Why ChatGPT ad clicks show up as Direct in GA4, and how to fix it

Ads Manager says your ChatGPT ads got clicks, but Google Analytics files the visits under Direct. Or you can’t tell them from organic visits that arrive from ChatGPT. Your analytics isn’t broken. GA4 calls a visit Direct when it can’t tell where it came from. The fix is to tag every ad, so each paid click carries its own source.

Figures about OpenAI are OpenAI’s. How GA4 behaves is from Google’s documentation.

In brief

• Add your own UTM parameters to every ad’s landing page URL.
• Use one convention everywhere, such as utm_source=chatgpt and utm_medium=cpc.
• Create a custom GA4 channel for it, and order it above Paid Other.
• Expect Ads Manager clicks and GA4 sessions to differ even when tagging is perfect.

Quick diagnosis

Match what you see in GA4 to the likely cause before changing anything.

What you see	Likely cause	Fix
Ad visits under Direct	The landing page URLs carry no UTMs	Tag every ad
Ad visits under Paid Other	The default channel catches cpc	Add a custom channel above it
Some tagged visits missing	A redirect or http hop dropped the parameters	Fix the redirect
Fewer sessions than clicks	Consent, blockers or slow pages	Expected. Compare ranges
Conversions differ	Different windows and models	Reconcile for the same dates

How GA4 decides a visit is Direct

In GA4’s default channel group, a visit is Direct when its source is “(direct)” and its medium is “(not set)” or “(none)”. Google describes that as traffic with no clear referral source.

Google’s documentation gives three groups of causes. First, missing information. That covers links with no UTM parameters and sites not integrated with the ad platform. It also covers redirects or URL shorteners that strip referral details. Second, real direct access, such as typed addresses and links in offline documents like PDFs. Third, ad blockers that interfere with the cookies used to identify where traffic came from.

Why ChatGPT ad clicks can end up there

Four things can combine to send a paid ChatGPT click to Direct.

OpenAI’s click reference isn’t for GA4

OpenAI’s Measurement Pixel reads a value called oppref from the landing page URL. OpenAI uses it to connect later conversions to the ad click. It’s a privacy-preserving identifier for OpenAI’s own measurement, not a source that GA4 reports.

The referrer isn’t documented

Without UTMs, GA4 falls back on the referrer. OpenAI doesn’t say what referrer an ad click passes, so tagging is the only way to be sure. We don’t assume it.

Paid and organic look alike

GA4’s default groups include an AI Assistant channel. Google assigns it when the medium is set to ai-assistant, which happens if the referrer matches Google’s list of AI assistants. Without UTMs, a paid click and an organic visit from the same referrer can’t be told apart.

Organic visits are partly tagged for you. OpenAI says ChatGPT adds utm_source=chatgpt.com to referral links from its search results, for sites that allow OAI-SearchBot. That’s useful, because organic ChatGPT traffic then has its own source. It also means your paid tags should use a different source value, such as chatgpt, so the two stay in separate rows.

Your own site can lose the tags

A redirect to a different URL can drop the parameters. So can a switch from https to http, and so can a consent tool that holds back analytics until a visitor agrees. Each of these can turn a tagged click into an untracked one.

The fix: tag every ad

OpenAI gives you two ways to add tracking parameters. You can type static UTMs into each ad’s landing page URL, and OpenAI says they persist on ad clicks. Or you can set landing page query parameters on a campaign, ad group or ad, using your own text or OpenAI’s macros. Ads Manager fills the macros in when the ad is delivered.

Macro	Filled with
{campaign_id}	The campaign’s ID
{ad_group_id}	The ad group’s ID
{ad_id}	The ad’s ID
{ad_account_id}	Your ad account’s ID

The most specific setting wins, in this order: the ad’s URL, then the ad, the ad group and the campaign. A parameter already in the ad’s URL isn’t overwritten by the others. For product-feed campaigns, add parameters through the landing page settings.

A convention to start from

This is our suggestion, set once at campaign level:

utm_source=chatgpt
utm_medium=cpc
utm_campaign={campaign_id}
utm_content={ad_id}

A click then lands on a URL like https://example.com/pricing/?utm_source=chatgpt&utm_medium=cpc, with the IDs filled in. IDs are unique, but they aren’t readable, so keep a note of which ID belongs to which campaign. Or type the campaign names as static text instead.

Whatever you choose, use it everywhere. One stray spelling splits your reports into two rows, and nobody remembers which one is which.

Give ChatGPT Ads its own channel

With a medium such as cpc, GA4’s default rules treat the visit as paid. Its Paid Other channel matches mediums such as cpc, ppc and paid. For a row of your own, create a custom channel group.

1. In GA4, go to Admin, then Channel groups, and create a new group. Starting from a copy of the default group keeps all the standard channels.
2. Add a channel called ChatGPT Ads, with the conditions source exactly matches chatgpt and medium exactly matches cpc.
3. Reorder the channels so ChatGPT Ads comes before Paid Other. GA4 puts traffic in the first channel whose definition it matches.
4. Save the group and pick it in your acquisition reports.

Google says custom channel groups apply to your reports retroactively. Tagged visits from before you created the channel show up in it too. You need the Editor role or above, and a standard property can have two custom groups besides the default.

A caution about the past. A channel group only regroups visits that already carried your tags. Visits recorded as Direct before you tagged the ads stay Direct. Tag first, and compare from the first tagged day.

Test the path before you spend

A tagged URL has to survive the whole journey. Test it before the budget goes in.

• Open your landing page URL with the UTMs added, and check that GA4’s real-time report shows chatgpt / cpc.
• Follow any redirects, and confirm the parameters reach the final page.
• Make sure the landing page is https all the way through.
• Check that your consent banner and tag setup let GA4 record visitors who agree to analytics.
• Confirm OpenAI’s Pixel still fires with the parameters present.

Our tracking guide covers the Pixel and the Conversions API in depth. The ad preview and URL builder can assemble a tagged landing page URL for you.

Don’t expect the numbers to match

Even with perfect tagging, Ads Manager clicks and GA4 sessions won’t agree. OpenAI’s explanation is that clicks measure ad interactions. Analytics sessions depend on page loads, redirects, consent, browser blocking, UTM handling, attribution windows and time zones. Compare the same date range in the same time zone, and use Ads Manager’s CSV exports to check campaign and ad-level activity.

Conversions differ too. Ads Manager credits conversions within its own windows. These are 7, 14 or 30 days after a click, and optionally one day after a view. It uses last-touch attribution. GA4 applies its own model.

If you want one view across channels, OpenAI lists integrations with measurement partners. They include Fospha, Hightouch, INCRMNTAL, LiveRamp, Northbeam, Rockerbox, Triple Whale and WorkMagic. Several ask you to add their own tracking parameters to the campaign’s landing page settings. Set those up deliberately, beside your UTMs.

OpenAI also has a Google Analytics 4 connection in beta, by early access through your account manager. Each ad account connects one GA4 property. It lets you compare attribution between Ads Manager and GA4. Event mapping isn’t available yet, and its event totals cover the whole property, so they aren’t campaign conversions.

We’d keep the UTMs either way, because they decide how GA4 labels each visit.

Questions