Can an AI agent monitor Google Merchant Center feed health?

Yes — it can check for disapproved products and surface which product groups they're concentrated in, which is often the real cause of a Shopping campaign's impressions quietly dropping.

Should every product be in the same Performance Max asset group?

Not if it can be avoided — blending a bestseller with a low-margin clearance item in one asset group gives Google's automation a noisier signal. Scoping by product category or margin tier is usually a better structure.

How does the agent compute true ROAS for an e-commerce store?

By joining Google Ads cost data with your own Shopify order and revenue data rather than relying solely on platform-reported, click-time-attributed conversion value, which doesn't account for refunds or returns.

Does the agent change PMax asset groups automatically?

No — it proposes a re-scoped structure for review. Every Performance Max configuration change pauses for your approval before it applies.

Is this different from just using Performance Max on its own?

Yes — PMax runs its own automation regardless. The agent's role is auditing feed health, structuring asset groups for a clearer signal, and computing real profitability from your own revenue data on top of what PMax does.

Playbooks

AI agent for Google Ads e-commerce: Shopping feed health, PMax by category, and true ROAS

TL;DR

E-commerce Google Ads runs on three things working together: a healthy product feed, Performance Max asset groups organized around what actually sells, and a ROAS number you can trust. An AI agent's job here isn't one optimization — it's reading all three in the same run, because a feed disapproval, a mis-scoped asset group and an inflated platform-reported ROAS all look like separate problems but usually share one root cause.

The three things that have to work together

Layer	What breaks silently	What the agent checks
Merchant Center feed	Disapproved products stop serving with no alert reaching anyone	Feed health, disapprovals, and which product groups they're concentrated in
Performance Max asset groups	One undifferentiated asset group blends your best and worst products	Asset-group scope, ad strength, and URL expansion/audience-signal configuration
ROAS reporting	Platform-reported conversion value drifts from what actually shipped, after refunds and returns	Ad cost joined against your own Shopify order data, not just Google's attributed value

Feed health first — nothing else matters if products aren't serving

A disapproved product isn't a Google Ads problem you'd find by looking at campaigns — it's a Merchant Center problem that shows up as a quiet drop in a Shopping campaign's impressions with no error surfaced anywhere obvious. "Check Merchant Center for disapproved products and tell me which product groups they're concentrated in" is the first goal worth running before anything else, because every downstream number assumes the feed is healthy.

Performance Max, scoped by what actually sells

The default PMax setup — one asset group for the whole catalog — blends a $200 bestseller with a $12 clearance item and lets Google's automation decide how to split budget between them with limited signal. Scoping asset groups around real product categories or margin tiers gives the automation a narrower, more coherent signal to optimize against.

1"Grade our PMax asset groups — which ones have Ad Strength below Excellent, and why?"
2"Which product categories are in the same asset group as our worst performers?" — the agent reads asset-group scope against product-level performance to find blending problems.
3"Propose a re-scoped asset-group structure by product category." — shown for review; every PMax configuration change pauses for approval before it applies.

True ROAS from your own store data

Google Ads reports conversion value from what it can attribute at click time — before returns, before refunds, before the discount code that shaved 20% off the order. For a store connected via Shopify, the agent can join ad cost against actual order and revenue data instead, which is what separates a real profitability number from a platform-reported one. See true ROAS from your own data for the underlying pattern, and how a Shopify connection works for the two connection tiers involved.

Where Performance Max's opacity still matters

Scoping asset groups well doesn't eliminate PMax's core trade-off — Google's automation still decides placement and timing inside each asset group without full visibility into why. For the deeper case on keeping a reviewable AI agent in that loop rather than trusting PMax's automation blind, see Beyond Performance Max: staying in control when Google's AI runs the campaign.

Frequently asked questions

Can an AI agent monitor Google Merchant Center feed health?: Yes — it can check for disapproved products and surface which product groups they're concentrated in, which is often the real cause of a Shopping campaign's impressions quietly dropping.
Should every product be in the same Performance Max asset group?: Not if it can be avoided — blending a bestseller with a low-margin clearance item in one asset group gives Google's automation a noisier signal. Scoping by product category or margin tier is usually a better structure.
How does the agent compute true ROAS for an e-commerce store?: By joining Google Ads cost data with your own Shopify order and revenue data rather than relying solely on platform-reported, click-time-attributed conversion value, which doesn't account for refunds or returns.
Does the agent change PMax asset groups automatically?: No — it proposes a re-scoped structure for review. Every Performance Max configuration change pauses for your approval before it applies.
Is this different from just using Performance Max on its own?: Yes — PMax runs its own automation regardless. The agent's role is auditing feed health, structuring asset groups for a clearer signal, and computing real profitability from your own revenue data on top of what PMax does.