Elbuz scores every active product from 0 to 100 on what an AI agent needs: barcode, photos, description, attributes and more. Click a problem to open exactly the products to fix.
When you need this:
- Planning a ChatGPT or Google feed — one number shows how much of the catalog is ready.
- Not sure where to start — categories are sorted by gain: the first raises the score most.
- Descriptions are missing — AI description writing opens from the category row.
- Categories aren't mapped to Google — each loses 10 points; the row action leads to mapping.
Where to start: Directories → Operations → Feed quality → “AI readiness”.
What “AI readiness” can do
- Catalog score and statuses — how many products are ready, partly ready and not ready.
- Eight criteria — for each one you see how many products fell short of full points.
- Category gain — how much the catalog score grows if you fill the category in.
- Jump to products — a click on a tile or chip opens “Products” with only the problem items.
- Category actions — AI description writing and mapping to Google categories right from the row.
What the tab shows
The “AI readiness” tab is the second one in the “Feed quality” window. At the top there is a line with the calculation date, “Calculated: …”, and the “How it is scored” and “Recalculate” buttons. The result is kept for an hour; “Recalculate” scores everything again right away — for example, after you've filled in products.
1
Catalog score2
Ready / partly / not ready3
Criteria tiles4
What to fill in5
Gain6
Row actions7
How it is scored- The average score of all active products, from 0 to 100.
- How many products have each status; click to open them in Products.
- How many products fall short of the full score on a criterion; hover for the scoring rule.
- Up to three criteria where the category loses most; click to open exactly those products.
- How many points the catalog score gains if the category is filled in completely.
- Show products that are not ready, AI description generation, Google mapping.
- The table of criteria and their points.
The first four tiles are “Catalog score” (the average score of all active products) and three statuses: “Ready”, “Partly”, “Not ready”. Then comes one tile per criterion: its number is how many products fell short of full points on that criterion; partial points also count as falling short. The hover tooltip reminds you of the scoring rule.
The table below lists categories: how many products they have, the average score and the ready / partly / not split as three colored labels. Next come the “What to fill in” chips and the gain (more on them below).
How a product score is calculated
A product score is the sum of points for eight criteria, 100 at most. The barcode weighs the most: Google and OpenAI use it to recognize the product. Description, attributes and photos are what an agent uses to answer a shopper's questions.
| Criterion | Full points | Partial | 0 points |
|---|---|---|---|
| GTIN barcode | 20 — a valid GTIN, EAN or UPC and no invalid ones | 10 — no barcode, but there is an MPN | an invalid barcode (even with an MPN) or nothing |
| Brand | 10 — manufacturer filled in | — | empty |
| Photos | 15 — 3 photos or more | 8 — 1–2 photos | no photos |
| Price | 10 — above zero | — | zero |
| Availability | 5 — stock status set | — | no status |
| Description | 15 — 500+ characters of text without tags | 8 — 150–499 characters | shorter than 150 or a copy of the title |
| Attributes | 15 | the category has required ones — share filled × 15; none required — 8 for 1–4 attributes, 15 for five or more | none; if there are required ones — none of them filled |
| Google category | 10 — the main category is mapped to a Google category | — | not mapped |
The full description is used; if it's empty, the short one. The barcode is checked the same way as in the “Feed errors” report: length, digits only, check digit, internal store numbers.
A product's status depends on more than the score:
- Ready — a score of 80 or more, with a price and a photo;
- Not ready — a score below 50, or no price, or no photo: without them OpenAI and Google won't accept the product at all;
- Partly — everything else.
Where to start: gain and “What to fill in”
The “Gain” column shows how many points the score of the whole catalog grows if the category is filled in completely. A large category with an average score often gives more than a small one with a poor score, so the table is sorted by gain: start from the first row. You can also sort by “Products” and “Score” — click the header.
The “What to fill in” chips in a row are up to three criteria where the category loses the most points, with the number of products. For example, “Description 20,586” means that 20,586 products in the category have a description shorter than 500 characters. A click on the chip opens exactly these products.
