# How to track Copilot referrals in Adobe Analytics: steps > See how to track Copilot referrals in Adobe Analytics using referring domains, a visit segment, channel rules and test clicks, with clear attribution limits. Published: 2026-10-26 Category: Tutorial Reading time: 6 min read ## How do you track Copilot referrals in Adobe Analytics? Inspect Referring Domain and Referrer Type, then build a segment for observed Copilot domains such as copilot.microsoft.com or copilot.com. Optionally add a Marketing Channel processing rule or classification; bing.com is ambiguous, and clicks without a recorded referrer can appear as Direct or Typed/Bookmarked. To track Copilot referrals in Adobe Analytics, inspect Referring Domain and Referrer Type, then build a segment where Referring Domain contains copilot.microsoft.com or copilot.com, checking the hostnames your implementation actually records. Optionally add a Marketing Channel processing rule or classification; bing.com can mix ordinary Bing search with Bing-hosted Copilot, and clicks without a referrer can fall into Direct or Typed/Bookmarked. Updated October. ## Which dimensions show a Copilot referral? Per [Adobe's documentation](https://experienceleague.adobe.com/en/docs/analytics/components/dimensions/referring-domain), Referring Domain derives from the recorded referrer. It is a useful starting point for identifying observed click-through traffic. Check the report suite's internal URL filters because incorrect filters can distort which domains appear. [Referrer Type](https://experienceleague.adobe.com/en/docs/analytics/components/dimensions/referrer-type) groups referrals into Adobe-maintained categories, including Conversational AI tools in its documentation. Inspect that category if present, but keep a domain view beside it: a category label is broader than Copilot and its membership can change. Treat copilot.microsoft.com and copilot.com as candidate hosts to verify, not an exhaustive permanent list. Different Copilot surfaces and redirects may produce different observations. Record the actual destination journey before adding another hostname to a rule. ## How do you inspect a Workspace freeform table? Open Analysis Workspace for the report suite collecting your website traffic. Choose a date range with enough activity to inspect, then create a freeform table with Referring Domain as its rows. Add Visits and your existing conversion or success-event metric where relevant. Search for the candidate domains and inspect the rows returned. Break down a candidate row by the recorded Referrer or landing-page dimension available in your implementation. This helps identify unexpected redirects and pages that receive the clicks. Record the suite, timezone and date range with the table. Keep a separate view of bing.com traffic. Without an additional reliable signal, it is not a Copilot-only cohort. Renaming the entire Bing row “AI referrals” would include ordinary search traffic and make the report look more complete than the evidence supports. ## How do you build the Copilot segment? Use Segment Builder to define a Visits container when the question is what happens during visits with an observed Copilot referral. Add Referring Domain contains copilot.microsoft.com OR Referring Domain contains copilot.com. Name it “Observed Copilot referral visits” and describe the candidate hosts and review date. Use the preview to inspect whether the definition selects the intended records, then save and apply it to the Workspace table. Per [Adobe's segment documentation](https://experienceleague.adobe.com/en/docs/analytics/components/segmentation/segmentation-workflow/seg-build), container scope changes what a segment includes. A Visitors container can bring in other visits and should not silently replace a Visits definition. Review every matched hostname before treating a contains condition as final. A substring can also match an unintended domain. Once you know the observed values, exact host matches or carefully bounded conditions make a production definition easier to audit. ## How can you add a Marketing Channel processing rule? If an administrator wants a shared channel view, review the report suite's Marketing Channels configuration. Adobe documents the route through Admin, Report Suites, Edit Settings and Marketing Channels. Check your admin console labels, which Adobe changes, rather than assuming every account has identical navigation. Create a Copilot-specific rule using the verified referring domains, set the channel and channel detail, and review its position against broader referral rules. Per [Adobe's processing-rule documentation](https://experienceleague.adobe.com/en/docs/analytics/admin/admin-tools/manage-report-suites/edit-report-suite/marketing-channels/mc-proc-rules), conditions determine those assignments. Confirm the rule's behavior on new collected hits before