AI chatbots lift sales for hard-to-choose products, and hurt easy ones
A shopping assistant raised purchase intent 44.7% for hard-to-evaluate products (think fit, size, and style), but lowered it for simple products with a small range. Add a chatbot or guided finder where choice is genuinely complex, not where it is easy.
Quick Summary
Across 4 experiments with 1,339 people, an AI shopping assistant raised purchase intent by 44.7% when the product was hard to evaluate, the kind with fit, size, and style to weigh up, like formal trousers. For products that are easy to judge and come in a small range, the chatbot actually lowered purchase intent. It got in the way.
So a chatbot is not a universal upgrade. Add a guided assistant or product finder where your catalog is genuinely complex or large, apparel, furniture, anything with many attributes, and skip it where shoppers can judge the product at a glance from a short list.
Key Finding
+45%
Increase in purchase intent when a chatbot suggested a hard-to-evaluate product such as formal trousers.
Anggraini, Demoulin & De Kerviler, 2026
Every store gets pitched a chatbot. This research answers the real question: when does one help, and when does it just annoy people? The deciding factor is how hard your products are to evaluate.
When products were hard to judge, requiring comparison across fit, style, and size, a chatbot lifted purchase intent 44.7%. When products were easy to judge and came in a small assortment (a handful of vitamin options), the chatbot lowered intent. For products that were easy to judge but came in a large assortment, engagement and intent rose again, because the value was in taming the choice overload.
Why the product type decides it
Hard-to-evaluate products create uncertainty: too many attributes, too many options, no obvious right answer. A chatbot helps by finding and structuring the relevant information, which restores a feeling of control. And because we converse with chatbots much as we do with people, a helpful exchange builds engagement that carries into the purchase. Where there is no uncertainty to resolve (a short list of obvious options), the chatbot adds a step without adding value.
What to do on your Shopify store
- Match the tool to the catalog. Add a guided assistant, quiz, or product finder for ranges with many attributes (clothing, footwear, furniture, skincare regimens) or large assortments.
- Skip it for simple, short ranges. If a shopper can judge the product from the page and there are only a few options, a bot is friction. Invest in the product page instead.
- Reduce choice overload when it does suggest. Have it group options by clear characteristics rather than dumping a long list.
- Use reassuring language, not pressure. "This suits a slim fit" outperforms "only 3 left" for building the rapport that drives the sale.
A caveat
This tested a narrow set of product types and did not model what happens when the bot makes a bad recommendation, which would quickly erode trust. Accuracy and tone matter enormously, so pilot it on one complex category and measure before rolling out.
This is the flip side of optimizing for AI that shops on the customer's behalf, covered in what biases AI agents toward your products. For guided selling on collection pages, see improving product discovery on Shopify.
Research: Anggraini, L., Demoulin, N.T.M. & De Kerviler, G. (2026). Journal of Business Research. IESEG School of Management and University of Lille. https://doi.org/10.1016/j.jbusres.2026.116182
Is this broken on your store?
Get a free UX review.
We review your store against 50+ conversion principles - including this one - and send you a detailed breakdown of what to fix and why.
Get a UX AuditFind your conversion leaks.
A focused expert review of your store with Figma redesigns and a Loom walkthrough. Pick one page or get the full picture.