A monthly AI-attributed revenue report answers one question: what traffic and revenue arrived from named AI assistants last month, compared with the rest of the store. Seven rows cover it: AI sessions, AI share of sessions, AI purchases, AI purchase revenue, AI conversion rate vs site, AI revenue per session vs site, and the top assistant by revenue.
- The number is a floor, not total AI influence. GA4 counts Google's AI Overviews and AI Mode in Organic Search and stripped-referrer visits in Direct, and it never sees orders completed inside Copilot, AI Mode or Gemini checkout. Read Shopify's agentic storefronts report beside it, never added to it.
- Build it on a GA4 custom channel group with one AI regex ordered above Referral, which applies to past data too, and a Data Studio report with a CASE field that names each assistant.
- Before you celebrate or panic about a month-over-month move, check the denominator, the order count, any regex edits and the same month last year. Small samples swing wildly.
- Give a CFO three lines: what moved, the likely reason, and the one action you'll take. Call it AI-attributed revenue, never AI revenue, and never add ad-platform conversions on top.
Short answer: a monthly AI-attributed revenue report is one page of numbers, refreshed on the first working day of each month, that shows sessions, purchases and revenue from AI assistants such as ChatGPT, Perplexity, Gemini and Copilot next to the same figures for the rest of the site. You build it in GA4 with a custom channel group that catches every named AI referrer, read it in Data Studio (formerly Looker Studio), and write three lines about it. The hard part is reading it honestly: it counts only what GA4 can trace to an AI assistant, so it's a floor.
This guide is for founders, ecommerce managers and agency analysts who get asked "how much revenue is AI bringing us?" It walks through our AI-Attributed Revenue Report Template section by section. The template is an email-gated CSV build sheet (33 steps, each with where to click, the setting or formula, why it matters and how you know it's done), not a one-click report copy, so every field stays yours to change. Platform rules below come from Google Analytics Help and Google's Data Studio documentation, opened on October 22, 2026 and listed under Sources.
- AI-attributed revenue
- Purchase revenue from sessions whose source GA4 can identify as an AI assistant, using last-click attribution within the session. It excludes purchases influenced by AI answers that arrived through Organic Search, Direct or another channel, so it understates AI's total effect.
- AI Assistant channel (GA4)
- A channel in GA4's default channel group, added in 2026. GA4 sets the medium to
ai-assistantand the campaign to(ai-assistant)when the referrer matches Google's list of AI assistants, which it describes as including ChatGPT, Gemini, DeepSeek, Copilot and Grok. Google's AI Overviews and AI Mode are excluded.
- Custom channel group
- Your own set of rule-based channels in GA4, created by copying the default group and editing it. Google says traffic goes into the first channel whose definition it matches, and that custom groups can be applied to reports retroactively.
What should a monthly AI revenue report include?
Fewer numbers than you'd think. The report exists to answer "is AI a real channel for us, and is it growing?", so every row either measures size, compares AI with the rest of the site, or points to where to act. These are the KPI rows the template logs each month.
| KPI | How it's calculated | The question it answers | Watch out for |
|---|---|---|---|
| AI sessions | Sessions where the Is AI field = AI | How many visits came from named AI assistants? | Undercounts AI Mode, AI Overviews and referrer-less visits |
| AI share of sessions | AI sessions ÷ all sessions | Is AI growing faster than the store? | Rises when the rest of the site falls |
| AI purchases | Ecommerce purchases where Is AI = AI (the purchase key event) | Do AI visitors buy at all? | Single digits make every rate unstable |
| AI purchase revenue | Purchase revenue where Is AI = AI, shown beside Shopify's agentic storefronts sales | What can finance tie to AI channels? | Misses in-chat checkout orders; GA4 nets out refunds |
| AI purchase conversion rate vs site | Ecommerce purchases ÷ sessions, AI vs Not AI | Does AI traffic convert better or worse here? | Don't swap in Session key event rate; it counts every key event |
| AI revenue per session vs site | Purchase revenue ÷ sessions, AI vs Not AI | Is an AI visit worth more than an average one? | One large order can double it in a small month |
| Top assistant by revenue | First row of the assistant table, sorted by revenue | Which assistant actually sells for us? | Mark any assistant under 100 sessions as low confidence |
Two KPIs get misread. GA4's Session key event rate counts sessions with any key event, so it won't match a purchase rate if sign-ups are key events; the template defines its own purchase rate. And GA4's Purchase revenue is net of refunds, per Google's definition, so it won't match gross sales to the dollar.
