Data, reconciliation & reportingCapability
A weekly report of which AI channel sold: Shopify's orders as the total, with GA4, server capture, Meta, Google Ads and ChatGPT Ads set against them.
Sounds familiar?
How I solve it
I build one weekly report that says which AI channel sold. Shopify’s orders are the total, and I check every other tool against them.
Rows are channels: ChatGPT, Google AI Mode and Gemini, Microsoft Copilot, Meta and Shop. Other AI referrers, such as Perplexity, come next. Shopify’s help doesn’t list Perplexity as a channel as of 2 October 2026, so it shows up only as a referrer. The last row holds orders the rule can’t place yet. Columns are systems: Shopify, GA4, your server-side capture of AI-checkout orders if you run one, and the conversions Meta, Google Ads and ChatGPT Ads recorded. When a column disagrees with Shopify, the cell says why, or says the gap isn’t explained yet.
Some gaps are platform behaviour, not bugs. Shopify says Google Analytics and custom pixels don’t fire in direct checkout on Google AI Mode and Gemini, Copilot and Meta. The buyer pays without opening your store, so GA4 has no session to take a source from. A server-side capture can send those orders with a channel label. GA4 also files AI Mode visits under Organic Search, not AI Assistant. The channel-by-channel guide has the dated details. Other gaps come from attribution windows, missing referrers or plain bugs, and the bugs are what I go after.
For agencies, the report goes out under your brand. The method is the order-by-order reconciliation I already use on ordinary channels.
Shopify’s help says orders from AI channels show in admin “with channel or referrer attribution”, and Meta orders are attributed to Meta. On the API side, as of 2 October 2026:
Order.attribution (API version 2026-07 and later) names the source that created the order, as a handle and a display name. The orders query’s channel filter matches that handle.Order.publication is the sales channel the order was created from. Order.app is the app that created it. Order.sourceName holds a source such as web, or a sales channel’s attribution handle.Order.channelInformation is deprecated. I don’t build on it.Order.customerJourneySummary keeps the first and last visit before the order, with referrer, landing page and UTMs, within 30 days. Shopify keeps this marketing attribution separate from the source that created the order.Order.attribution is filled for them at all.A ChatGPT buyer pays on your own checkout, so I expect that order to look like an Online Store order with ChatGPT in its visit data. Orders paid inside Google, Copilot or Meta skip the store, so I expect their visit data to be empty. Both are guesses until week one replaces them with a rule taken from your own orders.
The process
6 steps. Scope and a fixed price are agreed before the first one.
I need read access to Shopify, GA4 and the ad accounts you want in the report. The Agentic reports need the store owner, or a staff account with the Products > View and App and sales channel > Agentic permissions.
In week one I take recent orders that Shopify admin labels as AI-channel orders, read the same orders through the Admin API (attribution, app, publication, sourceName) and write down what marks each channel.
I compare per-channel sales from the rule with the report under Sales channels > Agentic and write down any difference before the first weekly report goes out. Shopify doesn't offer those reports for headless stores yet, so there the rule rests on the API alone.
I read Shopify orders through the API on a schedule, so nothing depends on a webhook arriving. Shopify itself advises a periodic reconciliation job next to webhooks. GA4 is split by its AI Assistant channel and Source Group, with shop.com added next to shop.app since Shop moved on 29 September 2026. Then come the server capture and the conversions Meta, Google Ads and ChatGPT Ads recorded.
I join order by order where both sides carry the order ID, and by totals where a platform reports only totals. Each gap gets a note: the cause where I can confirm it, or "not explained yet" and what I'll check next.
One page per store, the same layout every week, under your brand if you resell it. I re-check the channel facts behind it every month.
Handover
oppref click reference, and OpenAI says ChatGPT search referrals carry utm_source=chatgpt.com.Built with
Proof
Capability Work I do; the proof below is from related projects.
Before → after
There are no AI-channel client results yet, and I won’t stretch an older one to fit. The closest real work is measurement checked against real orders on ordinary channels. On a headless US store, the share of sales traceable to a source went from 6% to 76% after a server-side rebuild, read from the live GA4 and Google Ads APIs in August 2026. On a Greek herbal-tea brand, GA4 coverage checked against every real Shopify order went from 29% (84 of 290) before a server-side migration to 80% (176 of 219) after it, read in July 2026. The same project had first aligned one product ID across Merchant Center, GA4/Ads and Meta.
