ChatGPT Traffic Analysis: How to Measure Visits From ChatGPT

ChatGPT deserves its own analysis because it sends most AI referral traffic. BrightEdge reported that ChatGPT had 95.1% of AI referral traffic in August 2026. SE Ranking's panel of 101,574 sites puts its 2026 share lower, at 74.78%. The panels differ, but both say the same thing: for most sites, ChatGPT traffic is most of the AI story.

This guide covers ChatGPT only. For the channel setup that catches every assistant, read our full guide to measuring AI referral traffic in GA4. If you just want GA4 working today, the 15-minute AI referral setup is the shorter path.

How ChatGPT Traffic Reaches Your Analytics

Three signals can identify a ChatGPT visit. Each one fails in a different way, so it helps to know all three.

1. The referrer. A click from ChatGPT on the web carries chatgpt.com as its referring host. Old chat.openai.com links now answer with a 308 redirect to chatgpt.com, so expect little of that host in new data.
2. The tag. OpenAI's publisher FAQ says ChatGPT adds utm_source=chatgpt.com to its referral URLs. The tag travels in the address, so it survives even when a browser drops the referrer.
3. GA4's channel. Since 13 May 2026, Google's AI Assistant channel sets the medium to ai-assistant when the referrer is on its list of assistants, and Google names ChatGPT among them.

Google's processing rules say UTM values map straight onto source and medium, and that the referrer only fills gaps. Google doesn't document how its ai-assistant rewrite interacts with a source-only tag. So your property may show ChatGPT traffic in more than one row.

The chatgpt.com rows you may find

The paid row follows from Google's definitions: a paid-style medium with a source that isn't on Google's search, social, video or shopping lists falls into Paid Other.

Build a ChatGPT-Only View in GA4

You don't need a custom channel group for this. You need the source.

1. Find every row. Open Reports, then Acquisition, then Traffic acquisition. Set the primary dimension to Session source / medium and type chatgpt in the search box. Write down each row and its sessions for the last 28 days.
2. Start an exploration. In Explore, choose Free form. Add the dimensions Session source / medium, Landing page + query string and Session default channel group. Add the metrics Sessions, Engaged sessions, Engagement rate, Key events and Session key event rate.
3. Add two segments. Create a session segment where Session source exactly matches chatgpt.com, and a second where Session default channel group exactly matches Organic Search. Google's free-form docs allow up to four segments side by side.
4. Check new users. In the User acquisition report, the First user source chatgpt.com shows people whose first visit came from ChatGPT.

Exclude your ad rows before you read any rate. OpenAI's ads basics page says advertisers can add static UTM parameters to landing page URLs, and that they persist on ad clicks. Build those links with our UTM link builder and a medium of cpc, so paid clicks never blend into organic ChatGPT traffic.

The ChatGPT Traffic Scorecard

The ChatGPT traffic scorecard is our five-read monthly check. Each read answers one question, and together they fit on half a page.

Why the landing mix matters now

SE Ranking found that ChatGPT's worldwide referral share rose from 0.23% of all traffic in April 2026 to 0.32% in May, and that homepages drove most of the jump. Practitioners tie it to brand names in answers becoming clickable around 7 May. The same study reports that 60% of AI-referred visits land on homepages, against 17% from organic search. If you judge ChatGPT only by the posts it cites, you miss most of it.

Why quality needs a margin of error

ChatGPT sessions are few, so their conversion rate swings. Put the 95% margin next to the rate every time: 1.96 × √(rate × (1 − rate) ÷ sessions). If the range overlaps your organic search rate, the gap isn't proven yet.

Worked Analysis: Tallyhive's September

Tallyhive is a made-up invoicing app at tallyhive.io. The numbers are illustrative, but the arithmetic is real. In September 2026 it had 80,000 sessions, 42,000 of them from Organic Search.

Step 1: Gather the rows

Organic ChatGPT traffic is 1,260 ÷ 80,000 = 1.6% of sessions. Reading only the AI Assistant channel would have missed 110 of those 1,260 sessions, or 8.7%. The leak rate is the same 8.7%.

Step 2: Read the landing mix

Of the 1,260 organic sessions, 740 landed on the homepage: 58.7%. For Organic Search, the homepage took 9,000 of 42,000 sessions, or 21.4%. The rest of the ChatGPT sessions landed on /pricing (190), comparison pages (160) and the blog (170).

Step 3: Compare quality

ChatGPT traffic converted at 37 ÷ 1,260 = 2.94%. Organic Search converted at 756 ÷ 42,000 = 1.80%. The margin on the ChatGPT rate is ±0.93 points, so the true rate likely sits between 2.0% and 3.9%. Even the low end beats 1.8%, and a two-proportion test gives p ≈ 0.003. On this data, a ChatGPT visit is worth about 1.6 times a search visit for trial starts.

The ads tell a different story: 4 trials from 240 sessions is 1.67%, with a margin of ±1.62 points. That's too few sessions to judge.

Step 4: Find the page that converts

The pricing page converts at more than three times the homepage rate, and that gap holds up (p ≈ 0.001). But ChatGPT sends four times more visits to the homepage. Tallyhive's next move is to put its pricing and a comparison summary on the homepage, where ChatGPT traffic actually lands, and to track the homepage rate next month.

What ChatGPT Traffic Analysis Can't See

Every number above undercounts, in known ways.

• The prompt. Browsers default to sending only the origin to other sites, per MDN. You see chatgpt.com, never the question.
• Clicks with no referrer and no tag. Seer Interactive warns that clicks from in-app browsers and copied links land in Direct, so read the AI channel as a floor. OpenAI doesn't document what its desktop and mobile apps send.
• Journeys across devices. A buyer who reads the answer on a phone and signs up on a laptop looks like two people.
• Mentions without clicks. A buyer who reads your name and searches for it later shows up as branded search. Our GEO measurement guide covers a signup-form question that catches them.

One gap is smaller than it looks. Google's scope docs say sessions that start by direct entry are credited to the campaign values GA4 already holds for that user. ChatGPT's links carry a campaign source, so a buyer who clicks from ChatGPT and later types your URL in the same browser can still count toward chatgpt.com.

Two outside checks help. Hits from ChatGPT-User in your server logs show pages that ChatGPT read for a user, even without a click; our AI crawler log analyzer counts them per URL. And on paid Cloudflare plans, AI Crawl Control's referral charts count chatgpt.com referrals at the edge, before any analytics script runs.

Where Rankbox Fits

Rankbox doesn't analyze your GA4 data or track AI citations today. Everything above runs in GA4 and your logs.

Rankbox works on what earns the click. Answer-Space Research maps the questions buyers ask ChatGPT, and the Citation-Ready Writer writes source-backed articles that answer them, delivered to your site through Rankbox's API. The scorecard above then shows whether those pages bring ChatGPT traffic that converts. See plans and pricing.