What is UTM, and why do ChatGPT links add utm_source=chatgpt.com?

What is UTM, and why do marketers need it?

Suppose you run an online store and are launching a Singles' Day (11.11) sale.

You post the campaign link on Facebook, on Instagram and in your newsletter. All three channels point to the same product page. When the sale ends, you have a problem.

The campaign brought in 3,000 visitors, and 120 of them bought something. But did the Facebook post drive the most orders? Did Instagram do better? How much traffic came from the newsletter?

Without a record of the source, you can only estimate performance from the click counts each platform reports.

UTM exists to solve exactly this.

UTM is a set of parameters appended to a URL. Analytics tools read them to identify which channel, which campaign and even which ad brought a visitor to the site.

For example, links to the same product page can carry different parameters for each channel:

Put together, the three full links are:

# Facebook post
https://example.com/product?utm_source=facebook&utm_medium=social&utm_campaign=double11

# Instagram post
https://example.com/product?utm_source=instagram&utm_medium=social&utm_campaign=double11

# Newsletter
https://example.com/product?utm_source=newsletter&utm_medium=email&utm_campaign=double11

All three point to the same page; only the UTM parameters differ.

When people arrive through these links and the site has Google Analytics 4 (GA4) or a similar tool set up correctly, the parameters are recorded, so marketers can compare traffic and conversions by source.

However, UTM is not a tracking system in itself. It does not collect browsing history, and adding parameters to a URL does not send any data to Google.

UTM supplies the source label. Whether sessions, user behavior and conversion events get recorded depends on the analytics tool and tracking setup the site actually uses.

The five common UTM parameters

With that example in mind, here is how a UTM-tagged link is built.

A complete campaign link might look like this:

https://example.com/product?utm_source=instagram&utm_medium=social&utm_campaign=double11&utm_content=story_a

The question mark ? starts the URL's query parameters, and & separates one parameter from the next. If the original URL already has query parameters, add the UTM parameters with &.

These are the five most common UTM parameters:

In practice, the first three matter most. Google also recommends setting utm_source, utm_medium and utm_campaign whenever you build your own campaign URLs, so the data is classified completely.

The other two can be added as needed.

For example, you might publish two versions of an Instagram story and want to know which one drew more clicks.

Put story_a and story_b in utm_content, and the two versions can be compared separately.

How do source and medium differ?

The pair beginners mix up most often is utm_source and utm_medium.

Think of source as the platform the visitor came from, and medium as the type of channel the visit belongs to.

One detail is easy to miss: GA4 does not classify every utm_medium value correctly.

If the same kind of traffic is sometimes tagged social, sometimes organic_social and sometimes Social, the reports split it apart, or the values stop matching GA4's default channel rules.

That is why UTM naming conventions usually say: use lowercase English and fix one name for each source and medium.

If you have no convention yet, start from Google's URL builder and UTM best-practice guide instead of inventing new rules for every campaign.

Why does ChatGPT add UTM to links automatically?

Now back to the example at the top.

https://example.com/article?utm_source=chatgpt.com

It is easy to read.

utm_source is the traffic source, and chatgpt.com is the label ChatGPT uses for itself.

When someone reaches a site through a search result or citation link in ChatGPT, the site can use this parameter to recognize the click as coming from ChatGPT.

Site owners do not have to apply to OpenAI for this. ChatGPT adds the source information to some outbound links on its own.

According to Webflow's write-up of ChatGPT traffic attribution, OpenAI began adding UTM tags to citation links in ChatGPT answers in June 2026.

That is genuinely useful to site owners.

Before, an article indexed by Google would show up on a search results page, people would click through, and the site could watch Google organic traffic in GA4.

That pattern is changing.

Someone might first ask ChatGPT which GEO tools are worth watching in 2026, read the AI's summary of each product's strengths and weaknesses, and only then open one company's website.

For that company, the visitor still found it while searching for information online, but the click came from ChatGPT's conversation, not from a traditional results page.

That is what makes utm_source=chatgpt.com valuable.

It lets site owners separate ChatGPT visits instead of leaving them mixed into search, social or direct traffic.

One thing needs to be clear, though: ChatGPT adding UTM does not mean every link clicked in ChatGPT carries the parameter.