How to open the products with a problem
The “Products” window opens with a filtered set from:
- a status or criterion tile — products of the whole catalog with that status or without full points on the criterion;
- a chip in a category row — the same products, but only from that category;
- the row action “Show products that are not ready yet” — all products of the category except the ready ones.
1
Showing selected- How many products are in the selection; the cross clears it and brings back the whole catalog.
Above the “Products” table there is a “Showing selected: N” chip; its cross brings back the whole catalog. The filtered set holds at most 5,000 products — those with the lowest scores. When there are 5,000 or more, you'll see the message “Showing the first 5000 products — the lowest scores”: fill them in and open the set again — it is built from current data, no “Recalculate” needed.
If the “Products” window is already open, close it before clicking: an open window only comes to the front and doesn't show the new set.
Some bulk operations from this window run only on the products shown: “Product content search”, “Automatic photo and video creation with AI”, “Add to price monitoring” and export to an XLSX file. The filtered set is personal: other employees don't see it.
“Bulk text generation using AI” and “Translate into other languages” from the “AI & content” menu don't see the filtered set: they run on the category selected in the tree, or on the whole catalog without one. Start descriptions for a category with the row action “AI description writing for the category”.
How to fill in a category
On the right of a category row there are action buttons. You see only those that make sense for that category:
- “Show products that are not ready yet” — always there;
- “AI description writing for the category” — when the category has products without full points for the description. It opens the AI text generation window already set to this category; you start it there after checking the settings (how to start generation);
- “Map categories to Google” — when the category isn't mapped. It opens “Marketplace category and attribute matching”; there, AI matching maps hundreds of categories.
After your edits, click “Recalculate”: without it the tab shows a result saved up to an hour ago.
What the score doesn't take into account
The score is based on the product card data in the catalog's main language, not on what goes into a particular feed. Feed rules, template macros and translations don't affect the score. What a particular marketplace will reject is shown on the “Feed errors” tab.
A product counts only in its main category. The number of filled attributes comes from the card completeness recalculation; after a bulk attribute import it may lag until completeness is recalculated. The score doesn't check the link to the product page.
The score isn't the same as the “Product card completeness” bar in the card and the grid: that bar measures how complete the fields are according to your column settings, while this score measures AI feed readiness by fixed criteria.
Scenarios
Check whether the catalog is ready for ChatGPT
Open “Feed quality” → “AI readiness” and look at the “Catalog score” and the three statuses. “How it is scored” reminds you of the rules.
- The “Not ready” tile opens products without a price or photo, or with a score below 50 — the OpenAI feed won't accept them.
Improve descriptions in the largest category
In the first row of the table (the largest gain), click the “Description” chip or the “AI description writing for the category” action.
- The chip opens “Products” with the products whose description is shorter than 500 characters — for manual editing.
- The action opens AI text generation for this category; after generation, click “Recalculate”.
A category isn't mapped to Google
Click “Map categories to Google” in the row and map the category manually or with AI matching.
- After “Recalculate”, every product in the category gets +10 points for “Google category”.
FAQ
I filled in products, but the score didn't change. Why?
The result is kept for an hour. Click “Recalculate” — the score is calculated from current data right away.
Why does “Products” show only 5,000 products?
The filtered set is limited to the 5,000 lowest-scoring products. Fill them in and open the set again — the next ones will come up.
Do description translations and feed rules count?
No. The score is based on the card in the catalog's main language. What reaches a particular marketplace is shown on the “Feed errors” tab (limits).
The “Show products” buttons don't work — what's wrong?
The filtered set also needs the right to the “Products” window. Without it you can see the report but can't open the products.
Can I get the score through the API?
Yes, API v2 has an ai_readiness resource — category scores, read-only, with access through the “Feed quality” window.
Where to find it and who has access
- Menu
- Directories → Operations → Feed quality, “AI readiness” tab
- Access right
- the “Feed quality” window; to open products — also the “Products” window
- Prepare
- active products in the catalog
Related topics
- If you need to know what a particular marketplace will reject — Feed quality: which products a marketplace won't accept.
- If there are hundreds of categories without a Google category — How to match marketplace categories with AI.
- If you're preparing a feed for ChatGPT ads — How to export a product feed for OpenAI.
- If you need to generate descriptions in bulk — How to run AI text generation.
