Free checklist

Product Page Checklist for AI Readiness: 46 Checks

The 46 checks we run on product pages, grouped by the four dimensions of the AI Readiness Score, with a scorecard you can copy.

Get your free AI Readiness Audit

By Alberto Barberis, Founder and CEO of AndromedAI · Updated October 2026

The short answer

A product page checklist for AI readiness tests whether a product detail page gives AI shopping agents and search engines every fact they need to recommend it: complete attributes, the words shoppers use, answers to buying questions, specific metadata, and structured data and feeds that match the page. Score each check pass, partial or fail, then fix failures at the data source.

Product page checklist in 30 seconds

Use this product page checklist to confirm that a page states every fact a shopper might filter on, in readable text, and says the same thing on the page, in the markup and in the feed. That's what makes a product page ready for AI shopping agents.

01
Agents shortlist products by reading text. A fact that only lives in a photo or a size chart image doesn't exist for them.
02
The checks follow the four dimensions of the AI Readiness Score: Product Data Completeness, Keyword Coverage, Customer Intent Match and Shopping Metadata. A fifth group covers markup and feeds.
03
Most failing product pages fail on missing attributes, so run the completeness checks before the copywriting checks.
04
Audit by template and data field. One empty PIM field fails the same check on every SKU that uses it.
05
Score each check pass, partial or fail on 20 pages, fix what touches the most revenue, then re-score.

How to run this product page checklist on your catalog

Pick 20 product pages, score every check, and count failures per check across the sample. A check that fails on 15 of 20 pages is a template or data problem, and that's where you start.

66%

average machine-readability score of individual product pages, the lowest of all retail page types Adobe measured

A third of a typical PDP is hard for language models to readSource: Adobe Digital Insights

83%

of apparel sites don't provide sufficient sizing information

The most common completeness gap in fashionSource: Baymard Institute

20%

more clicks on average for retailers who added correct GTINs

One identifier, measurable liftSource: Google Merchant Center Help

1
Choose the sample
Five best sellers, ten mid-tail products from your two biggest categories, five recent launches.
2
Open three views per product
The page with JavaScript off, the Product markup in the Rich Results Test, and the item in Merchant Center. Agents read all of them.
3
Score each check
Partial means the fact exists in the wrong place, for example material in the description but missing from the material attribute.
4
Fix at the source
Sort by number of fails. Fix the PIM field, template or feed rule behind each systemic failure, then re-score the same 20 pages.

When we audit a catalog, launches are usually the weakest pages. They often go live straight from a supplier spreadsheet.

Store-wide crawling and indexing belong in the ecommerce SEO audit. The reasoning behind each check is in the product page optimization guide.

Product Data Completeness: 12 checks

Completeness asks whether every attribute a shopper could filter on is stated in text and in structured fields. Agents punish gaps here hardest: a product that doesn't confirm a constraint gets dropped from the shortlist.

C1 Product type is explicit
The page says what the product is in plain words, and product_type is 2 to 3 levels deep, broad to specific.
C2 Material and composition
Exact composition with percentages ("90% merino wool, 10% polyamide"), in text and in material.
C3 Dimensions and weight
Product dimensions with units, not packaging.
C4 Size and fit
Size range and fit notes, with the size chart in HTML text, never only as an image.
C5 Readable color
A color name shoppers use ("sage green"). "Colour 04" fails.
C6 Use-critical specs
The 3 to 6 specs that decide the purchase in this category: waterproof rating, capacity, wattage, compatibility, SPF.
C7 GTIN per variant
Each size and color has its own valid GTIN, or brand plus MPN. One GTIN shared across variants fails.
C8 Variants in sync
Selecting a variant updates image and price, and each variant has its own preselected URL.
C9 Care instructions
Washing or maintenance instructions in text.
C10 What's in the box
Included items listed, plus anything sold separately.
C11 Certifications
Certifications named with the certifying body.
C12 Same facts everywhere
Page, markup and feed carry identical prices and attributes.