Published studies disagree on whether AI visitors convert better (what the evidence says). The AI-vs-site rows answer it for your store, every month.
What does an AI revenue report leave out?
- Google's AI Overviews and AI Mode. GA4's documentation says the AI Assistant channel excludes them, so their clicks count in Organic Search, and nothing in GA4 splits them out. Estimating them is a Search Console job, covered in how AI Mode traffic hides in Organic Search.
- Visits with no referrer. When an app or browser strips the referrer, the session lands in Direct, and neither GA4's channel nor any regex can see it.
- Assistants not on the list. Google doesn't publish its full list of AI assistants, and a regex only matches what it names.
- Influence without the click. Someone who reads a ChatGPT answer and later searches your brand name shows up as Organic or Direct. Session-level last click can't credit the assistant.
- Orders completed inside an AI checkout. Shopify's help pages say Google Analytics and custom pixels don't fire in Microsoft Copilot's or Google AI Mode and Gemini's direct checkout; only server-to-server pixels do. Those orders never reach GA4.
- Paid ChatGPT clicks, if untagged. ChatGPT Ads traffic needs UTMs and its own channel, or it can blend into your AI row. Our ChatGPT Ads conversion tracking guide sets that up.
None of that makes the report useless. It makes it a floor you can trend, with a caveats box on page one. For the in-chat orders, Shopify's admin has its own view: Sales channels → Agentic reports sales, orders, online store sessions and conversion for each AI channel. Put those figures beside GA4's AI revenue, not added to it: ChatGPT sends Shopify shoppers to your own checkout, so some of the same orders appear in both.
How do you set up GA4 for the report?
GA4's built-in AI Assistant channel is a good baseline, and the custom channel group extends it to assistants Google's list may not recognize yet. You need the Editor role on the property.
- 01
Check the built-in channel
Open Reports → Acquisition → Traffic acquisition with Session default channel group as the dimension, and look for the AI Assistant row. Note last month's sessions, even if it's zero.
- 02
Create a custom channel group
In Admin, under Data display, open Channel groups and click Create new channel group. It starts as a copy of the default group, which stays untouched. Standard properties can have two custom groups.
- 03
Add one channel for every AI assistant
Click Add new channel, then + Add condition group. Choose the Source dimension, set it to matches regex and paste:
.*(chatgpt|openai|perplexity|gemini\.google|bard\.google|copilot|claude\.ai|deepseek|grok|meta\.ai|chat\.mistral|you\.com|phind|poe\.com).*GA4 regex conditions must match the whole value, hence the.*on both ends. - 04
Move it above Referral
Click Reorder, drag the AI channel above Referral (and above AI Assistant, so every assistant lands in one row), click Apply, then Save group. Google's own AI-assistant example gives the same instruction, because a session goes into the first channel it matches.
- 05
Confirm `purchase` is a key event
In Admin → Data display → Events, check that
purchasehas recent counts. Google markspurchaseas a key event by default. On Shopify, the GA4 setup through the Google & YouTube channel tracks ecommerce events automatically, but check rather than assume.
It catches any source containing those words, including a blog with "copilot" in its domain. A false positive you can see in a table is better than a real assistant you miss. Read the matched sources every quarter, and remember that ChatGPT Ads sessions tagged utm_source=chatgpt also match unless a paid channel sits above this one.
Because Google applies custom channel groups to past data, last month is reportable on day one. The flip side: Google also says an edited group's new definition applies to the underlying data immediately, so changing the regex quietly rewrites earlier months in your reports. That's why the monthly log matters. More on the channel itself, including what Google's AI Assistant rule does and doesn't catch, is in our GA4 AI Assistant channel guide.
How do you build it in Data Studio?
Data Studio reads GA4 directly, so the report refreshes itself. Google brought back the Data Studio name in 2026, and old Looker Studio links redirect to datastudio.google.com. You need at least Read & Analyze access to the GA4 property.
- 01
Connect GA4
Click Create → Data Source, choose the Google Analytics connector, authorize, pick the account and property, and click Connect. Then Create report and Add to report.
- 02
Add the AI assistant field
In the data source, add a calculated field named AI assistant: a
CASEstatement with oneWHEN REGEXP_MATCH(Session source, ...)line per assistant (ChatGPT, Perplexity, Gemini, Copilot, Claude, DeepSeek, Grok), then a catch-all line using the full regex that returns Other AI, andELSE "Not AI". The template has it ready to paste. - 03
Add Is AI and the two ratios
Is AI returns AI or Not AI from the field above. Revenue per session is Purchase revenue ÷ Sessions. Purchase conversion rate is Ecommerce purchases ÷ Sessions, typed as a percent.
