Track and analyze AI traffic

Transcript

Hey everyone, I’m Michelle and today we’re covering how to spot, sift out and analyze AI generated traffic and customer journey analytics with LR LMS like ChatGPT driving more web visits. It’s impacting data, think inflated page views, skewed conversions and user journeys. But don’t worry, CJR makes it easy to keep your insights human focused. We’ll cover quick detection tricks, reporting, and segments. Let’s jump in. We’ll start with detection using three signals. No heavy lifting required. First, user agents these are like digital fingerprints in the browser. Headers flag common ones like GPT bot or Cloud Bot. They’re often from eye crawlers. Next refers check where traffic is coming from. Referrers like chatgpt.com or perplexity AI are obvious AI referrals last query parameters look for UTM tagged with AI sources like UTM source equal to Gemini. Layer these for reliability one might miss, but together you’re covered. Now let’s implement these and capture these AI traffic signals and derive fields in your data view. These are your nondestructive classifiers. As a side note, a derived field created in a data view is stored in the underlying connection, and therefore it’s available in any other data view tied to that connection. For the first derived field, use a simple case when formula to get the rule started. We’ll look at the treatment for GPT ba. If the user agent contains this phrase, set the custom value as open AI agent or a value of your preference. Do the same for other AI agents in subsequent conditions. Review the documentation on Experience League for additional examples of detection signatures. Now to the AI refers. We handle this by setting up a new dry field using the URL pass function. For the first rule, the value we use is the web dot web refer dot URL attribute, which maps to the standard referred dimension in the reports. Selecting that value preset displays as URL in the role. Then choosing the Get host option will extract what we need. The second part of this field is classifying the output from the first rule. Uploading a CSV is handy for setting this up again. Review the documentation for additional examples of refer classification values. The last signal comes from the URL parameter UTM source and the referrer. This new drive field is very similar to the previous field, except it uses the Get query string value option for the refer in the URL parse function, and then extracts the UTM source value with the derived fields handled. We move on to building segments to aid in our analysis and reporting. Create a segment for LM and AI generated traffic using your flags, and then use this segment to exclude it from main reports. To purify your human data, add an alert for spikes, like if I had 10% of sessions so that you can stay ahead of the curve. All right, now it’s analysis time. Let’s keep it actionable in your workspace project. Compare AI versus human behaviors. Do bots consume content fast but bounce on carts? Use Fallout’s to see journey drops or cohort analysis for trends over time. Segment by channel two is mobile hit harder. This uncovers opportunities like optimizing for AI friendly summaries and then review quarterly as AI evolves. And as always, partner with your development team for adjustments. And there you have it. Track AI traffic confidently and CJA protect your data and fuel smarter marketing. Thanks for watching. See you next time.