The interface, how a link is displayed and the platform's own updates can all change what source information is passed along.

A tracking failure in October 2026 made that very visible.

When ChatGPT's UTM tags disappeared, traffic looked like it collapsed

In early October 2026, some site owners saw traffic from ChatGPT drop abnormally.

Oddly, total site traffic did not fall by the same amount. Instead, visits labeled Direct went up.

Later analysis suggested the problem was not that people stopped clicking through from ChatGPT, but that some of ChatGPT's outbound links had lost the utm_source=chatgpt.com tag.

Adobe's Cedric Huesler compared about 500 pages that had meaningful ChatGPT traffic and saw identifiable ChatGPT visits fall by about 31,000 a day, while Direct visits rose by about 26,000 a day.

Profound, an AI search analytics company, also tracked the citation links in ChatGPT answers.

According to Huesler, OpenAI confirmed to him that this was a known bug, not a deliberate removal of UTM tags. Search Engine Land later reported the episode.

What makes the episode worth noting is a problem web analytics often overlooks. When a traffic source cannot be identified correctly, a drop in the report does not necessarily mean fewer real visitors.

No source tag does not mean no traffic

Besides UTM parameters, analytics tools can use the HTTP Referer header to tell which page a visitor came from.

For example, when someone clicks a link on site A to reach site B, the browser may pass site A's address or origin to site B.

That information is usually called the referrer.

But the referrer is not always available. Browser privacy settings, how a link is opened, a site's Referrer Policy and the way some apps handle links can all keep it from being passed on.

In the ChatGPT incident, some clicks lost the UTM parameter and also came without a usable referrer.

When an analytics tool cannot identify where a visitor came from, it may file the visit under Direct.

So Direct does not only mean someone typed your address. It can also include visits whose source information was lost and could not be attributed.

If a site sees ChatGPT traffic drop sharply over a short period, it should not conclude that its GEO strategy has failed, and it certainly should not infer from a smaller number that AI has stopped recommending its content.

How to find ChatGPT traffic in GA4

Once you know what the source tag does, the next question is whether you can see these ChatGPT visitors in your own analytics.

If GA4 is set up correctly and the source information was recorded, the reports will show it.

1. In GA4, open Reports and go to Acquisition > Traffic acquisition.
2. Switch the primary dimension to Session source / medium.
3. Search for chatgpt.com to see the matching rows.

To look at what those visitors did, combine this with landing page, sessions, engagement and key events.

For example, imagine your site has two articles:

That suggests the two articles may be worth different things to the site. The first may suit readers meeting GEO for the first time; the second may reach people already comparing tools or ready to buy.

To know whether the difference is stable, you still need to consider sample size, who the users are and how it changes over time, rather than comparing a single conversion rate.

GA4 already has an AI Assistant channel

You no longer have to rely entirely on your own source filters to spot AI traffic.

GA4 now includes an AI Assistant default channel group that collects traffic from AI assistants such as ChatGPT, Gemini, DeepSeek, Copilot and Grok.

That means Google now treats AI assistants as a channel worth watching on its own, instead of mixing them in with ordinary referral links from other sites.

Still, AI Assistant and UTM work at different levels.

UTM gives GA4 something to identify the source with, but a URL carrying utm_source=chatgpt.com does not guarantee that every GA4 dimension shows the same result.

If the source tag or other referrer data is not passed on, GA4 may not put the visit in AI Assistant.

Google also currently counts traffic from AI Overviews and AI Mode as Organic Search, not AI Assistant. So the AI Assistant figure in GA4 is not the total traffic from all AI search.

For more on how the channel works, see LS-Labs’ earlier guide, Track AI referral traffic in GA4: the AI Assistant channel.

UTM can track AI traffic, but not why AI recommended you

So far, UTM helps a site recognize traffic that comes from ChatGPT.

From a GEO (generative engine optimization) point of view, there is another question.

If you know someone clicked through from ChatGPT, do you know how your site performs in AI search?

Not really.

Suppose 100 people ask ChatGPT which SEO tools are good. The answers to 30 of them mention your product, and 10 end up opening your site.

Your analytics may record some visits from ChatGPT, but that does not let you reconstruct the whole journey from GA4.

