How the AI Readiness Score works
The AI Readiness Score (the score) rates from 0 to 100 how completely your product information is set up in a form AI can easily read. This page explains the points and conditions for all 18 items, the verdict thresholds, when the score changes, and tips for raising it.
How the score is made up
The score is out of 100 points and is split into three sections. On the product diagnosis screen, items are listed by section.
| Section | Points | Items |
|---|---|---|
| Basic information | 40 pts | 6 items |
| Product attributes | 50 pts | 8 items |
| Purchase context | 10 pts | 4 items |
Three levels for each item
Each item is rated at one of three levels. "Structured" means the information is stored in a dedicated field on the product. This is different from information that only appears somewhere in the description text.
| Level | Points | Shown on the diagnosis screen |
|---|---|---|
| Structured | Full points for the item | Registered as structured data |
| Detected in the description | Half the item's points (rounded down) | Detected in the description (structuring it adds +5 pts) |
| Not registered | 0 pts | Not registered (worth 10 pts) |
For items detected in the description, the points are half the item's total, rounded down. A 10-point item gets 5 points, and a 3-point item gets 1 point. Items worth an odd number of points get less than half.
Points and conditions for all 18 items
The points and conditions in the tables below are defaults and may change.
"App field" in the tables means one of this app's product metafields (the fields saved when you apply an AI suggestion). For items that have an app field, applying an AI suggestion makes the item structured.
Keywords are searched for in the product title and the product description. The keywords in the tables are typical examples. Matching is not case-sensitive.
Basic information (40 pts)
| Item | Points | Full points (structured) | Detected in the description (half points) | 0 pts |
|---|---|---|---|---|
| Product title | 3 pts | 3 or more words in English, or 10 or more characters in Japanese | 1–2 words in English, or 1–9 characters in Japanese (1 pt) | The title is empty |
| Product description | 20 pts | 60 or more words in English, or 300 or more characters in Japanese, not counting formatting tags (such as bold or line breaks) | 10–59 words in English, or 50–299 characters in Japanese (10 pts) | Shorter than that |
| Product category | 10 pts | A Shopify product category (the standard classification) is set | No category, but a product type is entered (5 pts) | Neither is set |
| Vendor | 2 pts | The vendor is filled in. It can be the same as your store name | This level does not apply | The vendor is empty |
| Product images | 3 pts | 3 or more images or media | 1–2 (1 pt) | None |
| Price & variants | 2 pts | Every variant has a price greater than 0 | This level does not apply | A variant has a price of 0 or less |
The app detects whether a product is in English or Japanese automatically from the characters in its title and description. English products are measured in words separated by spaces, and Japanese products in characters.
Product attributes (50 pts)
| Item | Points | Full points (structured) | Detected in the description (half points) | Example keywords |
|---|---|---|---|---|
| Target customer | 10 pts | The app field Target customer has a value | A keyword is found (5 pts) | for women for kids children baby beginner adult |
| Use cases | 10 pts | The app field Use cases has a value | A keyword is found (5 pts) | ideal for perfect for great for suitable for |
| Materials | 10 pts | The app field Materials has a value | A keyword is found (5 pts) | material made of cotton leather stainless |
| Specifications | 10 pts | The app field Specifications has a value, or a variant option is named Size (Japanese equivalents also count) | A number with a unit, or a keyword, is found (5 pts) | 25cm 500ml size dimensions weight capacity |
| Key features | 4 pts | The app field Key features has a value | The description contains a bulleted list, or a keyword is found (2 pts) | features key features highlights benefits |
| Usage scenes | 3 pts | The app field Usage scenes has a value | A keyword is found (1 pt) | commute travel outdoor office everyday |
| Color | 2 pts | A variant option is named Color (Japanese equivalents also count) | A color word is found (1 pt) | color black white navy |
| Compatibility | 1 pt | A keyword alone earns full points | This level does not apply | compatible USB-C Bluetooth IPX4 waterproof |
Purchase context (10 pts)
| Item | Points | Full points (structured) | Detected in the description (half points) | Example keywords |
|---|---|---|---|---|
| Gift occasions | 3 pts | The app field Gift occasions has a value | A keyword is found (1 pt) | gift present birthday anniversary |
| Comparison points | 3 pts | The app field Comparison points has a value | A keyword is found (1 pt) | compared versus previous model |
| Cautions | 2 pts | The app field Cautions has a value | A keyword is found (1 pt) | caution warning care instructions |
| Shipping information | 2 pts | A keyword alone earns full points | This level does not apply | shipping delivery ships within |
Compatibility and Shipping information work differently
Compatibility and Shipping information don't yet have a structured-data level. Instead, if a keyword is found in the product title or description, the item gets full points straight away, without the half-points level. If no keyword is found, the item gets 0 points.
