Why AI chatbots fail shoppers, and what works instead
An AI chatbot on a store waits in a corner until a shopper opens it and types a question, and the shoppers who need help most are the ones who aren’t sure what to ask. What works instead is an answer on the product page itself, right where shoppers hesitate, with facts from the brand and proof from people who bought it.
Updated · By Gro, Stockholm
Chat widget vs AI overview on the product page
A chat widget is a separate window: the shopper has to notice it, open it and know what to ask. An AI overview on the product page, which is what Gro does, is part of the page. It sits under the price, next to the reviews or in the FAQ, asks one short question when a shopper hesitates, and shows where every answer comes from: the brand’s own facts, or what people who bought it say.
What the research says
- 01
About 1 in 100 visits buys
The median Shopify store converts 1.3% of its sessions; the top 10% of stores are above 3.6%. Many of the rest would buy, but aren’t sure yet.
Source: Littledata, median of 450 Shopify stores, 90 days to 7 Nov 2026
- 02
Shoppers trust ratings that look real, not perfect
Purchase likelihood typically peaks at an average rating of 4.0 to 4.7 stars and falls as ratings get close to 5.0. Honest proof, including the critical reviews, reads as more believable than a perfect score.
Source: Spiegel Research Center, Northwestern University, “How Online Reviews Influence Sales” (2017)
- 03
Inflated ratings come back as returns
Products shown with an average rating higher than their true rating were returned more often, and more reviews, and more helpful ones, went with fewer returns. Answers should come from people who bought it, not from a sales pitch.
Source: Sahoo, Dellarocas & Srinivasan, Information Systems Research 29(3), 2018
- 04
Shoppers already use AI to find products, not to check out
The share of Swedish online shoppers who use AI such as ChatGPT, Gemini, Perplexity or Copilot to find products went from 10% in 2024 to 20% in 2025. Only 3% think AI helps most in the cart and at checkout. Help belongs where they research: the product page.
How the options compare
| Question | AI chat widget | AI overview on the product page (Gro) | Static FAQ and reviews |
|---|---|---|---|
| Where it lives | A bubble in the corner of every page | Inside the product page: under the price, next to reviews, in the FAQ | Further down the page, or on its own page |
| Who starts it | The shopper opens it and types | It’s already on the page, and asks one short question when a shopper hesitates | The shopper scrolls and searches |
| Where facts come from | Whatever the bot was given; the source is often not shown | The brand’s catalog, pages, policies and files, marked with the store’s name | The brand, written once by hand |
| Where proof comes from | Usually none, or the bot’s own summary | Reviews from people who bought it, quoted and marked “People who bought it say” | The full review list, not sorted to the shopper’s question |
| Setup | Varies by vendor; often a script plus answers to keep up to date | One snippet on your pages; reads your catalog and the reviews you already have; set up with you | Written and updated by hand |
| How it’s measured | Usually chats started or tickets deflected | Sales compared against a share of visitors who never see Gro | Rarely measured on its own |
When a chatbot is the right choice
Chatbots are good at some jobs. Use one when:
- The question comes after the sale: order status, returns, delivery changes, account help.
- Shoppers already know what they want and only need a quick service answer.
- You need a hand-off to a live agent for complex cases, like custom orders or B2B quotes.
They can run side by side: Gro on the product pages before the sale, a support chatbot or helpdesk after it.
Questions
Are AI chatbots bad for online stores?
Not for everything. They’re good at service questions after a sale. Before the sale they depend on the shopper opening a window and knowing what to ask, which is hard when you’re not sure yet.
What is an AI overview on a product page?
A short answer that’s part of the product page, not a window on top of it. It sits where shoppers stop to check, like under the price or next to the reviews, and every line shows whether it comes from the brand or from people who bought it.
What is the best alternative to an AI chatbot for an online store?
For questions before the purchase: answers on the product page itself, with facts from the brand and proof from people who bought it. That’s what Gro does. Richer product pages, size guides and review apps with Q&A help too, but they can’t ask what’s holding a shopper back. For order and returns questions, a support chatbot or helpdesk is still the right tool.
Does Gro replace our support chatbot?
It doesn’t have to. Gro works on the product pages before the sale; your support tool keeps handling orders and returns after it.
How do you know it works?
A share of visitors never sees Gro, so sales are compared against a group that didn’t get it, on your own numbers.
Which stores is Gro for?
E-commerce, retail and consumer brands whose shoppers hesitate on the product page: size and fit, which product, will it work for me. It runs on Shopify, Centra or a custom storefront.
See Gro on your own store
Open the homepage and press “See Gro at work”, or book 30 minutes and we’ll show it on your own product pages.