- 04
Fix the reporting month
Under File → Report settings, set the default date range to the last full calendar month. On each scorecard and table, set Comparison date range to Previous period; add Previous year for seasonal stores.
Two rules from Google's function docs make the CASE field work. A CASE returns the result of the first WHEN that's true, so specific assistants go above the catch-all. And REGEXP_MATCH must match the whole value: Google's example shows 'A' failing against ABC123 while 'A.*' matches. Leave out the .* and nothing matches.
Google documents that the GA4 connector can't use GA4's own segments or comparisons, that a table filtered by a calculated field (such as Is AI) shows a null grand total, and that data-source calculated fields don't work in blended data. Use scorecards for totals, and build AI share of revenue from two scorecards or a chart-level field rather than a blend.
What are the four report pages?
- 01Summary. The KPI scorecards against the previous period, AI share of revenue, a 13-month trend so seasonality doesn't read as growth, and the caveats box.
- 02By assistant. A league table of assistants by revenue, plus a two-row AI vs Not AI benchmark.
- 03AI landing pages. The top 25 pages assistants send people to, and the AI landing pages with zero purchases: the cheapest fixes on the list.
- 04Organic and Direct context. The Organic Search trend (where AI Mode lives) and the Direct trend on product pages (where stripped referrers land), both labelled as indicators, not AI counts.
How do you read month-over-month moves?
Most bad readings come from one number in isolation. Run a move through this table first.
| What you see | Check first | Most likely reading |
|---|---|---|
| AI sessions up, AI share flat | All sessions; same month last year | The whole store grew: season, promotion or press. Not an AI gain. |
| AI share up, AI sessions flat | Sessions in the other channels | The rest of the site fell. Investigate that, not AI. |
| AI revenue jumps, sessions flat | AI purchase count; largest orders | A few big orders. Wait for a second month. |
| A new assistant appears or spikes | Matched sources; any regex edit; platform changes | A new referrer, a renamed domain or your own regex change, before it's a trend. |
| AI sessions down, Direct on product pages up | Direct landing pages; referrer changes | Possibly referrer stripping. A hint, not proof. |
| Organic down, AI rows flat | Search Console clicks and impressions | Possibly AI Mode or AI Overviews absorbing clicks. GA4 alone can't say. |
| AI conversion rate swings wildly | Sessions per assistant | Small sample. Under 100 sessions, treat it as noise. |
Don't act on an AI conversion rate built on fewer than 100 sessions in a month, and don't call a trend until it holds for two consecutive months and against the same month last year. The 100-session line is the template's small-sample flag, a judgment call rather than a statistical test: below it, two or three orders swing the rate by whole percentage points.
Also log changes on your side, such as a regex edit, a site migration or a new consent banner. They explain many surprises.
What should you tell a CFO about AI revenue?
A finance lead wants a defensible number, its trend, and whether to spend more. Give them three lines with the definition attached.
What moved: AI-attributed revenue was [amount], [x]% of online revenue, [up/down] [y]% on last month and [z]% on the same month last year. Why: [the specific cause you checked, such as a product page ChatGPT now sends traffic to, or a seasonal lift across all channels]. Action: [one thing, with an owner]. AI-attributed means sessions GA4 traced to a named AI assistant; it excludes AI Mode, AI Overviews, referrer-less visits and orders taken inside Copilot, AI Mode or Gemini checkout, so it's a floor. In-chat orders: see Shopify's agentic storefronts report.
- Say "AI-attributed", every time. It signals a measured floor, not a claim about all AI influence.
- Don't add platform numbers on top. ChatGPT Ads Manager credits conversions on its own windows, including view-through; adding them to GA4's AI row counts some sales twice. Report ad spend and ad-platform conversions on their own line.
- Reconcile once a month. Compare GA4's total Purchase revenue with your store's sales report for the same month and note the gap. If GA4 misses a share of orders store-wide, the AI row misses it too.
- Frame the trend, not the month. Six steady months beat one big one.
If the question becomes "what's at stake if we don't show up in AI answers?", that's a forecast, not a report. Keep it separate from the attributed number; our AI revenue calculator is built for that estimate.
How do you keep the report honest over time?
- Quarterly regex check. In an exploration, list the sources the regex matches, read the top 50, drop false positives and add new assistants in both GA4 and the Data Studio CASE field.
- Log every definition change. Because an edited channel group rewrites past months in reports, export the KPI rows each month and keep the log. The log is your history; the live report is today's definition.
