How Attributer categorises leads from AI

AI tools like ChatGPT have exploded in popularity over the past few years, with some sources putting ChatGPT’s user base at 800 million.

However, many studies have shown that despite these impressive usage numbers, they pale in comparison to Google. For instance, recent estimates say that Google Search has about 5 billion active users worldwide (approximately 90% of the internet population). We’ve also seen estimates that show Google drives about 300x more traffic to websites than ChatGPT does due to it’s lack of links to external sources.

Still, it’s likely these AI channels will continue to grow, and it’s something we think a lot about here at Attributer. So in this article, we wanted to outline how we’re thinking about leads from AI channels and how Attributer plans to handle them.

Our thinking

When it comes to our channels, we generally try to keep them aligned with Google Analytics. Reason being is we want people to be able to look in Google Analytics and say ‘I got 100 website visitors from Paid Search’. We then want them to be able to look in their CRM (with the Attributer data) and be able to say ‘I got 50 leads from Paid Search, 20 became customers, and they generated $10k in revenue’.

This ultimately gives marketers that full-funnel view they are looking for, but if our channels differed from Google Analytics, this wouldn’t be possible.

So our general philosophy has been to wait and see how Google Analytics approaches it, and so far they haven’t moved to make a ‘AI Referral’ channel (some smaller analytics tools like Piwik have, but they tools aren’t widely used). If/when they do, we will move to create an ‘AI Referral’ channel as well.

So what happens with AI referrals currently?

The world of AI chatbots is an evolving and somewhat complex landscape when it comes to attribution. So here’s a breakdown of the most popular ones:

ChatGPT

If a person is using ChatGPT in the browser, and they click a link and come to your website, the ‘HTTP Referrer’ will be chatgpt.com. This will cause Attributer to categorise it into the Referral channel as there is a referrer, but it doesn’t match the list of referrers for Organic Search or Organic Social. And if Channel = Referral then we write the domain in the Channel Drilldown 1 field. So ultimately, a lead that came from ChatGPT in the browser would look a bit like this:

• Channel = Referral
• Channel Drilldown 1 = chatgpt.com
• Channel Drilldown 2 = www.chatgpt.com
• Channel Drilldown 3 = None

However, when someone comes to your site from using ChatGPT in an app (iOS app, Mac OS X app, etc) then there is no referrer (because the visit started outside the browser). This would typically cause the visit to be categorised as Direct Traffic except that the ChatGPT apps append utm_source=chatgpt to the end of any clicks originating from their apps. We have a catch-all rule that essentially says ‘If UTMs are present but don’t match the rules for other channels like Paid Search or Paid Social, then make Channel = Other Campaigns’. As a result, they will get categorised into the ‘Other Campaigns’ channel. And when a lead is Channel = Other Campaigns then we write the value from the utm_source parameter into the Channel Drilldown 2 field (so it will say chatgpt.com). So ultimately, a lead that comes from a ChatGPT app will look a bit like this:

• Channel = Other campaigns
• Channel Drilldown 1 = chatgpt.com
• Channel Drilldown 2 = No campaign
• Channel Drilldown 3 = No terms

Gemini

If a person is using Gemini in the browser, and they click a link and come to your website, the ‘HTTP Referrer’ will be gemini.google.com. This will cause Attributer to categorise it into the Organic Search channel as the referring domain contains ‘google.com’. So ultimately, a lead that came from Gemini in the browser would look a bit like this:

• Channel = Organic search
• Channel Drilldown 1 = Other
• Channel Drilldown 2 = gemini.google.com
• Channel Drilldown 3 = None

However, when someone comes to your site from using Gemini in an app (iOS app, Android app, etc) then there is no referrer (because the visit started outside the browser). And because there is no Referrer and no UTM parameters in the URL, this causes the visit to be categorised as Direct Traffic. So ultimately, a lead that comes from a Gemini app will look a bit like this:

• Channel = Direct traffic
• Channel Drilldown 1 = None
• Channel Drilldown 2 = None
• Channel Drilldown 3 = None

What will happen in the future

As you can see, attribution from AI chatbots is a little tricky. This isn’t specific to Attributer, though; it affects all analytics tools (including Google Analytics).

Currently, anyone clicking through from an app (Gemini App, Claude, etc) is not going to have a HTTP referrer or any UTM parameters, so they are going to get categorised as Direct Traffic. ChatGPT appending utm_source=chatgpt to their links helps with this, and I expect Gemini will follow suit at some stage (otherwise Google Analytics becomes a lot less valuable to website owners), but for now, it’s impossible to track leads from AI referrals with 100% accuracy.