AI Product Finder for Your Online Store

An AI product finder lets shoppers on your online store type what they need, such as "a two-person tent under 4 pounds that I can pitch alone", and get the matching products from your own catalog. If your catalog is large or technical enough that shoppers do not know which product fits them, it is the third option next to category filters and a product quiz. This guide helps you choose between the three.

What is an AI product finder?

An AI product finder for your website lets visitors describe what they need in a chat and returns matching products from your own catalog, searched by price, attributes and availability. The term also covers two other things. Shoppers use consumer finders to search the whole market: Amazon's shopping assistant, launched as Rufus and renamed Alexa for Shopping in May 2026 per About Amazon, answers from Amazon's catalog, customer reviews and the web. Dropshippers use research tools to find products worth selling.

This guide is about the first kind: the finder on your store, which knows only your products and answers your shoppers. It matters most when a catalog has many variants and specs, such as outdoor gear, bike parts, lighting or lab supplies, and the shopper knows the job but not the product name.

Ultimo Bots is the AI agent platform where you describe in plain words what your agent should do and Mo, the builder in the AI Studio, builds it: skills on public APIs and MCP servers that do real work in your own systems. Searching your catalog is built in. The same agent also answers shipping, returns and sizing questions from your pages, on your website, WhatsApp, Messenger and Instagram.

The examples below use Tamarack Trail Outfitters, a fictional online camping and hiking gear shop on WooCommerce. Its owner, Nadia, sells around 1,400 products, and the tents alone come in 60 variants.

Filters, a quiz or an AI agent: which fits your catalog?

Filters serve shoppers who already know what they want, a quiz serves the questions you thought of in advance, and an AI agent serves the shopper who describes a need in their own words. Large catalogs get all three kinds of shoppers, but the third is the one who leaves when nothing fits.

Keep your filters: they work for the shopper who browses a category. The agent sits next to them as the chat on your store and takes the questions filters cannot hold, such as "which of these do I need for my trip?", in the language the shopper writes in.

The pricing row comes from the vendors' own pages, read on November 10, 2026. Doofinder, a site search and product discovery tool, prices its plans by monthly requests, and counts a request each time its server answers, which for search starts at the third keystroke. RevenueHunt, a product recommendation quiz, prices by monthly quiz responses; its paid plans start at $39 a month for 500.

Add your catalog: upload, crawl or sync

The agent needs your products as a catalog, and you can give it one in four ways. Upload a CSV, JSON or TSV export, or let the agent crawl your product pages; both are a snapshot of that day. A WooCommerce store with the Ultimo Bots WordPress plugin 1.3.0 or later keeps published products, prices and stock current a few minutes after you change them. A connected Wix store re-syncs daily on every paid plan; our Wix Stores guide walks through the setup.

In the portal the catalog sits under Products in the knowledge base. Two tips for a file: give it a title or name column, and use a column named "availability" or "stock" for in stock or not, because the agent reads that column as availability rather than a count. When you upload a new file, delete the old catalog, since the agent searches every catalog you keep. If you sell from a shop counter too, our retail guide covers the in-store side.

How the agent finds the right product

Every column in your catalog becomes something the agent can search by: price, weight, size, color, material, availability or any spec you include. It turns the shopper's sentence into filters, sorts the results, and knows the exact number of matches. Fuzzy wishes such as "something for a windy ridge" go to a meaning-based search instead.

A fictional shopper on Nadia's store asks "I need a two-person tent under 4 pounds for less than $300." The agent answers: "I'm Tamarack Trail Outfitters' AI assistant. Five of our two-person tents weigh under 4 pounds and cost less than $300. Here they are, lightest first." Product cards with photo and price follow in the website chat. The shopper adds "Which of those is freestanding?", and the agent narrows the same search to the two that are, without starting over.

• Exact counts, a short list. A search returns up to 15 products at a time. The agent shows about 8 at most, states the total and can ask a question to narrow down when there are more.
• Your catalog, nothing else. It recommends from the products your catalog returns, not from guesses about your store, and says so when nothing matches.
• Cards where they help. Product cards appear in the website chat, on the chat link and in the AI Studio's Test tab. On WhatsApp, Messenger and Instagram the same agent answers in text.
• A person when it matters. While a teammate is online in the live chat, a shopper can ask for a person, and the teammate joins the same chat with the full history.

