Weekly AI-attribution report
This mock shows the shape of the hands-on report. Every source label is backed by observable evidence. Orders with no referrer, UTM or customer declaration remain Unknown.
The week at a glance
| Metric | This week | Prior week | Change |
|---|---|---|---|
| Shopify orders | 428 | 401 | +6.7% |
| Shopify revenue | £31,840 | £29,510 | +7.9% |
| Orders with evidence of AI-assisted discovery | 17 | 12 | +5 |
| Revenue with evidence of AI-assisted discovery | £1,462 | £938 | +£524 |
| Share of weekly revenue | 4.6% | 3.2% | +1.4 pts |
| Orders with unknown source | 96 | 91 | +5 |
What changed: Verified AI-assisted revenue rose this week. ChatGPT produced the largest evidenced share. Nine further customers named an AI assistant in the post-purchase survey, including four orders whose browser referrer was missing. Those four are attributed by customer declaration, not inferred from Direct traffic.
Evidenced AI-assisted revenue
| Source | Orders | Revenue | Evidence mix |
|---|---|---|---|
| ChatGPT | 9 | £812 | 4 referrers, 2 UTMs, 3 declarations |
| Perplexity | 4 | £336 | 2 referrers, 2 declarations |
| Gemini | 3 | £241 | 1 referrer, 2 declarations |
| Claude | 1 | £73 | 1 declaration |
| Total | 17 | £1,462 |
OBSERVEDDECLAREDUNKNOWN
Observed: a recognised AI-domain referrer or campaign UTM persisted to the Shopify order. Declared: the customer explicitly named the source after purchase. Unknown: neither signal exists, so no source is assigned.
Orders behind the number
| Order | Revenue | Named source | Evidence |
|---|---|---|---|
| #10482 | £146 | ChatGPT | chatgpt.com referrer |
| #10491 | £89 | Perplexity | Customer declaration |
| #10503 | £212 | Gemini | utm_source=gemini |
| #10517 | £73 | Claude | Customer declaration |
| #10529 | £118 | Unknown | None |
The last order makes the boundary visible. It is not counted as AI-assisted revenue.
Survey health
| Metric | Result |
|---|---|
| Customers shown the question | 428 |
| Responses | 137 |
| Response rate | 32.0% |
| AI assistant named | 9 |
| Other free-text responses | 18 |
| No response | 291 |
Question used: “Where did you first hear about us?”
AI options: ChatGPT, Perplexity, Gemini, Claude, another AI assistant, the merchant's other main channels and free text.
What the merchant can do with this
- Protect the signal. Keep the survey running for another week before moving budget. Seventeen orders are useful, not yet a stable trend.
- Inspect ChatGPT discovery. Review the landing pages behind the four ChatGPT-referrer orders. Strengthen pages that repeat.
- Do not re-label Direct. The 96 unknown-source orders remain unknown.
Data quality and limits
- The report joins Shopify order value to surviving referrer/UTM evidence and customer declarations.
- A declaration is useful but self-reported. It stays separate from technical evidence.
- Referrers can be stripped by apps, browsers, copied links and multi-device journeys.
- Unknown is a valid result, not a gap filled with fingerprinting or probability.
- Refunds and cancellations should be deducted in the next report if they occur after cut-off.
Evidence standard: observed, declared or unknown. Never guessed.