AI agents don't browse.They read your feed.
Shopping agents recommend the products whose data they can read and trust — and skip the rest. A sophisticated catalog with a messy feed gets passed over. This gate scores the durable feed readiness every agent evaluates and names the one fault to fix first.
Over a million stores are already in AI shopping. The question is whether yours appears.
agents don't browse your store — they ingest your feed. They recommend the items they can read and trust, and skip the rest.
a basic catalog with a clean feed gets recommended; a sophisticated catalog with a poor feed gets skipped. The feed is what's graded.
a feed that contradicts your live page, or a missing identifier, makes an agent distrust the item — worse than a simple gap.
Auto-enrollment put your products in front of agents whether you prepared for it or not. What decides if an agent surfaces an item isn't how good your PIM is — it's whether the feed it reads is clean, complete, and internally consistent. This gate scores exactly that, item by item, and shows you the share of your catalog an agent would skip today.
Score one item's seven signals and watch the distrust gate fire.
Live gate · one item (tap a mark to cycle 0 → 1 → 2)
Item verdict
UNREADABLE
min 1 · mean 1.86
fix first: Feed-vs-page consistency
Resolve every contradiction — make the feed, the live page, and your schema agree on price, stock, title, and ids.
Grades durable feed readiness from your own marks — not today's checkout protocol. It connects to nothing, scores no people, and guarantees no placement.
The same verdict at the command line, across your whole catalog.
A zero-dependency Python engine grades every item offline and prints a per-item read plus the catalog rollup. The workbook reproduces every number it returns — three-way identical by construction.
==============================================================================
AGENTIC-COMMERCE PRODUCT-FEED READINESS GATE :: RedHub AI :: ACF-079
==============================================================================
Items graded: 5 Exposure rate (share UNREADABLE): 80%
CATALOG: CATALOG UNREADABLE (worst: Flagship running shoe)
------------------------------------------------------------------------------
[UNREADABLE] Flagship running shoe <- AGENT-DISTRUST GATE (Feed-vs-page consistency)
machine_readability=2 identifier_completeness=2 attribute_depth=2 variant_grouping=2 freshness=2 feed_page_consistency=1 trust_seller_block=2
min=1 mean=1.86 fix first: Feed-vs-page consistency
-> Resolve every contradiction — make the feed, the live page, and your schema agree on price, stock, title, and ids.
[AGENT-READY] Organic cotton tee
machine_readability=2 identifier_completeness=2 attribute_depth=2 variant_grouping=2 freshness=2 feed_page_consistency=2 trust_seller_block=2
min=2 mean=2.00 fix first: Identifier completeness
-> Give every item a stable id and a real GTIN/barcode plus brand; set a stable parent id for variant groups.
[UNREADABLE] Stainless water bottle
machine_readability=2 identifier_completeness=0 attribute_depth=2 variant_grouping=2 freshness=2 feed_page_consistency=2 trust_seller_block=2
min=0 mean=1.71 fix first: Identifier completeness
-> Give every item a stable id and a real GTIN/barcode plus brand; set a stable parent id for variant groups.
[UNREADABLE] Wireless earbuds
machine_readability=2 identifier_completeness=2 attribute_depth=0 variant_grouping=1 freshness=2 feed_page_consistency=2 trust_seller_block=1
min=0 mean=1.43 fix first: Attribute & answer depth
-> Fill structured attributes: title, description, price+currency, availability enum, condition, images, key specs.
[UNREADABLE] Yoga mat (variants)
machine_readability=1 identifier_completeness=1 attribute_depth=1 variant_grouping=0 freshness=1 feed_page_consistency=2 trust_seller_block=1
min=0 mean=1.00 fix first: Variant grouping
-> Group variants under a stable parent id with the distinguishing attributes (size/color/material) set.
------------------------------------------------------------------------------
PROTOCOL NOTE (planning only, not graded): Grade the durable feed first; wire whichever protocols your surfaces use second.
Grades DURABLE feed readiness, not today's checkout protocol. Connects to nothing,
scores no people. Placement is earned, not bought — no outcome guaranteed. Not legal advice.
==============================================================================Three rules keep the read honest.
MIN-as-headline
An item is graded by its weakest signal, never the flattering average — because an agent drops it on the single worst gap. One missing GTIN sinks an otherwise-perfect item. Weakest 2 is AGENT-READY, 1 is GAPS, 0 is UNREADABLE.
The agent-distrust gate
Two faults don't just make an item thin — they make an agent distrust it. A feed-vs-page contradiction (even minor drift) and a missing identifier each force UNREADABLE on their own, because an agent that can't tell which data is true deprioritizes the whole item.
Durable, not protocol-chasing
It grades the clean-feed layer every agent evaluates and reports checkout-protocol specifics for planning only — never gating on them. In-chat checkout changed in 2026 and will change again. The feed is the durable asset.
A readiness read from your own description, not a connection and not a guarantee.
What it is
- A deterministic read of a feed you describe — same verdict every time.
- A per-item AGENT-READY / GAPS / UNREADABLE call with a catalog rollup.
- Two dispositive gates on the faults that make an agent distrust an item.
- The fix-first fault for each item, with the fix note attached.
What it isn't
- Not a connection — it submits nothing and reads no live feed or platform.
- Not protocol-chasing — it grades the durable feed, not today's checkout flow.
- Not a guarantee — placement is earned, not bought; no surfacing or sale promised.
- It grades the feed, never a person, and isn't legal advice.
Scope: grades durable feed readiness from marks you enter and reports checkout-protocol status for planning only. It connects to nothing, submits nothing, ranks no people, and guarantees no agent surfacing or sale. Re-check protocol status before acting. Not legal advice.
Anyone whose products need to be found by an agent, not just a person.
- Ecommerce brands & DTC operators — to see which items an agent would skip before a competitor's cleaner feed wins.
- Marketplaces & multi-brand catalogs — to grade feed readiness across a large catalog and target the worst items first.
- Agencies & feed specialists — to turn "your feed isn't agent-ready" into a scored, prioritized fix list.
- PIM & merchandising teams — to prove that catalog sophistication doesn't help if the feed contradicts the page.
Be acted on by agents, and seen by people.
Straight answers before you buy.
Don't get skipped by the agent.
Make your feed readable.
One purchase, lifetime access, 12 months of updates. $79, once.
Grades durable feed readiness, not today's checkout protocol. Connects to nothing, submits nothing, scores no people. Placement is earned, not bought — no surfacing or sales outcome is guaranteed. Not legal advice.
Sold by RedHub AI LLC · Secured by Stripe · redhub.ai