BigQuery Reveals Which AI Agents Buy From Your WooCommerce Store
Quick Answer: GA4’s AI Assistant channel only detects 5 of 12+ AI referrer domains — BigQuery with server-side enrichment sees them all (Seresa, 2026). If your server-side tracking tags agent vs human at the order hook, BigQuery contains the cohort flag that GA4’s interface cannot display. One custom parameter set at the PHP level unlocks the entire agent commerce reporting capability, letting you split revenue, AOV, and conversion rate by human vs agent — and even by which AI platform drove the purchase.
In this article
- Can BigQuery identify AI agent purchases that GA4 cannot see?
- What data does BigQuery need to separate agent from human purchases?
- What can you learn from BigQuery agent purchase data?
- How do you query BigQuery for agent-attributed revenue?
- Can BigQuery identify which AI platform drove each agent purchase?
- How does BigQuery agent data feed into Looker Studio dashboards?
- What is the cost of BigQuery for agent commerce analysis?
- What is the first BigQuery query every WooCommerce store should run?
Can BigQuery identify AI agent purchases that GA4 cannot see?
Yes — BigQuery receives the raw GA4 event stream plus any server-side enrichment, so if your tracking tags agent vs human at the order hook, BigQuery contains the cohort flag that GA4’s interface cannot display. GA4’s AI Assistant channel only detects 5 of 12+ AI referrer domains — BigQuery with server-side enrichment sees them all (Seresa, 2026).
GA4’s standard reporting groups AI traffic into a single “AI Assistants” channel, and it only recognizes a handful of known referrers — ChatGPT, Perplexity, and a few others. Every agent purchase from an unrecognized AI platform lands in “(direct)” or “organic” and disappears into the noise. BigQuery does not have this limitation because it stores the raw event data with every parameter your tracking attached, including custom ones GA4’s interface has no way to surface.
Here’s the thing… the purchases are happening right now. The question is whether you can see them. GA4 was designed for browser sessions — it does not have a mental model for API-initiated purchases. BigQuery has no opinions about what a purchase should look like. It stores every event exactly as it arrived, and if you tagged the origin correctly, the data is already sitting there waiting for a query.
GA4’s AI Assistant channel detects only 5 of 12+ AI referrer domains — BigQuery with server-side enrichment captures all of them, revealing the full agent commerce channel (Seresa, 2026).
What data does BigQuery need to separate agent from human purchases?
BigQuery needs a custom event parameter or user property that your server-side tracking sets at the order hook — a field like purchase_origin with values human or agent — which then appears in the event_params array of every purchase event. One custom parameter set at the PHP level unlocks the entire agent commerce reporting capability in BigQuery (Seresa, 2026).
The parameter itself is simple: when a purchase event fires at the woocommerce_payment_complete hook, your server-side tracking checks whether the order arrived through a browser session or an MCP API call. If it came through MCP, the purchase_origin parameter is set to agent and — if identifiable — a second parameter agent_platform carries the AI platform name (ChatGPT, Claude, Perplexity). These parameters travel with the GA4 Measurement Protocol event and land in BigQuery exactly where you put them.
Without this parameter, you can still approximate agent identification by querying for known AI user agents or referrer domains in BigQuery. But approximation misses purchases from agents that use standard user agents or arrive through API endpoints with no referrer header. The server-side tag gives you ground truth at the source instead of inference after the fact.
Related: Qualified Future Conversions: What WooCommerce Store Owners Need to Change
What can you learn from BigQuery agent purchase data?
BigQuery can show you agent revenue by AI platform, agent vs human conversion rates, agent AOV compared to human AOV, which product categories agents favor, agent purchase timing patterns, and agent session behavioral signatures. Shopify AI orders carry 14% higher AOV — BigQuery lets WooCommerce stores discover their own platform-specific agent commerce metrics (Latency Studio / Shopify, 2026).
That 14% AOV premium is a Shopify-wide average. Your store’s numbers may be higher or lower — and you cannot know until you measure. BigQuery gives you the resolution to answer questions GA4 cannot: do agent purchases have different return rates? Do they cluster around specific products? Do Claude users buy different things than ChatGPT users? Is the agent AOV premium seasonal or consistent?
