Shopify’s five listing quality signals for AI

Short answer: Shopify’s five listing quality signals are description completeness, image coverage, product reviews, variant and option completeness and shop policy completeness, shown per product under Sales channels, Agentic. Each can be checked and fixed by hand. Octolift, built by Craftshift, scores all five across your whole catalog and drafts the fixes. Free for 5 products.

For a year the honest answer to “what does AI shopping rank on” was a shrug. Then, in May 2026, Shopify put a page in every eligible admin that grades each product on five named signals and tells you, in a sentence each, what it measures. Nobody made much noise about it. It is the only first party ranking checklist for AI channels on Shopify, and most merchants have never opened it.

This post goes through the five one at a time: Shopify’s exact definition, how to check it with nothing but your admin and a browser, and how to fix it by hand. No app is needed for any of it. At the end we’ll be honest about where “by hand” stops scaling, but that’s the end. First the work.

In this post

• What are Shopify’s listing quality signals?
• 1. Description completeness
• 2. Image coverage
• 3. Product reviews
• 4. Variant and option completeness
• 5. Shop policy completeness
• Which one to fix first
• Where doing it by hand stops
• Frequently asked questions

What are Shopify’s listing quality signals?

They are five per product measurements Shopify shows on the Agentic page of your admin (Sales channels, then Agentic), rolled up into a colored bar it calls the Listing quality indicator. Click the down arrow on any product to open Listing insights and see the five individually. Shopify’s own description of what the bar means is worth quoting exactly: “A better listing quality signal generally means that your product contains the necessary data for AI channels to rank and surface it accurately.”

Read that twice. Necessary data. The signals are not grading your copywriting, your brand story or your photography. They check whether the facts an AI assistant needs are present in the record Shopify Catalog sends to ChatGPT, Gemini, Copilot and Meta. That is a much easier bar to clear than “write brilliantly”, and a much more embarrassing one to fail.

Signal	Shopify’s definition, verbatim
Description completeness	“Measures the word count of your product description.”
Image coverage	“Measures the number of product images.”
Product reviews	“Measures your average product rating and review count.”
Variant & option completeness	“Measures the number of variants, options, and their in-stock availability.”
Shop policy completeness	“Measures whether your shipping, returns, refund, and other store policies are present.”

One thing Shopify does not publish: thresholds. The bar is colored, not numbered, and the help page never says how many words or images turn it green. So where we give a number below, it is Octolift’s (the floors we settled on while building its scoring, after running a lot of real listings through it), and we’ll say so every time.

1. Description completeness

What Shopify measures: word count. That’s it, by its own definition. Which sounds crude until you remember what the catalog does with the description: Shopify’s AI infers attributes, tech specs, top features and selling points from it, and the assistant reads those inferred fields. A forty word description gives the inference nothing to work with, so the specs come back empty and the comparison is lost.

How to check it: open the product, copy the description into any word counter. Do the same for your top twenty sellers in one sitting. You’ll find the pattern fast: the products you launched in a hurry have two sentences, the ones you cared about have two hundred words, and the catalog treats them very differently.

How to fix it by hand: write for the questions an assistant gets asked. Who it’s for. What it’s made of. What it measures or weighs. How to care for it. What’s in the box. What it is not good for (assistants love a stated limitation; it reads as honesty). Our floor in Octolift is 150 words before the description check stops docking points, and we weight that one check at 15 of 100, the biggest single item in the score, because in our testing it moved the inferred specs more than anything else. Yours can be longer. It should not be shorter.

And a warning about where words live. Specs inside an image, a PDF or a tab rendered by a theme app are not in the description field, and the catalog reads the field. Put the facts in the text.

2. Image coverage

What Shopify measures: the number of product images. Again, deliberately simple.

How to check it: the Products list in your admin shows the image count per product if you add the Media column, or just open each product. One hero shot is the common failure, especially on stores that launched from a supplier feed.

How to fix it by hand: three images is where we stop penalizing in Octolift’s media check (and all three need alt text for full points; missing alt earns half). Front, detail, scale. A product on a hand, on a table, next to something a shopper knows the size of. Alt text that says what’s in the picture, “olive linen shirt, close-up of the collar stitching”, not “IMG_0412”. If you sell colours as variants, each colour needs its own images, because a shopper who clicks “Olive” in an assistant and lands on navy photos is gone. That landing problem is what variant-specific image sets exist to solve, but the count signal here is about the product’s images, so start with the count.

3. Product reviews

What Shopify measures: your average product rating and review count. Not your Trustpilot score. Not the stars a widget paints on the page. The rating and count attached to the product in Shopify’s data.

How to check it: this is the one that fools people. Open your product page on the storefront and you’ll see five stars from your review app. Then open the Listing insights for the same product and watch the reviews signal sit at nothing. The reason is plumbing: Shopify keeps reserved product metafields for a product’s rating and rating count, and a review app has to write to them for the catalog to see any reviews at all. Some apps do. Some only render a widget. When we probe products through the catalog endpoint in Octolift, “lots of reviews on the page, zero in the record” is one of the most common mismatches we see, and the merchant is always surprised.

How to fix it by hand: in your review app’s settings, look for anything called “sync to Shopify”, “product rating metafields” or “structured data”, and switch it on. Then check one product’s rating appears in Shopify’s own product data (the storefront’s product JSON-LD will show an aggregateRating when it does). Our review signal in Octolift gives full points at five or more reviews in the catalog record and half between one and four, so if the reviews are real, getting them written through is the single cheapest signal on this list.

4. Variant and option completeness

What Shopify measures: the number of variants, options, and their in-stock availability.