using it in reporting. Retain the segment for checking the underlying domains. Do not assume a channel-rule change rewrites past data. Record the effective date so a reporting change is not mistaken for a sudden increase in Copilot visits. ## When would a classification be useful instead? If your implementation already stores referring-host values in a classification-enabled reporting variable, a classification can group those existing values under a maintained label. Check which variable is available and who owns its lookup before choosing this approach. Do not assume the standard Referring Domain dimension is directly classifiable. Per [Adobe's classification documentation](https://experienceleague.adobe.com/en/docs/analytics/components/classifications/classifications-overview), classifications categorize variable values and can be used with supported custom dimensions. Confirm that prerequisite before promising a classification-based report. Keep a simple mapping record: observed hostname, assigned group, reason and review date. Leave ambiguous hosts unassigned or label them as mixed. A classification reorganizes recorded evidence; it cannot discover a missing referrer or separate ordinary Bing clicks from Copilot without another signal. For a one-off investigation, the saved segment may be sufficient. Add shared configuration when its reporting purpose is clear, rather than building several competing definitions of the same traffic cohort. ## How do you verify the setup with a test click? Find or obtain a genuine Copilot answer linking to a page you control. Record the surface, browser or app, timestamp and destination. Follow the link normally with the site's expected consent settings, then inspect the landing analytics request using your usual browser debugging tools. Check the referrer received, the report suite receiving the hit, and the destination after redirects. After your normal processing delay, inspect the corresponding Workspace result and, if configured, the Marketing Channel assignment. Repeat for another surface when your buyers use both web and app experiences. Do not manufacture a referring host to demonstrate that Copilot sends it. A synthetic analytics hit can test a rule, but only a real click tests the actual referral journey. Keep those two checks labeled separately. ## What remains outside this referral report? An app may open your page without a referrer. Adobe documents Typed/Bookmarked when referrer data is absent; a configured channel may call that Direct. Existing persistence and channel rules can affect the final assignment, so inspect your implementation before interpreting a label. Report the cohort as observed Copilot referrals, with ambiguous Bing and missing-source traffic separate. SearchSeal saves answer evidence rather than human referral traffic. Use Adobe Analytics for visits and outcomes, server logs for request diagnostics, and saved answers for the source links an assistant returned. The [AI visibility measurement guide](/blog/how-to-measure-ai-visibility) explains why those observations should remain distinct. ## Frequently asked questions ### Which domains should a Copilot referral segment include? Start by checking copilot.microsoft.com and copilot.com, then verify the hosts your own click journeys record. The candidates are not a complete or permanent list, and production matching should exclude unrelated hosts. ### Can I classify all bing.com visits as Copilot? No. Bing referrals can include ordinary search and Bing-hosted Copilot. Keep them as mixed unless a separate reliable signal identifies the Copilot journey. ### Why did my Copilot test click appear as Direct? The app or redirect may have supplied no usable referrer. Adobe can show Typed/Bookmarked for missing referrer data, while Marketing Channel labels and persistence depend on your setup. ### Do I need a Marketing Channel rule to inspect Copilot visits? No. A saved domain-based segment and a Workspace table can support an initial review. Use a shared channel rule when you need a maintained channel view and can verify the assignments. ## Verified sources - [Adobe Analytics Referring Domain dimension](https://experienceleague.adobe.com/en/docs/analytics/components/dimensions/referring-domain) - [Adobe Analytics Referrer Type dimension](https://experienceleague.adobe.com/en/docs/analytics/components/dimensions/referrer-type) - [Adobe Analytics Segment Builder](https://experienceleague.adobe.com/en/docs/analytics/components/segmentation/segmentation-workflow/seg-build) - [Adobe Marketing Channel processing rules](https://experienceleague.adobe.com/en/docs/analytics/admin/admin-tools/manage-report-suites/edit-report-suite/marketing-channels/mc-proc-rules) - [Adobe Analytics classification capabilities](https://experienceleague.adobe.com/en/docs/analytics/components/classifications/classifications-overview)