The commercial implication is significant. If agent purchases carry a materially different AOV or conversion rate, they need a different ROAS target in your bidding strategy. Blending agent and human conversions into one number means your Smart Bidding is optimizing against an average that represents neither cohort accurately — a problem you cannot diagnose until you split the data.
How do you query BigQuery for agent-attributed revenue?
Query the GA4 events export table filtering for event_name = 'purchase', then group by your custom purchase_origin parameter — the result gives you total revenue, transaction count, and average order value split by human vs agent cohort. The query runs against standard GA4 BigQuery export tables — no special schema or additional data pipeline required beyond the server-side tag (Seresa, 2026).
The SQL is straightforward. You unnest the event_params array to extract purchase_origin, then aggregate ecommerce.purchase_revenue by that dimension. A basic version looks like this:
SELECT purchase_origin, COUNT(*) as transactions, SUM(purchase_revenue) as revenue, AVG(purchase_revenue) as aov FROM events, UNNEST(event_params) WHERE event_name = 'purchase' AND key = 'purchase_origin' GROUP BY purchase_origin
This gives you two rows: human and agent. From there, you can layer on time dimensions (daily, weekly, monthly trends), product dimensions (which SKUs agents prefer), and platform dimensions (ChatGPT vs Claude vs Perplexity revenue) — all from the same table, all using the same parameter. The query complexity scales with the question, not with the infrastructure.
Shopify AI orders carry 14% higher AOV — BigQuery lets WooCommerce stores discover their own platform-specific agent commerce metrics instead of relying on cross-platform averages (Latency Studio / Shopify, 2026).
Related: Five GA4 Volume Thresholds Your WooCommerce Store Fails
Can BigQuery identify which AI platform drove each agent purchase?
If your server-side tracking captures the agent platform identity (ChatGPT, Claude, Perplexity, etc.) from MCP request headers or referrer data and passes it as a custom parameter, BigQuery can break down revenue by AI platform. ChatGPT commands 76.85% of AI referral traffic — BigQuery reveals whether that share holds for your store specifically or if other platforms convert better for your products (AuthorityTech / Statcounter, 2026).
The 76.85% number is a traffic share, not a conversion share. Your store may find that ChatGPT drives the most visits but Claude drives the highest-value purchases, or that Perplexity users convert at a higher rate because their search intent is more specific. Without the platform-level breakdown in BigQuery, you are treating all AI agents as one undifferentiated channel — the same mistake marketers made with “social” traffic before they split it into Facebook, Instagram, and TikTok.
The implementation sits in your server-side tracking. When an MCP request arrives, the agent typically identifies itself in the request headers or through OAuth metadata. Your tracking extracts that identifier, normalizes it to a platform name, and sends it as the agent_platform custom parameter on the purchase event. BigQuery stores it; your query groups by it.
How does BigQuery agent data feed into Looker Studio dashboards?
BigQuery tables connect directly to Looker Studio as a data source — create a dashboard with filters for purchase_origin that let clients toggle between human performance, agent performance, and blended views. Looker Studio + BigQuery + server-side cohort tags = the complete agent commerce reporting stack for WooCommerce (Seresa, 2026).
The dashboard setup mirrors any BigQuery-to-Looker Studio workflow: create a custom SQL data source in Looker Studio that unnests the purchase_origin and agent_platform parameters, then build scorecards (agent revenue, agent AOV, agent transaction count), time series (agent revenue trend vs human revenue trend), and comparison tables (by platform, by product category).
The value of the dashboard over ad-hoc queries is the audience. A SQL query answers one analyst’s question. A Looker Studio dashboard answers the store owner’s standing question — “how much of my revenue is coming from AI agents, and is it growing?” — without requiring anyone to write SQL. The dashboard refreshes automatically, and the store owner checks it the same way they check any other performance report. Transmute Engine feeds the cohort tags into the GA4 Measurement Protocol event at the PHP hook, so the data arrives in BigQuery already tagged and ready for the dashboard — no manual tagging, no ETL layer, no data warehouse configuration beyond the standard GA4 export.
What is the cost of BigQuery for agent commerce analysis?