What does not count toward the score
- The AI summary is not part of the score. The summary condenses information you already have and does not add anything new
- Product metafields you added yourself (fields that do not belong to this app) are not used for scoring. No matter how carefully you fill them in, your score does not change
Registered items versus points
On the product diagnosis screen, below the AI Readiness Score, you see how many of the 18 items are registered out of the total. Under Diagnosis, each section heading shows a count on the right, such as 4 / 6 registered. Both counts include structured items and items detected in the description. Because items detected in the description earn only half their points, a product can have many registered items and still score low.
Verdict thresholds
Each product gets a verdict based on its score.
| Verdict | Score |
|---|---|
| Good | 80–100 pts |
| Needs Improvement | 50–79 pts |
| Poor | 0–49 pts |
These thresholds are defaults and may change. If they change, the new thresholds apply the next time you open a screen. You do not need to rescan.
Products that do not have a score yet are shown as Not diagnosed. When you filter the product list by verdict, products that are Not diagnosed do not appear. For details, see Finding and filtering products.
Catalog score
The Catalog score on the home screen is the average score of all scanned products, rounded to the nearest whole number.
- Products are not filtered by status. Draft and archived products are included in the average along with active ones
- While a scan is running and no product has been scored yet, the average cannot be calculated, so the catalog score shows "—"
- If the scan finishes and there are no products at all, the catalog score section is not shown
When the score changes
| When | What happens |
|---|---|
| You create, update, or delete a product in Shopify | That product's score is recalculated automatically. Deleted products are removed from the list and the score |
| You apply an AI suggestion | The score is recalculated right away |
| You open a product's diagnosis screen | The score is recalculated on the spot and saved again if it differs from the previous value |
| You run Rescan | All products in your store are read again and rescored |
| The points or detection conditions change | A rescan starts automatically the next time you open the app, and products are rescored against the new criteria |
| The verdict thresholds change | Scores stay the same. Only the verdicts are shown using the new thresholds |
The product diagnosis screen shows a last-diagnosed date and time. This is not when the score was calculated. It is when the product was last updated in Shopify. After you update a product, the score is recalculated without you having to reopen the screen.
You cannot run Rescan again while a scan is running. You also cannot run it within one hour after the last scan finished.
Tips for raising your score
- Start with the items that show the largest +5 pts or −10 pts on the diagnosis screen. That number is how many points the item is missing
- Create a suggestion with Optimize with AI, review it, and apply it. Items you apply become structured and earn full points
- Set a Shopify product category. Product category is worth 10 points
- Make your product description at least 60 words long (300 characters for Japanese products). Product description is worth 20 points
- Add at least 3 product images
- If your variants are split by size or color, name the options
SizeorColor - Mention shipping and compatibility in the product title or description. A mention alone earns full points for Shipping information and Compatibility
The score is a guide to how completely your product information is set up in a form AI can easily read. A higher score does not guarantee that your products will be shown or recommended in AI shopping experiences such as Muse.
Troubleshooting
- I added keywords, but my score barely went up: Items detected in the description earn half their points (rounded down), so a 3-point item gains only 1 point. To earn full points, apply an AI suggestion so the item becomes structured
- I updated a product, but the score didn't change: Automatic recalculation is processed in order. Wait a moment, then reopen the product's diagnosis screen. Opening the diagnosis screen recalculates the score on the spot. If the score still doesn't match the product, run Rescan
- The catalog score is lower than expected: Draft and archived products are included in the average
- My score changed one day without my doing anything: The points or detection conditions may have changed, triggering an automatic rescan