Core Concept · Machine Perception Layer

What does AI actually think you sell?

This is not asking whether your copy is well written. It is asking what an AI shopping agent actually concludes your store sells, who it is for, and what problems it solves, based only on what it can extract and test against real buyer questions.

Category Core ConceptMeasures Machine PerceptionFeeds Into Technical Factor
AI scans a store atomfoundry-demo-store.com Building Merchant Understanding Object...
categoryskincare / organic
target_customersensitive skin
use_casepartial match
confidence0.94
AI Interpretability 8.7 / 12 This is what AI concluded, not what the copy said.
Illustrative example, not a specific measured result.
Definition
AI Interpretability measures whether an AI shopping agent can build an accurate, structured understanding of what a store sells, who it is for, and what problems it solves. AI does not grade your writing. It builds its own internal model of your business, called a Merchant Understanding Object, and Interpretability measures how accurate that model actually is.

This model feeds directly into the Technical factor of the AI Commerce Score, published as Core Technical & Interpretability at 15% weight in the Methodology. The 0 to 12 scale on this page is the detailed diagnostic used to build that model in the first place, six sub-signals, tested against real buyer intent, and checked by more than one method so the score does not rest on a single model's opinion.

The core data structure

The Merchant Understanding Object

This is not a summary written for humans. It is the structured object AI actually builds and reasons from, after reading a homepage, category pages, product pages, About, FAQ, structured data, navigation, and robots or llms.txt. Every field below gets filled in, or left uncertain, based only on what AI could actually extract.

merchant_understanding_object.json
brand: "Lumina Skin"
category: "organic skincare"
products: ["cleanser","moisturizer","serum"]
target_customer: ["sensitive skin","reactive skin"]
use_cases: ["eczema","rosacea","fragrance-free routines"]
problems_solved: ["irritation","dryness","flare-ups"]
price_position: "mid-range, under $60"
product_attributes: ["fragrance-free","dermatologist-tested"]
geography: ["US","EU"]
constraints: ["no synthetic fragrance"]
confidence: 0.94

This is the object every downstream layer works from, the Buyer Intent Matrix, the AI Interpretability Score, and eventually the AI Commerce Graph. Get this object wrong, and everything built on top of it is wrong too.

Illustrative example, not a specific measured result.
Same product, two different objects

The same store can score completely differently

Two stores selling identical products can produce very different Merchant Understanding Objects, based purely on how clearly their content is written. Here is what AI actually extracts from ambiguous copy versus specific copy.

✗ Store A: ambiguous copy
"We believe in the power of nature. Our products are crafted with love and the finest ingredients. Feel the difference today."
categoryUNKNOWN
target_customerUNCLEAR
use_caseNOT_FOUND
confidence0.12
AI Interpretability 2 / 12
✓ Store B: specific copy
"Organic skincare for sensitive skin. Fragrance-free. Dermatologist-tested. For eczema, rosacea, and reactive skin types. Under $60."
categoryskincare / organic
target_customersensitive skin
use_caseeczema, rosacea
confidence0.94
AI Interpretability 10 / 12
The commercial impact: a store scoring 10 out of 12 matches roughly 12 to 15 relevant buyer queries out of 15 tested. A store scoring 2 out of 12 matches 1 to 2. Every point of AI Interpretability compounds directly into Recommendation Share.
Illustrative example, not a specific measured result.
Why interpretability drives recommendations

The Buyer Intent Matrix

Once the Merchant Understanding Object exists, it gets tested against real buyer queries. This is what that test looks like for the clear-copy store above.

Buyer intentMerchant understandingMatch
"skincare for sensitive skin"sensitive skincare
96%
"eczema moisturizer"eczema
91%
"fragrance-free skincare"fragrance-free
94%
"affordable skincare"under $60
88%
"organic skincare"organic
93%
"natural skincare brands"uncertain match
54%
Illustrative example, not a specific measured result.
How the 12 points break down

The AI Interpretability Score, six signals

Six sub-signals combine into the 0 to 12 score. Click a card to see what it actually checks.

