By Steve Merrill, Founder of WRKNG Digital -- June 13, 2026

73% of AI referral traffic shows up as direct traffic in GA4 right now. According to Adobe Analytics (2026), without custom event tracking you're flying completely blind on your fastest-growing traffic source.

These 11 events fix that.

1. ai_referral_session

This is the foundation event. Fire it when the referrer domain matches chatgpt.com, perplexity.ai, gemini.google.com, or copilot.microsoft.com. Everything else on this list depends on it being set correctly first. Without it, you can't segment any downstream behavior by AI source.

2. ai_product_view

Fire this when a product page loads inside an ai_referral_session. It tells you which specific products AI assistants are sending people to see. That data feeds directly into feed improvement -- you'll know which products are already getting AI traction and which ones aren't being recommended at all.

3. ai_add_to_cart

AI-referred shoppers who add to cart are signaling real purchase intent. Track this separately from your standard add_to_cart event so you can measure AI conversion rates in isolation. Stores tracking AI attribution see 2.4x more actionable data for feed decisions than those relying on default GA4 channel groupings.

4. ai_checkout_initiated

When an AI-attributed session reaches checkout, that's your clearest signal of high-intent traffic. Fire this on the checkout initiation step, not order confirmation. It lets you compare AI funnel drop-off against paid and organic channels with clean separation.

5. ai_purchase_completed

This is the revenue attribution event. Fire it on order confirmation for any session that started with an ai_referral_session. Pass order value and product IDs as event parameters. This is what turns AI visibility work from a brand exercise into a measurable revenue line.

6. chatgpt_shopping_click

ChatGPT Shopping carousel clicks carry traceable UTM parameters -- but only when merchants set them up in advance through their product feed. Track this event specifically using those UTM values (utm_source=chatgpt, utm_medium=shopping). Without it, ChatGPT Shopping traffic blends into generic AI referral data and you lose the product-level signal entirely.

7. perplexity_buy_click

Perplexity's Instant Buy referrals behave differently from standard AI assistant clicks. The user intent is further down the funnel -- they've already read a comparison or summary. Fire this separately from your general ai_referral_session event so you can track Perplexity-specific conversion rates. Perplexity's shopping feature is expanding fast and the data will matter.

8. google_ai_mode_click

Google AI Mode traffic and standard organic traffic share the same referrer domain. Without a dedicated event, they're indistinguishable in GA4. GA4 added native AI assistant channel grouping in May 2026, but you still need this event to correctly segment AI Mode vs. traditional search clicks at the session level.

9. agent_session_identified

This one's forward-looking. Fire it when you detect a non-browser user agent -- headless Chrome patterns, known AI agent strings, or anomalous session behavior with no mouse movement. Agentic browsing is already happening at small scale. Tagging those sessions now means you'll have historical data when it becomes a material traffic source.

10. ai_bounce_event

Fire this when an AI-attributed session ends in under 10 seconds. A fast bounce from an AI referral means the product page didn't match what the AI described or recommended. That's a product data problem, not a traffic problem. Tracking these gives you a direct diagnostic for which products need better structured data or clearer descriptions.

11. ai_comparison_exit

This event fires when session path analysis shows a user landing on a product page from an AI source, then navigating directly to a competitor or back to an AI result page. It's an exit with intent to compare -- different from a standard bounce. High rates on specific products tell you where your positioning or product data is losing the AI-recommended comparison.

How We Chose This List

Each event was chosen against one test: does it produce a decision? Not just data -- a decision. Events that tell you what's happening without telling you what to do next didn't make the cut.

The list covers the full AI attribution funnel: source identification, behavioral signals, conversion tracking, and diagnostic events for when something's broken. We also weighted events by what GA4's current AI channel grouping misses by default, since that's where the biggest blind spots sit right now.

FAQ

Q: Does GA4 handle AI attribution automatically now?

Partially. GA4 added native AI assistant channel grouping in May 2026. But it doesn't capture platform-specific events, agent sessions, or behavioral signals like bounce rate and comparison exits. You need both.

Q: How do I fire events only for AI sessions without slowing down the site?

Set the ai_referral_session event first and write it to a session-scoped custom dimension. All downstream events check that dimension instead of re-parsing the referrer on every page. One referrer check per session. That's it.

Q: ChatGPT Shopping UTM parameters -- where do I set those up?

In your product feed, through the Merchant Center or feed management tool you're using. Add utm_source=chatgpt&utm_medium=shopping&utm_campaign=shopping to your product URLs before submitting the feed. Shopify's product feed documentation covers UTM append options. ChatGPT Shopping will pass those through on click.

Q: What do I do with ai_bounce_event data once I have it?

Sort by products with the highest AI bounce rate. Those are your feed improvement priorities. The AI described or recommended something your product page didn't confirm. Fix the description, title, or structured data for those products first.

Q: Is agent_session_identified reliable yet?

Not perfectly. User agent detection has false positives -- some automation tools and crawlers trigger it. Treat it as a directional signal, not a hard count. Tag it now, build the baseline, and calibrate thresholds as you accumulate data over 60-90 days.

If your store isn't tracking AI traffic separately, you're making feed and content decisions without the data that matters most right now. See how WRKNG Digital helps Shopify stores get AI-ready -- start here.