The AI revenue hiding in your 'Direct' bucket.
Dark Agentic Commerce Traffic — sessions from ChatGPT, Gemini, Perplexity, Rufus, and Copilot that GA4 misclassifies as Direct. We resolve them with referrer fingerprinting + crawler co-occurrence at 91% average confidence.
Sessions from ChatGPT, Gemini, Perplexity that GA4 misclassifies as Direct. Resolved via referrer fingerprinting + crawler co-occurrence.
Two signals. One re-attribution.
DACT combines referrer fingerprinting with crawler co-occurrence — every re-attributed session ships with a full evidence trail.
We watch the AI crawlers hit you
GPTBot, Google-Extended, PerplexityBot, ClaudeBot, OAI-SearchBot — every visit is logged with timestamp, URL, and bot identity.
Then a human session lands as 'Direct'
When a buyer clicks an AI citation, GA4 typically loses the referrer. We fingerprint the session by referrer headers, UA, and crawler co-occurrence on the same URL.
We re-attribute it to the right AI agent
Every match emits a confidence score. Sessions above 70% are re-attributed to ChatGPT, Gemini, Perplexity, Rufus, or Copilot — with full sessions/orders/CVR/AOV/GMV per platform.
Now your dashboards tell the truth
Marketing finally sees how much revenue AI is actually driving. Finance can defend the budget. Optimization queue stops chasing the wrong levers.
Full-funnel attribution per platform.
Sessions, orders, CVR, AOV, and GMV — for every AI surface, broken out the way GA4 should have.
| Platform | Sessions | Orders | CVR | AOV | GMV |
|---|---|---|---|---|---|
| ChatGPT | 5.7K | 231 | 3.42% | ₹4.7K | ₹5.05L |
| Gemini / AI Mode | 3.9K | 126 | 3.50% | ₹2.3K | ₹5.80L |
| Perplexity | 1.9K | 57 | 3.46% | ₹2.4K | ₹1.97L |
| Amazon Rufus | 1.2K | 42 | 3.39% | ₹2.4K | ₹1.15L |
| Microsoft Copilot | 691 | 20 | 3.44% | ₹2.4K | ₹45.5K |
Sample data from a live D2C customer · 30-day window
Why DACT is now a P0 for every D2C brand.
of GA4 'Direct' on commerce sites is actually AI-driven, on the customers we've measured.
of previously invisible AI revenue, recovered by a brand doing ₹50L/month direct.
changes to GA4, Shopify, or your warehouse — DACT annotates, it doesn't replace.
Frequently asked
What does DACT stand for?
Dark Agentic Commerce Traffic. It's the slice of buyer traffic that originates from AI assistants (ChatGPT, Gemini, Perplexity, Rufus, Copilot) but lands in GA4 as 'Direct' or 'Unassigned' because the AI surface didn't pass a referrer header.
Why does GA4 misclassify AI traffic as Direct?
ChatGPT, Gemini, and Perplexity often strip or omit the HTTP Referer header when a buyer clicks a cited link. GA4 falls back to 'Direct / None'. The session is real and converting — it's just attributed to the wrong channel, which makes AI look smaller than it is and breaks the case for AI investment.
How does the DACT Resolver actually re-attribute sessions?
Two signals working together. First, referrer fingerprinting: even when GA4 sees 'Direct', the session often has subtle headers (sec-ch-ua, sec-fetch-site, in-app browser tokens) that match a known AI surface. Second, crawler co-occurrence: if GPTBot or Google-Extended fetched the same URL within the conversion window, that's strong evidence the session came from an AI citation. Both signals combine into a confidence score.
What confidence threshold do you use?
We re-attribute at 70%+ by default. Average across the funnel is 91%. Customers can dial the threshold up or down based on their tolerance for false positives. Every re-attributed session is logged with its evidence trail — referrer, crawler timestamps, co-occurrence URL — so finance and analytics can audit it.
Does this require changes to my analytics stack?
No. DACT runs alongside GA4, Shopify, or your warehouse. We don't replace your attribution model — we annotate it. Sessions stay in GA4 as 'Direct'; we surface the re-attributed view in the Asva dashboard and push it into your warehouse via a clean event stream.
How much revenue does DACT typically recover?
On the customers we've measured, 18–24% of GA4 'Direct' on commerce sites is actually AI-driven. For a brand doing ₹50L/month direct, that's ₹9L–₹12L/month of revenue that was previously invisible. The exact number depends on category, content depth, and AI citation density.