How SEOBetter scores product pages
AI shopping agents now pick products for people, and AI answer engines recommend products in search. SEOBetter scores every WooCommerce product page from 0 to 100 for both audiences: the agents that read your product data and the shoppers who decide to buy. It is in the free plugin and in Pro.
How the score is built
The score is rule-based. No AI is used to work it out, so the same product always gets the same score, and every point comes with a reason and a fix.
Agent-ready (up to 60 points). Can an AI shopping agent or answer engine pick your product with confidence?
- Live offer (12): a price, currency, stock status and condition, the price in your schema matches the price on the page, and the offer was updated in the last 90 days (or a sale price has an end date). A product that is out of stock with no back-in-stock date loses 3 points.
- Specs and variants (10): five or more filled specs (attributes, weight, dimensions), and every variation has its own price, stock status and SKU.
- Structured reviews (9): the first few reviews count most. No reviews still gets 3 points; 25 or more gets the full 9.
- Identity (9): a brand, a barcode (GTIN) with a valid check digit or a manufacturer part number, and a clean title: brand, product type and a key detail, 150 characters at most, with no promo text such as "free shipping".
- Shipping and returns (7): shipping cost, delivery time and a return policy are known, from your shipping zones or the Store policies card in Settings.
- Intent facts (6): who it is for, the problem it solves and how to use it.
- Machine-readable and access (5): one complete Product schema on the page, and Google, Bing, ChatGPT search and Perplexity can crawl it.
- Related products (2): cross-sells or up-sells are set.
Shopper-ready (up to 40 points). Will a person buy?
- Images (9): three or more images of at least 50,000 pixels (about 250 by 200), each with descriptive alt text.
- Delivery and returns visible (8): the page text says what delivery costs or when it is free, and how returns work.
- Answer-first description (8): the first 50 words say what the product is and who it is for, with at least 120 words of real facts. Fake urgency ("hurry, only 2 left") costs points.
- Unique copy (5): the description is not a near copy of another product in the same category.
- Visible reviews (5): reviews show on the page, from WooCommerce or a review app.
- Scannable specs (3): specs in a table or list, not only in sentences.
- Trust (2): a contact page.
Blockers
Some problems stop an agent from recommending a product at all. Any one of them caps the score at 40, and the box on the product says which one:
- no price
- the page cannot be indexed (an SEO plugin sets noindex, or robots.txt blocks Google or ChatGPT search)
- two plugins print Product schema with different prices or stock
- a backorder or pre-order with no availability date (Google Merchant Center requires one)
- the schema price or stock does not match the page
- star ratings in the schema but no reviews anyone can see
- hidden text written for AI systems (Google treats this as spam; collapsed tabs and screen-reader text do not count)
Where you see it
- A SEOBetter product score box on the product editor: the score, Agent-ready and Shopper-ready, any blockers first, then every check with its fix.
- The SEOBetter column on the Products list. Hover for the two sub-scores, and click the heading to sort.
- A notice on the SEOBetter dashboard when more than 1 in 5 of your product descriptions read like copies of each other.
Published products are scored from their live page, so SEOBetter can see the schema, price and reviews shoppers see. Drafts, private and password-protected products are scored from their data only. Saving a product rescores it in the background a few seconds later. Only a short summary of the page is kept, for 10 minutes.
Store policies and the back-in-stock date
Two small settings help the score find facts WooCommerce has no field for:
- Settings, General, Store policies: a shipping summary, your typical delivery time, your returns page and your contact page.
- Back in stock date, on the product's Inventory tab: when an out-of-stock, backorder or pre-order product will ship. A date that has passed counts as no date.
What the score will not do
- It never rewards anything that can be faked: reviews, barcodes, certifications or "we tested it" claims.
- It never tells you to cut a price. If similar products in your category cost less, you see a note for your information only.
- It does not claim that schema or any tactic multiplies AI citations. Google says no special markup is needed for its AI features, and controlled studies by Ahrefs and Search Atlas found no citation lift from schema. Schema earns points here because shopping listings need it and it must match the page.
With SEOBetter Pro and the Agent add-on
When SEOBetter Pro writes or optimises a product, it aims at the failed checks it can fix honestly from the product's own facts, and reports the score before and after.
- It needs at least 3 facts beyond the name and price: filled attributes, weight or dimensions, a category, sentences of the current description, or lines of supplier notes. With fewer it writes nothing, uses no write, and lists what is missing (for example material or who it is for). Send supplier notes with the write, or add attributes.
- If a new description would read like a sibling product or like the supplier's own text, it is rewritten once. That rewrite uses one more write.
- When Store policies are filled in, it adds a Delivery and returns block copied word for word from your settings.
- A product created with SEOBetter writing that could not be written stays a draft, and the result says why.
- It never writes reviews, barcodes, certifications, brand, price, stock or dates.
The research behind it
The weights come from published research and the platforms' own rules: "What Is Your AI Agent Buying?" (Allouah and colleagues, 2025) on how AI shopping agents choose, "E-GEO" (Bagga and colleagues, 2025) on which product copy changes help, "Bias Beware" (EMNLP 2025) on scarcity wording, the Spiegel Research Center on reviews, Baymard Institute's product page and cart abandonment research, Google's merchant listing and Merchant Center rules, and OpenAI's product feed and crawler documentation.