How we rate AI e-commerce tools

Every search tool here understands what a shopper means, handles synonyms, suggests as you type, builds filters, lets you pin or bury products and reports on searches, so none of that earns a point. What separates them is ordering results per shopper and searching by image.

Each section adds up to 10, then counts for 20 of the 100.

How many store platforms it works with, its own Shopify App Store listing, webhooks, and how many apps it connects to.

Each section adds up to 10, then counts for 25 of the 100.

What the cheapest paid plan costs each month, a free plan, a trial, whether there is a setup fee, and what the price grows with. A flat fee you can budget scores best, a fee that rises with orders, sessions or catalogue size scores in the middle, and a share of your revenue or ad spend scores lowest, because it grows with your sales. Where a vendor sells a tier and publishes no figure, the page says so rather than showing a number nobody published.

Each section adds up to 10, then counts for 35 of the 100.

Whether it trains AI on your store's product and order data, SOC 2 and GDPR, and how many ways there are to reach support.

Each section adds up to 10, then counts for 20 of the 100.

Every one of these recommends products, cross-sells, sorts collections, personalises the homepage, runs A/B tests, puts recommendations in email, respects stock and reports the revenue it influenced. What is left: offers after checkout, bundles, and targeting shopper segments.

Each section adds up to 10, then counts for 20 of the 100.

How many store platforms it works with, a Shopify App Store listing, a public API and how many apps it connects to.

Each section adds up to 10, then counts for 25 of the 100.

What the cheapest paid plan costs each month, a free plan, a trial, whether there is a setup fee, and what the price grows with. A flat fee you can budget scores best, a fee that rises with orders, sessions or catalogue size scores in the middle, and a share of your revenue or ad spend scores lowest, because it grows with your sales. Where a vendor sells a tier and publishes no figure, the page says so rather than showing a number nobody published. A required annual contract loses points.

Each section adds up to 10, then counts for 35 of the 100.

Whether it trains AI on your store's data, SOC 2, onboarding included in a plan, and how many ways there are to reach support.

Each section adds up to 10, then counts for 20 of the 100.

All three write in bulk, take a spreadsheet, place keywords, keep a brand voice, write in other languages, make images and publish back to the store. What separates them is copy shaped for a named marketplace such as Amazon, and a screen to review everything before it goes live.

Each section adds up to 10, then counts for 20 of the 100.

How many store platforms and marketplaces it works with, a Shopify App Store listing, a public API and how many apps it connects to.

Each section adds up to 10, then counts for 25 of the 100.

What the cheapest paid plan costs each month, a free plan, a trial and whether a contract term is required. All three price by the number of products.

Each section adds up to 10, then counts for 35 of the 100.

Whether it trains AI on your product data, SOC 2, GDPR, onboarding and how many ways there are to reach support.

Each section adds up to 10, then counts for 20 of the 100.

Every one of these bids for you and builds campaigns. What separates them: how many ad networks it runs, excluding wasted search terms by itself, bidding by hour or day, your own rules, Amazon DSP, making the ad creative, and a tier where the vendor runs the ads for you.

Each section adds up to 10, then counts for 35 of the 100.

How many store platforms and marketplaces it works with, a Shopify App Store listing, a public API and how many apps it connects to.

Each section adds up to 10, then counts for 15 of the 100.

What the cheapest paid plan costs each month, a free plan, a trial, whether there is a setup fee, and what the price grows with. A flat fee you can budget scores best, a fee that rises with orders, sessions or catalogue size scores in the middle, and a share of your revenue or ad spend scores lowest, because it grows with your sales. Where a vendor sells a tier and publishes no figure, the page says so rather than showing a number nobody published. Most of these charge a base fee plus a share of ad spend they do not publish.

Each section adds up to 10, then counts for 30 of the 100.

Whether it trains AI on your account data, GDPR, single sign-on, onboarding and how many ways there are to reach support.

Each section adds up to 10, then counts for 20 of the 100.

What it tells you: which channel sold the order, profit after costs, its own tracking pixel, lifetime value by cohort, your own dashboards, demand forecasts, stock planning, benchmarks against other stores and competitor prices. Every one of them sends alerts and exports the data, so that earns nothing.

Each section adds up to 10, then counts for 35 of the 100.

How many store platforms it works with, a public API, webhooks and how many apps it connects to.

Each section adds up to 10, then counts for 20 of the 100.

What the cheapest paid plan costs each month, a free plan, a trial, whether there is a setup fee, and what the price grows with. A flat fee you can budget scores best, a fee that rises with orders, sessions or catalogue size scores in the middle, and a share of your revenue or ad spend scores lowest, because it grows with your sales. Where a vendor sells a tier and publishes no figure, the page says so rather than showing a number nobody published.

Each section adds up to 10, then counts for 25 of the 100.

Whether it trains AI on your store's data, SOC 2, GDPR, onboarding and how many ways there are to reach support.

Each section adds up to 10, then counts for 20 of the 100.

Every one of these makes a page from a prompt, so that earns nothing. The biggest difference is whether it takes the payment itself or builds pages on someone else's store. Then a drag-and-drop editor, templates, editing the code, A/B tests, an app marketplace, more than one store or region, hosting and finding products to sell.

Each section adds up to 10, then counts for 35 of the 100.

Where it runs (web, phone, desktop), a public API, webhooks, and how many apps it connects to. A platform with its own app marketplace is not scored on integrations as well.

Each section adds up to 10, then counts for 15 of the 100.

What the cheapest paid plan costs each month, a free plan, a trial, whether there is a setup fee, and what the price grows with. A flat fee you can budget scores best, a fee that rises with orders, sessions or catalogue size scores in the middle, and a share of your revenue or ad spend scores lowest, because it grows with your sales. Where a vendor sells a tier and publishes no figure, the page says so rather than showing a number nobody published.

Each section adds up to 10, then counts for 30 of the 100.

SOC 2, GDPR, PCI DSS for the ones that take payment, and how many ways there are to reach support.

Each section adds up to 10, then counts for 20 of the 100.

All three answer from your own catalogue, capture what shoppers tell them, start a conversation by themselves and report the revenue they influenced. What separates them: a chat on the store, a product quiz, order and return questions, handing over to a person, other languages, and how many email and SMS tools it feeds.

Each section adds up to 10, then counts for 35 of the 100.

How many store platforms it works with, webhooks and how many apps it connects to.

Each section adds up to 10, then counts for 15 of the 100.

What the cheapest paid plan costs each month, a free plan, a trial, whether there is a setup fee, and what the price grows with. A flat fee you can budget scores best, a fee that rises with orders, sessions or catalogue size scores in the middle, and a share of your revenue or ad spend scores lowest, because it grows with your sales. Where a vendor sells a tier and publishes no figure, the page says so rather than showing a number nobody published.

Each section adds up to 10, then counts for 30 of the 100.

SOC 2, single sign-on and how many ways there are to reach support.

Each section adds up to 10, then counts for 20 of the 100.