That is because an ordinary UTM source tag cannot tell you how many people asked the AI, how often the AI mentioned your brand, where your product appeared, or why someone decided to click after reading.

Now suppose a user reads ChatGPT's description, does not click the citation, opens Google instead, searches for your brand name and only then reaches your website.

That visit may be attributed to Google organic search, not to ChatGPT, which introduced the brand in the first place.

Adding a few more UTM parameters will not fix this. It is a limit of traditional web traffic analysis itself.

Marketers have long judged SEO by search impressions, click-through rate and on-site conversions.

As AI search spreads, people may finish part of their research, product comparison and even an initial purchase decision before they ever reach a website. It resembles zero-click search, a shift LS-Labs also discussed in its AI search retrieval study.

So to understand how a brand performs in AI search, you also need to look beyond site traffic: whether the brand appears in AI answers, which questions trigger a mention, and whether the AI describes the brand accurately. To check how your own site shows up in AI answers, you can run a free scan with LS-Labs GeoCheck. For how AI answers describe and caveat brands, see ChatGPT can cite a comparison and still question its claims.

This information does not conflict with UTM tracking; the two complement each other.

For site owners, a more complete measurement looks at brand visibility in AI answers alongside what visitors actually do and convert on the site.

The first shows how AI presents the brand; the second shows whether that traffic brings real value.

How site owners can start using UTM

Back in everyday digital marketing, UTM is simpler than it looks and needs no programming.

If you run social accounts, buy ads or send a newsletter, start here.

1. Tag the external campaign links you control

For example, your Instagram bio link, Facebook campaign posts, article links in your newsletter and campaign entry points that partners share.

To keep naming consistent, generate links with the Google Campaign URL Builder.

2. Use one naming convention per channel

Do not use facebook one time and fb the next, and do not mix campaign names that differ only in capitalization in the same report.

If you want to compare organic social posts with paid ads, use medium values that clearly separate the two, and check that they match GA4's classification rules.

3. Do not add UTM to your own internal links

UTM is for traffic arriving from external channels. Adding campaign tags to internal links in your site navigation can overwrite or disturb the original attribution and make the analysis meaningless.

Measure clicks on internal buttons, navigation menus and on-site recommendation modules with GA4 event tracking instead.

4. Review conversions by source regularly

The point of UTM is not more rows in a report; it is knowing which channels deserve continued investment.

Rather than comparing traffic volume alone, look at which visitors read more, sign up, submit forms or complete a purchase. To turn GA4 and Search Console numbers into a regular weekly report, you can also try LS-Labs SEO Pulse.

5. Check outbound parameters and tracking regularly

Sites that depend heavily on AI search traffic should periodically check the actual outbound URLs from ChatGPT and other platforms, along with the matching source data in GA4.

If your site uses redirects, short links, cross-domain jumps or other third-party services, make sure those steps do not strip the UTM parameters.

What UTM shows about traffic in the age of AI search

UTM is not new. Its logic has always been simple: use URL parameters to make it easier to see where visitors come from.

But seeing utm_source=chatgpt.com on ChatGPT's outbound links gives a mechanism familiar from digital ads and social marketing a new meaning.

It shows that the starting point for finding information is spreading from search engines and social platforms to AI conversations.

Marketers now need to look beyond Google rankings and understand which steps AI takes part in, from discovering a brand to comparing products and converting.

UTM helps identify some clicks along that path, and GA4 helps explain what people do after they arrive, but neither can reconstruct the full AI search decision on its own.

The October 2026 ChatGPT tracking failure is also a reminder that when traffic sources rely more and more on labels supplied by platforms, a change in a report is not necessarily a change in user behavior.

Learning UTM is only the first step. What matters more is understanding how these numbers are produced, what they can prove, and which user behavior current analytics cannot observe directly.

Sources

• Google Analytics: URL builders and UTM best practices
• Google Analytics: Traffic-source dimensions and manual tagging
• Google Analytics: Default channel group
• Profound: ChatGPT's Missing UTM Tags, October 8–16
• Webflow: Why Your AI-Attributed Traffic May Have Dropped, updated 16 October 2026.
• Search Engine Land: ChatGPT Drop in UTM Tracking Was a Bug