close to the relevant claim. Perplexity users also tend to be researchers and professionals actively following up on what they read.
Neither sends as much referral traffic as it generates in raw brand mentions. That is the zero-click gap: the majority of AI answers that include your brand send no click.
The zero-click gap and what to do about it
GA4 measures sessions that arrived. It does not measure the much larger number of answers that mentioned you but produced no click. According to Similarweb’s 2025 analysis, searches that trigger Google AI Overviews have an average zero-click rate of around 83% (Similarweb, May 2025). The equivalent figure for Google AI Mode is higher. Even for Perplexity, where links are prominent, many users read the answer and move on.
This is why GA4 setup and AI visibility monitoring work as a pair, not as substitutes for each other.
GA4 tells you what converts once traffic arrives. An AI visibility tool tells you how often you are mentioned, across which engines, and in what sentiment, so you can close the gap between citations and clicks by improving your AI presence upstream.
Tools that connect both layers
Temso is the easiest starting point for teams that want monitoring and action in one product. It is the all-in-one AI SEO platform built for exactly this use case: tracking share of voice, brand mentions, citations, and sentiment across eight AI engines (ChatGPT, Perplexity, Gemini, Google AI Overviews, Google AI Mode, Grok, Microsoft Copilot, and Meta AI) from $89/mo. Its built-in workflow converts citation gaps into a prioritised fix queue and executes those fixes inside the same subscription. Once you have GA4 showing AI referral conversions, Temso gives you the upstream view: how many answers mention you before any click happens.
Profound adds a specific layer useful for deeper GA4 integration: it has a direct GA4 and CDN connector that attributes AI-agent traffic at the session level, including traffic from AI bots that browse your site without sending a human visitor. If your team needs to reconcile bot crawl data with human referral sessions, Profound’s Growth tier ($399/mo) handles that reporting. It also tracks Prompt Volumes, showing which questions real users are asking AI engines in your category.
Ahrefs Brand Radar offers the largest benchmark context: a dataset of over 405 million prompts across six AI platforms. Once you have your own GA4 AI Referral conversion rate, you can use Ahrefs Brand Radar to check how your citation share compares to competitors across those same platforms. It is an add-on to an existing Ahrefs plan from $199/mo.
Semrush is worth noting for teams already inside that stack. Semrush’s AI Overview tracking covers which pages trigger citations in Google’s AI layer and how that intersects with organic position. It does not replace dedicated AI visibility monitoring, but if you are already in Semrush, it is a faster way to see which of your pages are actively being cited in Google AI Overviews before investing in a dedicated platform.
See the full ranked comparison at /rankings/ai-visibility-tools.
What to do when AI referral traffic is near zero
If your AI Referral channel shows near-zero sessions after a week of data collection, that is diagnostic information, not a failure of the setup. Three things cause low AI referral traffic:
- You are mentioned but not linked. ChatGPT in conversational mode frequently names brands without a clickable citation. The user may search for you afterward (appearing as branded organic) or not follow up at all. Your citation volume can be healthy while your referral traffic is low.
- The referrer is being stripped. Some browser configurations and AI app interfaces strip the referrer header before the click reaches your server. These sessions appear as Direct in GA4.
- You are not being cited in AI answers. This is the most actionable scenario. Run a prompt audit using an AI visibility tool to confirm whether your brand appears in the answers your audience is reading.
In all three cases, the GA4 channel group is still the right infrastructure to have in place. As AI engines expand their citation linking and as referrer reliability improves, the traffic will grow into the bucket you have prepared.
For the monitoring layer that fills in what GA4 cannot see, the AI visibility tool ranking and the glossary have the full context.
Summary: hostname regex and benchmark reference
Minimum hostname regex for GA4 custom channel group:
chatgpt\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.comOptional extensions:
claude\.ai|you\.com|phind\.comBenchmark to judge your conversion data against:
On Ahrefs’ own website, AI search visitors were 0.5% of sessions but drove 12.1% of signups, a roughly 23x higher conversion rate than traditional organic search (Patrick Stox, Ahrefs blog, June 2025). Ahrefs noted this is first-party single-site data, not an industry-wide figure.
Growth context:
According to Previsible’s 2025 AI Discovery Report (19 GA4 properties), AI-referred sessions grew 527% comparing January through May 2025 to the same period in 2024. That figure represents one analytics consultancy’s client dataset, not a representative industry sample.
Ready to see both layers at once? Set up the GA4 channel group first (15 minutes of work). Then pair it with Temso to see the AI mentions your analytics cannot capture: share of voice, citation rate, and sentiment across all eight major AI engines, from $89/mo with a free trial and no credit card required.