The data often exists and never reaches the page. On Shopify, custom metafields such as fit or fabric weight sit in the admin while the theme ignores them, and they won't reach Merchant Center unless an app or feed rule maps them.

Most fashion sites we look at publish size charts as a JPG. For an agent asked about "true to size", an image of a table says nothing.

Google's product variants documentation expects each variant to have a distinct URL that preselects it with the right image and price. Our GTIN guide covers products with no barcode.

With only a week, fix C2, C4 and C6 in your top categories. They cover most of the constraints shoppers type into AI prompts.

Keyword Coverage: 8 checks

Keyword coverage asks whether the page uses the words shoppers actually type and say. Brand language rarely matches demand, so pages with beautiful copy often score low here.

K1 Head term in the title
The title contains the generic name shoppers search ("linen shirt"), beyond the model name.
K2 Modifiers with demand
The modifiers people search for (gender, material, fit, use) appear early.
K3 Shopper vocabulary
Market terms sit next to branded names: "puffer jacket" alongside "the Brava down shell".
K4 Synonyms once
Common variants appear naturally, such as "sneakers" and "trainers" on a store selling to the US and UK.
K5 Long-tail values written out
Values from long queries are spelled out: "210 cm", "16-inch laptop".
K6 Keywords beyond the title
Terms also appear in bullets and FAQ answers.
K7 Local-language keywords
Each market uses its own researched search terms. A literal translation of the English keyword fails.
K8 No stuffing
Each term appears where it reads naturally.

We'd skip keyword density targets. We've never seen a product page win because a phrase appeared six times instead of three. Google's Merchant Center title guidance points the same way: put the decisive details first, since users usually notice only the first 70 or so characters.

K7 is where multi-market catalogs fail quietly. A translated title can be grammatically perfect and use a word nobody searches in that country. Our guide on how to write product descriptions shows how to work keywords into copy naturally, and the product description templates by category give each category a structure with room for them.

Customer Intent Match: 9 checks

Intent match asks whether the page answers the questions behind the purchase: who it's for, when to use it, what problem it solves and how it compares. That's what lets an agent justify a recommendation.

I1 Who it's for
The page names the buyer ("for first-time marathon runners").
I2 When and where
Occasions, seasons or conditions are stated: "muddy day hikes", "office to dinner".
I3 Problem it solves
One or two sentences in the shopper's own words.
I4 Who it's not for
Honest limits. Agents use them to avoid bad matches.
I5 Comparison anchor
How it differs from your nearest alternative, so an agent can choose between them.
I6 Product FAQ in visible text
4 to 8 real questions, such as sizing and care.
I7 Benefits backed by specs
"Keeps feet dry" sits next to "waterproof membrane, taped seams".
I8 Reviews in HTML
Review count and rating visible, with fit feedback for apparel.
I9 Returns near the buy box
Return window and delivery estimate on the page as well as the footer.

AI assistants split a prompt into constraints. Google calls this query fan-out, issuing multiple related searches across subtopics and data sources. OpenAI says a product appears in ChatGPT's carousel when ChatGPT perceives it to be relevant to your intent.

ChatGPT

Demo illustrativa

ChatGPT

Illustrative example

The cheaper boot lost on one missing fact, width, which fails C4 and I1 together.

On I6 we disagree with a lot of current advice. Google stopped showing FAQ rich results in 2026 (Google Search Central), and many teams dropped product FAQs. Yet AI systems still read those answers, and Merchant Center accepts up to 30 question and answer pairs per product. Baymard reports that 70% of sites get FAQs and Q&As wrong, so the bar is low. More in FAQ schema on product pages.

Shopping Metadata: 8 checks

Shopping metadata covers titles, headings, meta descriptions and URLs. Search engines and agents read these fields first, and they're the ones most often left on a template default.

M1 Unique page title
Each product has its own <title> built from product type, a key attribute and brand.
M2 H1 names the product
The first visible heading says what the product is, not a campaign slogan.
M3 Specific meta description
Real attributes plus one reason to buy, roughly 140 to 160 characters.
M4 Front-loaded feed title
The Merchant Center title puts decisive details in the first 70 characters and stays within 150.
M5 Readable, stable URL
A slug with the product type that survives renames.
M6 Titles agree
The feed title, page title and H1 describe the product with the same key terms.
M7 Descriptive alt text
Alt text names the product and its color.
M8 Market match
Title and meta description use the page's language with local units.

