AI Commerce University · Guide 05

AI Retrieval Signals

The 8 signals behind your AI Commerce Score

AI does not evaluate a store from one signal. It builds an understanding from multiple machine-readable, semantic, and external signals. Atom Foundry measures those signals across every scanned store.

45
/ 100
AI Commerce Score
industry average
Discover
Understand
Trust
Match
Recommend
6,789 stores scanned · v3.0 · 8 factors
The whole model, first

How the score is built.

AI Commerce Score
Discovery
Technical
Visuals
Understanding
Structured signals
Trust
Trust signals
Intent match
Recommendation
External authority

8 measurable signals, weighted by how much evidence each one gives AI to work with. All 100 points add up to the score you see on every scan.

Semantic Visuals and Image Clarity
15%
AI Structured Signals
15%
Core Technical and Interpretability
15%
AI Trust and Transaction Confidence
15%
Commerce and Feed Accuracy
15%
User Intent Match
10%
Recommendation Confidence
10%
External Authority Signals
5%
Each factor, explained

Every signal, one by one.

01
Semantic Visuals and Image Clarity
15 points
Can AI understand what your product actually looks like?
Product images
+
Alt text
+
Product context
AI vision
Semantic clarity
Image accessibility Descriptive alt text Image and product consistency Semantic description Visual clarity
Why it matters: AI systems increasingly interpret product imagery directly. The image itself becomes part of the commerce signal, not just decoration next to the real information.
Typical failure
product-image-4.jpg
Machine-readable
Black heavyweight organic cotton oversized sweatshirt
02
AI Structured Signals
15 points
Does AI get facts, or does it have to guess?
Product
name
description
image
brand
offers
price
currency
availability
Organization
name
logo
identity
Review
ratingValue
reviewCount
AI gets facts instead of guesses.

Covers JSON-LD Product, Offer, and AggregateRating schema, Organization schema for merchant identity, and llms.txt quality. See AI Structured Signals for the full breakdown.

03
Core Technical and Interpretability
15 points
Can AI crawlers actually access and parse the store?
robots.txt
GPTBot
ClaudeBot
HTML
DOM
Headings
Core Web Vitals
Accessible
Extractable
Interpretable

Covers robots.txt configuration, Core Web Vitals via the PageSpeed API, and DOM structure quality including heading hierarchy. A technically broken store is an invisible store. See AI Readability and AI Interpretability.

04+05
AI Trust and Commerce Accuracy
15 + 15 points
The two factors with the most direct business impact.
Trust
Can AI verify you?
Reviews
Returns
Identity
Checkout
Policies
Commerce
Can AI safely describe the offer?
Price
Currency
Availability
Product
Feed
Trust tells AI "you are safe." Commerce tells AI "this is what you're selling."

Return policy in crawlable HTML, checkout accessibility for autonomous agents, price visible without JavaScript, and inventory consistent between feed and live page. Together these two factors fail in more than 60% of scanned stores. See AI Trust Confidence.

06
User Intent Match
10 points
Does your positioning actually fit what the buyer asked for?
Buyer prompt"Best organic collagen for endurance athletes"
Store A
"Premium wellness products for everyone."
31 / 100
Intent match
Store B
"Organic collagen supplements for endurance athletes."
94 / 100
Intent match
AI does not need your keywords. It needs semantic alignment between what the buyer asks and what your store represents. Measured via embedding comparison against real buyer prompt language for that category.
07
Recommendation Confidence
10 points
The social proof signals AI can actually parse.
81
/ 100
Reviews
Identity
↓ feeds ↓
Confidence to recommend
Intent Match
Does this store fit the question?
Recommendation Confidence
Can AI confidently endorse it?

AggregateRating schema with a real ratingValue and reviewCount from a verified platform, plus sentiment analysis on accessible review content. See Recommendation Confidence.

08
External Authority Signals
5 points
Evidence that exists outside your own storefront.
RedditReviewsPressCitationsMentionsExternal databases
External evidence
Current weight: 5%. We keep this factor intentionally small because external authority is harder to measure deterministically than on-site commerce signals. See Hidden Authority for more on how this evidence gets discovered.
The important part

A score is not a diagnosis.

Two stores can land on the same number for completely different reasons. Here is one example, factor by factor.

Example store
Semantic Visuals
14/15
Structured Signals
13/15
Technical
8/15
Trust
6/15
Commerce
7/15
Intent
8/10
Confidence
6/10
Authority
2/5
Total
64/100
AI Commerce Score: 64
But the real diagnosis: your biggest constraint isn't visibility. It's verification.
Score zones

What your score means.

85-100
Highly recommendable
AI can confidently recommend.
70-84
Moderately recommendable
Strong foundation, inconsistent advantage.
50-69
Low confidence
AI can read you, but critical signals are weak.
0-49
AI visibility risk
Too many gaps for reliable recommendation.
What to do about it

What changes the score.

Directionally, not precisely. Exact point deltas depend on the store, so this shows which factor each change moves rather than a specific number.

Add Product schema
↑ Structured Signals
Server-render price
↑ Commerce Accuracy
Add crawlable returns
↑ Trust
Sharpen positioning
↑ Intent Match
Improve image semantics
↑ Visual Clarity
Improve reviews
↑ Confidence
Build external authority
↑ Authority
FAQ

Frequently asked questions.

What are the 8 AI retrieval signals?
The 8 factors in AI Commerce Score v3.0 are: Semantic Visuals and Image Clarity (15%), AI Structured Signals (15%), Core Technical and Interpretability (15%), AI Trust and Transaction Confidence (15%), Commerce and Feed Accuracy (15%), User Intent Match (10%), Recommendation Confidence (10%), and External Authority Signals (5%). Maximum score is 100.
Which AI retrieval signal matters most?
All five 15% factors are equally weighted. AI Trust and Transaction Confidence and Commerce and Feed Accuracy fail most often across scanned stores. Fixing prices in HTML and making return policy server-rendered typically provide the fastest score improvement.
How is the AI Commerce Score calculated?
The AI Commerce Score v3.0 is 100 points across 8 factors. Above 85 is High Confidence. Between 70 and 84 is Moderate Confidence. Between 50 and 69 is Low Confidence. Below 50 is AI Visibility Risk.
Continue the curriculum

Up next.

Now measure your own signals.

Your store may look perfect to humans and still have critical gaps in the signals AI uses to understand it.

Free · 8 factors · 10 seconds