For where this report sits in a wider measurement plan (GA4, Shopify, Search Console and server logs, and what each can see), read the AI traffic and attribution guide. For ChatGPT specifically on Shopify, see how to track ChatGPT traffic and revenue.
If the report shows assistants sending visitors to pages that don't sell, that's usually product data and page content, the work we do in Agentic Commerce Setup for Shopify stores.
FAQ
What is an AI-attributed revenue report?
- It's a monthly report of sessions, purchases and revenue that GA4 can trace to AI assistants such as ChatGPT, Perplexity, Gemini and Copilot, set against the rest of the store. It uses session-level last click, so it measures a floor: AI Overviews and AI Mode clicks sit in Organic Search, and visits without a referrer sit in Direct.
Can I use GA4's AI Assistant channel instead of a regex?
- You can start with it. GA4's AI Assistant channel groups referrers from Google's list of AI assistants, which Google describes as including ChatGPT, Gemini, DeepSeek, Copilot and Grok, but the full list isn't published. A custom channel group with a regex catches named assistants the default may miss. Many stores report both and compare them.
Does the custom channel group work on past data?
- Yes. Google says custom channel groups can be applied to reports retroactively, so you can report last month on the day you build the group. The same applies in reverse: editing the group's rules changes how earlier months appear in reports, which is why the template has you export and log the monthly KPI rows.
Why doesn't GA4's AI revenue match Shopify's numbers?
- GA4 counts only orders whose purchase event reached it, credits them to the session's source, and nets out refunds. Shopify records every order, including in-chat checkouts on Copilot, AI Mode and Gemini, where Shopify says GA4 doesn't fire. Compare store-wide totals monthly, and read Shopify's agentic storefronts report beside GA4's AI row.
How often should I update the AI regex?
- Every quarter, and whenever a new assistant launches or an existing one changes domain. List the sources the regex matches in a GA4 exploration, read the top 50, remove false positives and add missing assistants. Update the GA4 channel and the Data Studio CASE field together so the two stay in step.
Should ChatGPT Ads revenue go in the AI revenue report?
- Report it on its own line. Tag ChatGPT Ads with UTMs and give them a paid channel placed above your AI channel in the custom group, so paid clicks don't inflate organic AI numbers. Don't add Ads Manager's conversions to GA4's AI revenue: the two use different attribution, and adding them double-counts sales.
Is there a Data Studio (Looker Studio) template I can copy?
- Ours is a build sheet rather than a one-click copy: a CSV of 33 steps covering GA4 channel group settings, Data Studio calculated fields to paste, the four report pages, the KPI rows and the monthly routine. Building it yourself keeps every field and filter yours to change.
Sources
- 1.[GA4] Default channel group · Google Analytics HelpAI Assistant channel rule and named assistants; excludes AI Overviews and AI Mode
- 2.[GA4] Custom channel groups · Google Analytics HelpCreate, add and reorder channels; first match wins; applies retroactively; 2 groups on standard properties
- 3.[GA4] Mark events as key events · Google Analytics Helppurchase is a key event by default; Admin → Data display → Events
- 4.[GA4] Analytics dimensions and metrics · Google Analytics HelpPurchase revenue net of refunds; Ecommerce purchases; Session key event rate
- 5.Setting up Google Analytics 4 · Shopify Help CenterGA4 via the Google & YouTube channel tracks ecommerce events automatically
- 6.Connect to Google Analytics · Google Cloud documentation (Data Studio)Connector steps, permissions, no GA4 segments or comparisons
- 7.About calculated fields · Data Studio HelpData-source calculated fields can't be used with blended data
- 8.CASE (Searched) · Google Cloud documentation (Data Studio)CASE returns the first true WHEN
- 9.REGEXP_MATCH · Google Cloud documentation (Data Studio)Full-value match
- 10.Set a date range for your report · Data Studio HelpFile → Report settings default date range; Previous period comparison
- 11.Selling on Microsoft Copilot · Shopify Help CenterGoogle Analytics and custom pixels don't fire in Copilot direct checkout
- 12.Selling on Google AI Mode and Gemini · Shopify Help CenterGoogle Analytics and custom pixels don't fire in AI Mode and Gemini direct checkout
- 13.Managing agentic storefronts · Shopify Help CenterSales, orders, sessions and conversion per AI channel
Founder of GigaCommerce, part of Gigaverse. Works with mid-market Shopify and Amazon merchants on agentic commerce installs, AI-ready catalogs, and Commerce GEO.
LinkedIn