Google builds title links from several sources, including the <title> element, the main visible heading and og:title (Google Search Central). If your H1 says "Summer Essentials" and your title says "Linen Shirt", you've handed Google two competing answers.

Merchant Center may also show a customized dynamic title built from your landing page. If Shopping displays titles you never wrote, treat that as a hint: yours didn't match the query well enough.

On M3 we'd take a template that pulls clean fields over thousands of hand-written meta descriptions nobody maintains.

Structured data and feeds: 9 checks

This group checks the machine-readable layer: Product markup plus the Merchant Center and ChatGPT feeds. It's where a good page gets confirmed or contradicted.

T1 Valid Product markup
Product with name, image and offers, passing the Rich Results Test.
T2 Recommended properties
brand, gtin, sku, color, material, size and aggregateRating where they apply.
T3 Markup matches the page
Every value in the markup is visible on the page.
T4 Variants marked up
ProductGroup with variesBy and hasVariant, matching the variant URLs.
T5 Shipping and returns
shippingDetails and hasMerchantReturnPolicy, or the same policies set at organization level.
T6 Product highlights
4 to 6 product_highlight values per item, under 150 characters each, benefits only.
T7 Conversational attributes
Question and answer pairs sent wherever your catalog has them.
T8 Content in the initial HTML
Specs and FAQ load with the page, not fetched after a click.
T9 Fresh feed
Price and availability update in the feed when they change on the site.

Google's merchant listing documentation recommends material, color, size, gtin and pattern on Product, the same facts agents filter on. Google also says there's no special schema.org structured data needed for AI Overviews or AI Mode, and recommends markup that matches the visible text. So T3 beats adding exotic properties.

Feeds now carry more than the page does. Google introduced conversational attributes in Merchant Center at Google Marketing Live 2026, and they are rolling out globally. OpenAI's product feed spec caps title at 150 characters and description at 5,000. For T6, Google's product highlight rules ban keywords and prices, exactly what most teams try to put there first.

T8 trips up headless and script-heavy themes. Turn JavaScript off and reload: if the specs tab is empty, many crawlers see it empty too. For every feed requirement, use the Google Merchant Center checklist; for markup examples, the structured data guide.

Product page audit scorecard and what to fix first

Copy the scorecard into a spreadsheet with one row per check and one column per sampled page. Rank failures by pages and revenue affected, then fix the systemic ones at the source.

Check	Dimension	Pass	Partial	Fail
C1 to C12	Product Data Completeness	Fact in text and in the structured field	Fact in one place only	Missing or only in an image
K1 to K8	Keyword Coverage	Shopper terms throughout the copy	Terms in title only	Brand vocabulary only
I1 to I9	Customer Intent Match	Answered in visible text	Answered vaguely or only in reviews	Not answered
M1 to M8	Shopping Metadata	Unique and specific	Templated but accurate	Default or duplicated
T1 to T9	Structured data and feeds	Valid and matches the page	Present with warnings	Missing or contradicts the page

Score pass as 2, partial as 1, fail as 0. A page under half the maximum on completeness needs data work before any copywriting.

Priority	Typical failure	Where to fix
1	Page and feed contradict each other (C12, T3)	Feed rules and templates, this sprint
2	Missing decisive attributes (C2, C4, C6)	PIM or product admin, top categories first
3	No audience, use cases or FAQ (I1, I2, I6)	Description and FAQ fields, by revenue tier
4	Templated titles and empty meta descriptions (M1, M3, M4)	Title and meta templates
5	Highlights and conversational attributes (T6, T7)	Merchant Center feed

Common advice says start with titles because they're quick. We'd start with contradictions: a price mismatch can get an item disapproved in Merchant Center.