Category clarity
2.7 / 3
Click to flip
Category clarity
Does AI know the exact product category, not just a vague theme like "lifestyle" or "wellness"?
Target customer
2.4 / 3
Click to flip
Target customer
Does AI know who this is actually for, not just who might theoretically buy it?
Use-case clarity
1.8 / 2
Click to flip
Use-case clarity
Does AI know the specific situations where this product gets used?
Problem clarity
1.2 / 2
Click to flip
Problem clarity
Does AI know what problem this solves for the buyer, not just what the product is?
Attribute clarity
0.4 / 1
Click to flip
Attribute clarity
Does AI know the specific attributes buyers filter by, like fragrance-free or organic?
Intent consistency
0.2 / 1
Click to flip
Intent consistency
Does the picture stay the same across every page, or does the store contradict itself?
Composite for the example above8.7 / 12
Illustrative example, not a specific measured result.
How it's actually measured

Not just an LLM's opinion

AI Interpretability is not a single model guessing how good your copy sounds. It runs through a fixed, mostly deterministic pipeline before a score comes out the other end. It cycles on its own, or click a step to jump straight to it.

AI pulls the homepage, category pages, product pages, About, FAQ, structured data, and robots or llms.txt. This is the raw material, nothing has been interpreted yet.
What AI actually reports back

The AI Interpretability report

This is the kind of report the pipeline above produces for a single store, in plain language.

AI thinks you sellOrganic skincare for sensitive skin
AI thinks your customer isPeople with sensitive or reactive skin
AI understands these use caseseczema · rosacea · fragrance-free routines
AI is uncertain aboutanti-aging positioning
Strongest buyer intent matchsensitive skincare · 96%
Weakest buyer intent matchnatural skincare · 54%
8.7/ 12 · AI Interpretability
Illustrative example, not a specific measured result.
Where this leads

Interpretability is one layer of five

Being understood correctly does not guarantee a recommendation. It unlocks the next layer AI has to check. Click through to keep going.

1AI Readability
2AI Interpretability
3AI Trust Layer 4Recommendation Confidence
5Recommendation
Where this data goes

Every object becomes a graph node

The Merchant Understanding Object built for a single store does not get thrown away once a score is calculated. It becomes one node in the AI Commerce Graph, the dataset connecting merchants, signals, and AI recommendation outcomes across every store measured. AI Interpretability is the layer where that data model actually begins.

This is also the direction AI Commerce Intelligence is built around: individual factor pages like this one measure a single signal, but every measurement compounds into one shared dataset rather than staying a one-off audit.

FAQ

Questions about AI Interpretability

What is AI Interpretability?
AI Interpretability measures whether an AI shopping agent can build an accurate structured understanding of what a store sells, who it is for, and what problems it solves. AI builds this understanding into a structured Merchant Understanding Object, then tests it against real buyer queries. It is scored 0 to 12 across six sub-signals, and it feeds into the Technical factor of the AI Commerce Score.
What is a Merchant Understanding Object?
It is the structured summary AI actually builds from a store's pages: brand, category, products, target customer, use cases, problems solved, price position, attributes, geography, constraints, and a confidence score. It is what AI reasons from, not the store's own marketing copy.
Is the score just an LLM's opinion?
No. The pipeline combines intent tests against real buyer queries, a language model's evaluation for completeness and contradiction, and deterministic rule-based checks that confirm structured data is present and the category matches a known taxonomy. All three combine into the final score, not a single model's read of the copy.
What causes low AI Interpretability?
Low AI Interpretability is usually caused by ambiguous product category language, missing use-case descriptions, too much focus on brand story versus product specifics, no clear target customer in the copy, and product descriptions that describe how something looks rather than what it does and who it helps.
Is AI Interpretability the same as Semantic Clarity?
They are related but measure different things. Semantic Clarity measures whether your positioning language contains specific buyer-intent signals. AI Interpretability measures whether AI can build an accurate Merchant Understanding Object from your store, and whether that object survives contact with real buyer queries. A store can have specific language but still score low if the overall picture AI builds is incomplete or contradictory.
Is AI Interpretability its own factor in the AI Commerce Score?
No. It is a diagnostic layer that feeds into the Technical factor, published as "Core Technical & Interpretability" at 15% weight, in the AI Commerce Score. The 0 to 12 scale on this page is the detailed sub-model used to measure it in depth; see Methodology for how it rolls up.

See what AI actually thinks you sell

Check your AI Interpretability.

Get a free AI Commerce Score and see the exact Merchant Understanding Object AI builds from your store.

Free · No signup · No credit card · Results in 10 seconds