By Ayan Chakraborty, Founder, SBA Innovation Pvt Ltd, Kolkata. Last updated . Reviewed by Ayan Chakraborty.
How do I measure applications that come from ChatGPT referrals?
Create a custom channel group in GA4 that matches assistant referrers, then report sessions and key events by that group. In Admin, open Data display, then Channel groups, copy the default group and add a channel named AI assistants with a rule on source matching the pattern below. Place it above Referral so the traffic is not swallowed by the default channel.
Make sure your application start and application completion events are marked as key events, so the channel report shows applications, not only visits. Reconcile monthly against the source field in your loan origination system, because GA4 sees only what the browser reports. Assistant traffic that arrives without a referrer lands in Direct, so the figure you report is a floor, not the real total; say so on the dashboard.
chatgpt\.com|chat\.openai\.com|perplexity\.ai|gemini\.google\.com|claude\.ai|copilot\.microsoft\.comHow do I report AI-referred applications reliably?
Build one report from the AI assistants channel with sessions, application starts and completed applications, and keep the referrer list identical between baseline and current periods, so changes reflect performance rather than a longer list. Add a landing page dimension to see which guides draw assistant traffic, and link that to your prompt-run log to see which citations produce visits.
Do not model the missing traffic to flatter the number. State the limit plainly: sessions are undercounted by design, and application starts, not approvals, are what you can measure from this channel. Compare approval rates for the cohort against the site average only when the sample is large enough to be meaningful, and avoid presenting a rate from a small sample as durable.
- Create the AI assistants channel group and place it above Referral.
- Mark application start and completion as key events.
- Report sessions and key events by channel and landing page.
- Reconcile monthly against the LOS source field.
- Keep the referrer list unchanged between comparison periods.
- State the undercount on the report.
From our client work: AI-referred sessions and applications
For a consumer lending NBFC, AI-referred sessions rose from 610 to 4,180 a month and application starts from AI referrals from 34 to 386 over eight months, measured with a maintained referrer and UTM list.
Method: Engagement 031: GA4 sessions from a maintained referrer list; application starts are the first funnel event after the eligibility screen, deduplicated by device and session, and reconciled against the LOS source field.
What did not work: Assistants that strip the referrer are invisible in GA4, so the true figure is higher than reported; we chose the conservative measurement rather than modelling the gap.
Read case study 031 Prompt set, baseline, assumptions and footnotes.