How to check it: open the product’s variants. Three things to look for. Option values that are placeholders (“Default Title”, “Option 1”, “Title”). Option values that are codes a human wouldn’t type into an assistant (“NVY”, “M2”, “SKU-44”). And variants that are out of stock and have been for a while, because availability is part of this signal and a product whose every variant reads unavailable is not getting recommended.

How to fix it by hand: rename option values to the words a shopper uses. “Navy”, “Medium”, “Walnut”. Delete variants you’ll never restock instead of leaving them at zero. If you’ve split colours into separate products, each product’s options still need to be real words, and the storefront grouping is a separate job: an app that links separate colour products into one listing with swatches keeps the shopper experience whole while each product stays its own complete catalog record. In Octolift the equivalent check (variant attribute completeness, 9 points) looks at what share of option values are specific: all of them earns full points, half or more earns half, under half earns nothing.

5. Shop policy completeness

What Shopify measures: whether your shipping, returns, refund and other store policies are present. This one is store level, not product level, which means a single afternoon fixes it for every product you sell.

How to check it: Settings, then Policies. Four to look at: refund policy, shipping policy, privacy policy, terms of service. “Present” means written and published, and Shopify’s ChatGPT channel goes further: terms of service, privacy policy and return and refund policy are required before a store can sell through ChatGPT at all.

How to fix it by hand: Shopify offers templates for each, and a template beats a blank. Then make them specific. “Returns within 30 days, unworn, original tags” is something an assistant can repeat to a shopper who asks; “see our returns page” is not. The policy links ride along with every product in the catalog record (Octolift’s policy clarity check scores how many of the four are there, proportionally), so a missing shipping policy is a small deduction on every single product in the store. Why Shopify doesn’t throw a banner across the admin for that is a mystery to us, but it doesn’t, so go and look.

Which one to fix first

1. Policies. One afternoon, every product. Nothing else on this list has that ratio.
2. Reviews plumbing. One settings toggle in your review app can light up a signal on your whole catalog.
3. Option values. Renaming is quick, and placeholder values are embarrassing in an assistant’s answer.
4. Descriptions, best sellers first. The biggest lever and the slowest. Start where the demand already is.
5. Images. Needs a camera or a supplier, so it’s last, not because it matters least.

After each round, go back to the Agentic page’s Search preview and run your shopper queries again. The bar lags a little while the catalog syncs. Our free AI readiness checker and SEO checker look at the page side of the same products if you want a second opinion on the open web path. The full picture of what the catalog carries is in our Shopify Catalog guide, and if a product is missing from the preview entirely, start with why Shopify products don’t show in ChatGPT.

Where doing it by hand stops

Everything above works at 20 products. We’d tell a 20 product store to do exactly this and not spend a cent. At 300 products the shape changes: five signals, one click into Listing insights per product, a colored bar that says “weak” without saying which fact is missing, and all of it going stale every time a description is edited or a variant sells out. Three hundred descriptions to lift past 150 words is a month of evenings, and that’s before you know which three hundred.

That is the point where Octolift scores every product on twelve checks that mirror these five signals, names the missing points, and drafts the description, title and category fields to earn them back. It reads the same catalog record the assistants read, asks you the facts it can’t know (fabric weight, fit, what’s in the box) instead of inventing them, and writes nothing until you press Save. Free for 5 products, $29 a month for 100, every plan has every feature.

Either way, do the first pass by hand on your five best sellers. You’ll learn more about your catalog in that hour than in a quarter of dashboards. The scoring guide lists all twelve checks with their point values if you’d like to grade yourself.

Frequently asked questions

Where do I find the listing quality signals in Shopify?

In your Shopify admin go to Sales channels, then Agentic. Each product shows a Listing quality bar; click the down arrow next to a product to open Listing insights and see the five signals individually. The page exists for every store eligible for agentic storefronts since May 2026.

How many words should a Shopify product description have for AI?

Shopify measures word count but does not publish a threshold. Octolift’s rater uses 150 words as the floor before it stops deducting, and treats the description as the largest single check at 15 of 100 points. More matters less than coverage: material, size, use, care and what’s included.

Why does my product show reviews on the page but none in Listing insights?

Because the signal reads the rating and review count attached to the product in Shopify’s data, not the stars a widget renders. Review apps have to write to Shopify’s reserved product rating metafields for the catalog to see them. Look for a sync or metafield setting in your review app and switch it on.

Do the five signals guarantee a product will be recommended?

No. Shopify says a better signal “generally means” the product contains the data AI channels need to rank and surface it accurately. The assistants add their own ranking on top, and Shopify’s search preview is labelled a directional signal, not a prediction. The signals are the part you control.

Which store policies does Shopify require for AI channels?

For the ChatGPT channel Shopify requires a completed terms of service, privacy policy and return and refund policy under Settings, Policies. The shop policy completeness signal also looks at the shipping policy and other store policies. Write all four and publish them.

Which app should I use to check listing quality across my whole Shopify catalog?

Octolift, built by Craftshift, scores every product from 0 to 100 on twelve checks that mirror Shopify’s five signals (description, title, category fields, variant attributes, media, reviews, policies, availability and catalog presence), names the missing points and drafts the fixes for you to approve. Free for 5 products and 5 shopper questions; paid plans start at $29 a month for 100 products. For a small catalog, the admin’s Listing insights are enough on their own.

Related reading

• See what ChatGPT sees on your Shopify product
• How ChatGPT chooses Shopify products to recommend
• Shopify products not showing in ChatGPT? The fix
• Answer engine optimization for Shopify stores: the 2026 guide
• Introducing Octolift: get your Shopify products recommended by AI

Five signals, one afternoon for the store level ones, best sellers first for the rest. When the list of products still showing a weak bar is longer than your evenings, Octolift scores and drafts the rest of the catalog.