The website chat shows "You are chatting with an AI assistant" above the first message. The sample chats here also use one behavior, saved word for word under Behaviors > Add Behaviors in the AI Studio's "Your agent" panel: "If none of your replies so far has said you are an AI assistant, say so briefly in this reply. The welcome message at the start does not count, since the visitor may not have seen it. If asked whether you are human, say plainly you are an AI."

If you want the whole store in one agent, with order status and support beside product search, our ecommerce guide covers that wider job.

Easier to pitch? Your reviewers already know

The question that decides a technical purchase is often one your catalog cannot answer: what do people who bought it say? For Nadia it was tents. Shoppers compared two models with almost the same specs and asked which one is easier to pitch alone. Her product data has weights and pole types; the answer sits in her customer reviews.

Nadia collects reviews with Yotpo, which connects to WooCommerce through its Yotpo Product Reviews plugin. Per its API reference on November 10, 2026, a public endpoint returns one product's reviews when a request carries the store's app key and the product's ID on your store. The store owner finds the app key in Yotpo under Account Settings > General Settings. The endpoint returns up to 150 per page, can filter by star rating, and new reviews can take up to 3 hours to appear. With Ultimo Bots, Mo builds on any public API that works with an API key or token, so the reviews become a skill:

1. 1

   Nadia describes the job to Mo

   In the AI Studio she writes: "When a shopper asks what reviewers say about a product, or compares two, read up to 12 of its published reviews in Yotpo through the API. Use only the rating, title and text, never a reviewer's name or email. Say how many reviews you read and answer only from them."

2. 2

   Mo reads the docs and asks for the key

   Mo reads Yotpo's API reference and asks for the app key in a secure field, where it is stored encrypted.

3. 3

   Mo builds and tests the lookup

   Mo builds the steps it needs, finding the product's ID on your store and then reading its published reviews in Yotpo, runs them against the real store and reports what the agent can now do.

In a fictional chat, a shopper asks "Between the Ridgeline 2 and the Cove 2, which one do reviewers say is easier to pitch?" The agent answers: "I'm Tamarack Trail Outfitters' AI assistant. I read 12 published reviews of each tent. Reviewers find the Ridgeline 2 easier: 7 of its 12 mention pitching it alone in a few minutes, and several credit the color-coded poles. Four Cove 2 reviewers say the rainfly is hard to attach alone in wind."

The skill reads published reviews, the same for every shopper, so no customer account or order is involved. Each answer that uses it costs 1 AI credit on Ultimo Core, the standard level, like any other answer. The skill takes one of your plan's skills (Smart 3, Boost 10, Ultimo 30), and building it with Mo uses more credits than a quick question.

Other review platforms with a public API and a key work the same way, and so do your own systems: a size guide service, a compatibility database for parts, a rental calendar. Describe the job to Mo in the AI Studio, or connect an MCP server. Ready-made skills cover Wix, WooCommerce, Stripe, Cal.com, Calendly, HubSpot and Mailchimp.

What does an AI product finder cost?

As of November 10, 2026, Ultimo Bots Smart costs $29 a month or $19 billed yearly with 500 AI credits, Boost $59 or $39 with 2,000 and Ultimo $249 or $199 with 10,000. A standard answer uses 1 AI credit, product search included, and teammates are free. Agents installed from Wix are billed through the Wix App Market, which shows your price. See pricing.

How that compares with per-seat and per-resolution tools is covered in what an AI agent costs and in our price check of low-cost options.

Which one should your store add first?

Start with the shoppers you lose, not with a tool. If most visitors browse a category and buy, your filters serve them, and the agent still takes their shipping, returns and sizing questions. If they describe needs in their own words, leave long searches with no results, or email you "which one should I get?", add the AI agent next to your filters.

To see what the agent does for those shoppers, take ten of those long searches and emails as a test set. Connect your WooCommerce or Wix store or upload a catalog export, then ask the ten in the Test tab of the AI Studio (test chats use AI credits too). Each answer you would not give yourself points to a column your catalog lacks. Each question only your reviewers or your own systems can answer is the first skill to describe to Mo.

Questions and answers