Headless Shopify automotive-repair retailer, US market
6% to 76%
sales traceable to their source 6% → 76% after a server-side rebuild; before it, 573 of 612 sales were invisible to every report.
Read the case: 16 Days Before Shopify's Deadline: Rebuilding Measurement for a Headless StoreHerbal-tea brand, Greece
29% to 80%
GA4 coverage against every real Shopify order 29% → 80% across a server-side migration, after one product ID was aligned across Merchant Center, GA4/Ads and the Meta catalogue.
Read the case: One Product ID Everywhere, and 29% → 80% Conversion CoverageClient words
No client words here. My reviews are about tracking and reporting on ordinary channels, not AI channels, and I won't stretch one to cover this page.
What you can look at
I have nothing from a client to show here yet. What you can open today is the free AI Shopping Readiness Check. Its “AI orders in your analytics” row explains why orders paid inside AI Mode, Copilot or Meta can only be checked from inside your accounts. The Tracking Watch page describes the daily reconciliation this report builds on.
No AI-channel report has been delivered to a client yet. The proof above is adjacent measurement work on ordinary channels, anonymised, with figures read from live platform APIs in July and August 2026. Platform facts on this page were read in Shopify, Google, OpenAI and Meta documentation on 2026-10-02. They change often, so I re-check them every month.
Price and timeline
Tracking Watch on a Technical Partnership is USD 1,000 per month: 8 engineering hours in an agreed area, one active task. For one store, the weekly report, the gap notes and the monthly re-check run inside it. If the first setup needs more than those hours, I quote the build separately before starting.
A server-side capture of one store’s AI-checkout orders is a Focused Fix from USD 450, with a fixed quote before it starts. Several stores or a white-label report across an agency’s client list is a Custom Engineering Project from USD 1,500, quoted in writing. If you need a one-time answer for one store, a Tracking Health Check is USD 395: read-only, one agreed question, up to three systems, a written verdict two working days after kickoff and access. Discussing the task is free. The rest of the AI-channel work is on the AI shopping channels page.
QuoteFixed before work starts
A note from Daniilbefore you decide
If Shopify’s Agentic report shows no AI-channel sales, wait. A weekly page of zeros costs you a partnership and tells you nothing. The free readiness check and a monthly look at Sales channels > Agentic are enough until orders show up.
If all your AI orders come through ChatGPT and GA4 already agrees with Shopify on them, you need a channel group in GA4, not a report. The AI referral traffic guide covers that.
And if what you want is to be recommended more often in ChatGPT or AI Mode, this isn’t it. Nobody can promise that, and I don’t.
— Daniil
Questions
Shopify's orders. Shopify shows AI-channel orders in admin with channel or referrer attribution, so its count is the total. GA4, the server capture and the ad platforms are compared against it. Each difference gets a written note: the cause where I can confirm it, or "not explained yet" and what I'll check next.
From the order itself. Shopify's Admin API documents Order.attribution, publication, app and sourceName for where an order came from, and customerJourneySummary for the visits before it. Which values Shopify's own AI channels write there is not documented as of 2 October 2026. So in week one I read your real AI-channel orders and write the rule down before the first report.
Not documented as of 2 October 2026. Shopify says Google Analytics and custom pixels don't fire in those direct checkouts, and neither Google nor Meta says whether such orders count as ad conversions. The report shows what each platform actually recorded. It doesn't assume either way.
Not the old ones. GA4's Measurement Protocol accepts events backdated by up to 72 hours, so an order older than that can't be added. It still shows in the report, because the report reads orders from Shopify.
Partly. As of 2 October 2026, Shopify says direct checkout in Google AI Mode and Gemini and in Copilot is shown only to US-based shoppers, and Meta's only to shoppers in the US, Canada and Mexico. For a UK or EU store the report covers AI assistant visits that end on your own checkout, plus ChatGPT Ads, which launched in the UK on 11 August 2026 and in 31 European countries in late August.
The first answer is free, within one working day. Or write directly: next@taskfordaniel.com