AndromedAI / OptimizerAI Readiness 38/100

title

Before~~Corvo Trail Boot Brown~~

AfterCorvo Ridge Women's Waterproof Hiking Boot, Full-Grain Leather, Wide Fit Available

description

Before~~A tough boot built for every adventure.~~

AfterFull-grain leather hiking boot with a waterproof membrane and a deep-lug outsole for muddy day hikes. Regular and wide fit. Runs true to size.

attributes

Before~~color: brown~~

Aftercolor: chestnut brown; material: full-grain leather; waterproof: yes; width: regular, wide; weight: 540 g per boot

faq

Before~~none~~

AfterGood for wide feet? Yes, choose the wide fit. For winter mountaineering? No, they're built for three-season day hikes.

Illustrative example, AI Readiness Score from 38 to 89

The after version passes ten more checks, and every added fact came from the brand's existing product sheet.

Product page checklist with AndromedAI

AndromedAI runs these checks across the whole catalog instead of a 20-page sample, fixes the failing fields and publishes them back to your store and feed.

The four dimensions here are the four in the AI Readiness Score, which is how the platform applies Agentic Commerce Optimization to every SKU.

Checklist group	AndromedAI	What it does
Score your pages	Free AI Readiness Audit	Scores product pages on the four dimensions and shows what's missing
Completeness, keywords, intent	Optimizer	Extracts attributes, rewrites titles, copy and FAQ, adds use cases and intents within your Brand Kit
Missing pages and markets	Creator	Creates complete product pages from brand or supplier data in 12 languages
Feed and publishing	Integrations	Publishes to Shopify, Google Merchant Center and other platforms, and generates Merchant Center conversational attributes

Changes pass AI checks and approval workflows, with optional auto-approval above an AI Checker score of 4.0.

+1,800%

clicks from AI chats

Bomboogie

+80%

add-to-cart in two months

Altaforma Milano

-33%

bounce rate

Ausilium

FAQ

It should test four things on every page: complete attributes in text and structured fields, the words shoppers actually search, answers to buying questions such as who the product is for, and specific titles and meta descriptions. Add checks for Product markup and Merchant Center feeds, because agents compare the page with both and can drop products whose data disagrees.

Start with 20 pages from your top categories: five best sellers, ten mid-tail products and five recent launches. Failures repeat across templates and data fields, so a small sample shows most systemic problems. Once you fix those at the source, re-score the same 20 pages before widening the sample.

Product selection relies mainly on text and structured data, so a fact that only appears in an image is easy to miss. Put sizes, specs and size charts in HTML text and in feed attributes, never only in a photo or a JPG table. Alt text should also name the product and its color.

Google stopped showing FAQ rich results in 2026, so the markup no longer earns extra space in search results. The questions and answers are still worth publishing as visible text, because AI systems read page content and Merchant Center accepts question and answer pairs as product attributes. Aim for 4 to 8 real questions per product.

Re-run it every quarter, and again after any template change or large catalog import, since both can break a field across hundreds of pages at once. Check new launches before they go live, because pages built straight from supplier spreadsheets tend to fail the most completeness checks.

Glossary

AI Readiness Score: AndromedAI's score for how ready a product page is for AI agents, built on four dimensions of product data.
Product Data Completeness: The share of relevant attributes stated on the page and in structured fields.
Customer Intent Match: How well a page answers the questions behind a purchase, such as who it is for and when to use it.
Query fan-out: The way AI search splits one shopper prompt into many related searches.
product_highlight: A Merchant Center attribute for short benefit statements, 4 to 6 recommended per product.
Conversational attributes: Merchant Center product data, such as question and answer pairs, built for experiences like AI Mode.

Keep reading

The strategy behind every check on this list

The store-wide audit for crawling, indexing and facets

Every feed requirement, check by check

A ready copy structure for each product category

Sources (14)

Parlano di noi

Choose the sample

Open three views per product

Score each check

Fix at the source

What should a product page checklist include for AI readiness?

How many product pages should I audit?

Do AI shopping agents read product images?

Is FAQ schema still worth adding to product pages?

How often should I re-run the checklist?

Keep reading