Convertmax can classify events and sessions as agentic commerce when the request includes explicit attribution hints or when the worker recognizes known AI-driven referrers.
This guide focuses on how to send those hints intentionally so downstream reporting stays clean and explainable.
Normalized fields stored on each event
When agentic attribution is present, Convertmax stores these fields on the event record:
agentic_sourceagentic_confidenceagentic_sessionagentic_protocolagentic_channelagentic_evidence
Typical normalized sources include chatgpt, copilot, google_ai, perplexity, shopify_agentic, acp, ucp, and unknown.
Accepted request shapes
You can provide agentic hints in any of these forms:
- root-level request fields such as
agentic_sourceandagentic_channel data.agentic_*aliases inside the event payload- a nested object at
data.agentic - request headers such as
X-Agentic-SourceandX-Agentic-Protocol
Detection order
Convertmax applies attribution in this order:
- Explicit metadata from your integration, protocol, or commerce channel
- Known AI referrers such as ChatGPT, Copilot, Perplexity, or Gemini
- Lower-confidence heuristics reserved for future use
That means your server-side hints win over referrer inference.
Preferred payload example
{
"event_type": "convert",
"visitor": "visitor_123",
"session_id": "session_abc",
"data": {
"page": "https://example.com/checkout/success",
"agentic": {
"source": "chatgpt",
"confidence": "high",
"session": true,
"protocol": "acp",
"channel": "shopify_agentic",
"evidence": ["protocol:acp", "channel:shopify_agentic"]
}
}
}
Header-based example
curl -X POST "https://event.convertmax.io/v1/track/" \
-H "Authorization: Bearer <api_key>" \
-H "Content-Type: application/json" \
-H "X-Agentic-Source: chatgpt" \
-H "X-Agentic-Confidence: high" \
-H "X-Agentic-Session: true" \
-d '{
"event_type": "convert",
"visitor": "visitor_123",
"session_id": "session_abc",
"data": {
"page": "https://example.com/checkout/success"
}
}'
Protocol and channel behavior
Some values trigger strong attribution automatically:
agentic_protocol=acporagentic_protocol=ucppromotes the event to a high-confidence agentic sessionagentic_channel=shopify_agenticdoes the same- channels containing
agentic,chatgpt,copilot, orperplexityalso contribute evidence
If a valid source is provided, Convertmax defaults the session to agentic and records evidence for that source.
Recommended implementation pattern
- send explicit agentic metadata from the system that knows the order source best
- use
agentic_protocolfor standards likeacporucp - use
agentic_channelfor commerce-platform labels such asshopify_agentic - pass
agentic_evidencewhen you want to preserve why the attribution was assigned
If you already send standard Convertmax events, you do not need a new endpoint. Add these fields to your existing track, server, or pixel requests and the worker will normalize them for storage.