GA4’s BigQuery export is free for up to 1 million events per day on the free tier, and the queries needed for agent commerce analysis typically scan small amounts of data — most WooCommerce stores will spend under $5/month on agent-specific BigQuery analysis. 1 million events/day free tier covers most WooCommerce stores — agent commerce reporting adds negligible cost (Seresa, 2026).
The cost model works in your favor for two reasons. First, the BigQuery export itself is free — Google does not charge for writing GA4 data into BigQuery. Second, BigQuery charges by data scanned on queries, and a query that filters for purchase events and groups by one custom parameter scans a tiny fraction of your total event volume. Even stores with hundreds of thousands of monthly sessions will find that their agent commerce queries cost pennies.
The more relevant cost is the server-side tracking that tags the events correctly in the first place. Without the purchase_origin parameter, you can still run approximate queries against referrer data, but the accuracy drops and the query complexity rises. The tagging investment is a one-time setup that unlocks a permanent reporting capability — the ongoing cost is effectively zero once the data flows.
What is the first BigQuery query every WooCommerce store should run?
Count purchase events by referrer source domain where the source matches known AI platforms — this gives you an immediate baseline of how much AI-attributed revenue you already have, without any server-side tracking changes. The baseline query uses only standard GA4 BigQuery export data — you can run it today before implementing any new tracking (Seresa, 2026).
The baseline query filters purchase events where traffic_source.source contains known AI domains: chatgpt.com, perplexity.ai, claude.ai, gemini.google.com, copilot.microsoft.com, and others. It will miss agent purchases that arrived through API calls (which have no referrer), but it gives you a floor — the minimum amount of AI-driven revenue you are already earning.
If the floor is non-trivial, you have a business case for implementing the server-side purchase_origin tag. If it is near zero, either agent commerce has not reached your category yet or — more likely — the agents are arriving through API channels the referrer-based query cannot see. Either way, you need the server-side tag to know for certain. The baseline query takes 10 minutes to write and run. The answer tells you whether agent commerce is already in your revenue mix or still arriving.
Key Takeaways
- GA4 misses most agent purchases: its AI Assistant channel only detects 5 of 12+ AI referrer domains — BigQuery with server-side enrichment catches them all.
- One custom parameter unlocks everything: a
purchase_originfield set at the PHP hook lets BigQuery split all commerce metrics by human vs agent. - Agent AOV may differ significantly: Shopify AI orders carry 14% higher AOV — BigQuery lets you discover your store’s own numbers instead of relying on cross-platform averages.
- Platform-level breakdown matters: ChatGPT commands 76.85% of AI traffic but other platforms may convert better for your specific products.
- Cost is near zero: the free BigQuery tier covers most WooCommerce stores, and agent commerce queries scan negligible data.
- Run the baseline query today: you can count AI-referred purchases in your existing BigQuery data right now, before any tracking changes.
Yes — BigQuery receives the raw GA4 event stream plus any server-side enrichment, so if your tracking tags agent vs human at the order hook, BigQuery contains the cohort flag that GA4’s interface cannot display.
BigQuery needs a custom event parameter or user property that your server-side tracking sets at the order hook — a field like ‘purchase_origin’ with values ‘human’ or ‘agent’ — which then appears in the event_params array of every purchase event.
BigQuery can show you agent revenue by AI platform, agent vs human conversion rates, agent AOV compared to human AOV, which product categories agents favor, agent purchase timing patterns, and agent session behavioral signatures.
Query the GA4 events export table filtering for event_name = ‘purchase’, then group by your custom purchase_origin parameter — the result gives you total revenue, transaction count, and average order value split by human vs agent cohort.
If your server-side tracking captures the agent platform identity from MCP request headers or referrer data and passes it as a custom parameter, BigQuery can break down revenue by AI platform.
BigQuery tables connect directly to Looker Studio as a data source — create a dashboard with filters for purchase_origin that let clients toggle between human performance, agent performance, and blended views.
GA4’s BigQuery export is free for up to 1 million events per day on the free tier, and the queries needed for agent commerce analysis typically scan small amounts of data — most WooCommerce stores will spend under $5/month on agent-specific BigQuery analysis.
Count purchase events by referrer source domain where the source matches known AI platforms — this gives you an immediate baseline of how much AI-attributed revenue you already have, without any server-side tracking changes.