eComXLoading
0%

AgentWatch for Shopify

Know when AI shopping channels stop sending buyers.

AgentWatch tracks AI-referred sessions, orders, and revenue by channel, checks the store surfaces shopping agents rely on, and opens an incident with evidence when something drops. Read-only. It never touches your products or theme.

admin.shopify.com/apps/agentwatchLast 30 days
HealthyLast health check 12 min ago

AI-referred revenue

$18,420

+14%

AI-referred sessions

6,312

+9%

AI-referred orders

241

+11%

Global Catalog requests

48.7k

-3%

Sessions by channel

  • ChatGPT3,184
  • Google AI Mode and Gemini1,702
  • Microsoft Copilot893
  • Shop533

Sample data

The problem

AI channels don't send error emails. They just go quiet.

A broken catalog endpoint, a revoked analytics permission, a channel that stops referring buyers. None of it shows up in your daily sales until a week of revenue is gone. By the time you notice the dip, you're guessing at the cause.

ChatGPT sessions, last 14 days7-day median
Warning

ChatGPT sessions -41% vs 7-day median

Incident opened with evidence

Email alert sent

How it works

From installed to monitored in three steps.

01

Install with read-only access

AgentWatch asks for two scopes: read reports and read products. No theme block, no storefront script, nothing to configure.

02

History and health load first

On install, AgentWatch backfills 30 days of AI-commerce metrics and runs its seven health checks. You see a baseline on day one.

03

Get told when something breaks

Daily metrics and scheduled checks run in the background. A meaningful drop opens an incident and, if you want, sends an email.

Commerce surface health

Seven checks on the surfaces AI agents use to shop your store.

Shopping agents read your catalog, look up products, check cart support, and pull your policies through machine-facing endpoints. AgentWatch checks each one on a schedule, or on demand from the Health page. The cart check only reads what's advertised. It never creates a cart.

Warning

2 failures in a row

Critical

3 failures in a row

Resolved

2 healthy runs in a row

admin.shopify.com/apps/agentwatch/healthHealth run
  • Shopify analytics accessHealthy 412 ms
  • Global Catalog analyticsHealthy 538 ms
  • AI commerce connectionHealthy 186 ms
  • Product discovery toolsHealthy 224 ms
  • Product visibilityHealthy 301 ms
  • Cart capabilityHealthy 158 ms
  • Policies and FAQsHealthy 197 ms

Monitoring built around the signals Shopify already reports.

01

AI revenue by channel

Sessions, orders, and revenue for ChatGPT, Google AI Mode and Gemini, Microsoft Copilot, and Shop. Trends over 7, 30, and 90 completed days, compared with the previous period.

02

Agent checkout, split out

See sales the AI agent completed directly for the buyer, separate from last-click orders.

03

Global Catalog exposure

Store-level search requests plus your top 20 products by Global Catalog demand, with period-over-period change.

04

Incidents with evidence

Baseline, observed value, percentage change, and a full evidence timeline. Acknowledge, resolve, or reopen.

05

Email alerts that don't spam

One email when an incident opens, one if it turns critical, one when it recovers. Deduplicated.

06

No data is not zero

Missing data stays marked as missing. AgentWatch never reports a gap in history as an outage.

Detection rules

Incidents open on real drops, not noise.

Detection is deterministic. No LLM guessing. Each completed day is compared with the median of the seven days before it, and low-volume signals are ignored entirely.

Completed day
7-day median
Volume gate
-30% Warning
-60% Critical
SignalMinimum baselineWarningCritical
AI-referred sessions, per channel20 sessions/day median-30%-60%
Global Catalog search requests100 requests/day median-30%-60%

Auto-recovery: two completed days back at 80% or more of the baseline.

Read-only by design

No customer-level data. Ever.

AgentWatch reads Shopify's aggregate reports and nothing else. It never stores customer names, contact details, individual orders, or individual sessions. It never edits products, themes, or settings. Diagnostic packets are sanitized, so you can share them with support safely.

FAQ

Questions merchants ask

Data, detection, and privacy.

From Shopify. AgentWatch reads aggregate AI-referred sessions, orders, and revenue through ShopifyQL, grouped by Shopify's agentic referring channel. Orders and revenue use Shopify's last-click attribution. Nothing is tracked on your storefront.

Analytics use completed days, so today's numbers arrive tomorrow. Health checks run on a schedule throughout the day, and you can run them manually from the Health page with a one-minute cooldown.

Shopify reports agent checkout for Google AI Mode and Gemini, Microsoft Copilot, and Shop, but not for ChatGPT. AgentWatch shows No data for those columns instead of a misleading zero.

No. An incident needs eight consecutive completed days of history and a baseline above the minimum volume. Health incidents need two failures in a row. One flaky response never pages you.

No, and it doesn't claim to. AgentWatch reports what Shopify measures and checks the surfaces agents use. It tells you when something changes so you can act on it. It does not edit your catalog or promise placement.

English, French, Japanese, Simplified Chinese, and Brazilian Portuguese. Alert emails and diagnostic packets are in English.

Uninstalling triggers Shopify's mandatory data deletion. You can also delete AgentWatch's monitoring history yourself from Settings at any time.

Ready when you are

Stop finding out about AI channel drops a week late.

Install AgentWatch and get a 30-day baseline on day one. Free to install.