Open GA4 and filter your sessions by the source chatgpt.com. Whatever number appears is a floor, not a total. Some of those visitors sit in direct. Some are scattered across five different source labels. None of them have their own channel.
That is a measurement problem, and it makes LLM visibility look smaller than it is. Fix the report first, then judge the channel.
What ChatGPT traffic looks like inside GA4
ChatGPT passes a referrer on most outbound clicks, so GA4 usually records the source as chatgpt.com with the medium set to referral. Many links also carry a utm_source=chatgpt.com parameter, which gives you a second, more reliable signal.
Older sessions may still show chat.openai.com. Traffic from the ChatGPT desktop and mobile apps is less consistent. Some of it arrives with no referrer at all and lands in direct.
Expect three buckets.
Clean referral sessions from chatgpt.com
Tagged sessions carrying the UTM parameter
Untagged visits hiding in direct
You can only separate the first two. Plan for the third by treating your count as a minimum.
Step 1. Check what GA4 already recorded
Go to Reports, then Acquisition, then Traffic acquisition. Change the primary dimension to Session source / medium and search for "chatgpt" and "openai".
Write down the sessions, engagement rate and key events for each label you find. More than one label means your data is fragmented. It is also the reason a custom channel is worth ten minutes of setup.
Step 2. Build a custom channel for ChatGPT
Go to Admin, then Data display, then Channel groups. Copy the default group so you never overwrite it.
Add a new channel and name it ChatGPT. Set the condition to Source matches regex and paste this pattern.
chatgpt\.com|chat\.openai\.com
Drag the new channel above Referral in the order. GA4 assigns each session to the first channel that matches, so position matters. Skip this and the sessions stay in Referral.
Custom channel groups apply to historical data, so your trend line starts on day one.
Step 3. Build one exploration that answers real questions
Open Explore and start a free-form report. Add Session source / medium and Landing page as dimensions. Add Sessions, Engagement rate, Average engagement time and Key events as metrics.
Filter to your ChatGPT channel and save the report.
Now read it in this order. Landing pages first. Engagement second. Conversions last.
If ChatGPT sends people to three comparison pages and your pricing page, you have learned which content the model trusts. Protect those pages. Keep them accurate, fast and current. If it sends people to a 2023 blog post with outdated pricing, you have found a leak.
Step 4. Tag the links you control
You cannot tag what ChatGPT writes. You can tag what you publish. Add UTM parameters to links in your profiles, press pages and any partner content that models tend to quote.
Pick one naming convention and stick to it. Source in lowercase. No spaces. Mixed capitalization splits one source into several rows, and you will spend an afternoon cleaning it up.
What GA4 will never tell you
GA4 counts visits. It does not count mentions.
A buyer who asks ChatGPT for a vendor shortlist and sees your name without clicking creates zero sessions. A buyer who sees a competitor instead creates zero sessions as well. Your report shows the same flat line for both.
That is where LLM brand visibility tracking fills the gap. It runs the prompts your buyers actually type, records whether you appear, in what position, and which sources the answer cites. Place that data beside your GA4 channel and you can see the full path from mention to click to conversion.
A 15-minute weekly check
Compare ChatGPT sessions week over week
Note the top three landing pages
Compare conversion rate against organic
Flag one page to improve
One small fix every week beats a quarterly audit. Models re-weight their sources constantly, and a page that earned clicks in September can stop earning them by November.
Set up the channel before your next report
Create the ChatGPT channel today. It takes ten minutes and gives you a number you can defend in the next leadership meeting.
Then find out why that number is what it is. LLM Search Console shows which prompts mention your brand and which competitors get named instead.
Subscribe to the Substack newsletter and get one tested AI search tactic every Monday, in time for your standup.