When AI chatbots boost vs hurt sales

This insight is brought to you by… Science Says Platform

The Science Says newsletter keeps you up to date, and I hope you’re loving it 💜

If you’re a marketing pro, we built the Science Says Platform to be the place you turn to before any project to check “What is the latest evidence I should know about, so I get the best possible results?”

 You’ll find all our 300+ insights, like:

• Discount coupons: when to give % off vs $ amount off

• When to use vouchers vs free gifts for promotions

• Make your discounts a bit harder to get

Only paid Platform members can access them all 24/7, along with real-world case studies, the exclusive playbook of brand names, and direct access to our team of PhDs and me to answer your questions.

Want to sponsor Science Says? Here’s all you need to know.

📝 Context

Topic: AI | Ecommerce
For: B2C. Can be tested for B2B 
Research date: April 2026
Universities: IÉSEG School of Management, University of Lille

You’re the owner of a sports apparel ecommerce brand with 10,000+ products and dozens of brands. You just got a proposal to implement a chatbot that would help people find clothes most fitting to their needs and style. It got you thinking, “would my customers really use a chatbot to shop for clothes?”

Science says yes, and they might be even more likely to buy what it suggests.

P.S.: If you’re implementing a chatbot and want to make it even more effective you can tweak how it speaks. Research shows making it use interjections (e.g. oh wow, aw) makes people up to 49% more satisfied with it.

📈 Recommendation

Add a chatbot to your website if your product is hard to evaluate (e.g. has different elements to consider like fit, size, style). Avoid if your product is easy to evaluate and has a small assortment size (e.g. a Vitamin with 20+ options).

People will have higher intentions to buy your product when suggested by a chatbot.

🎓 Findings

• AI chatbots boost purchase intention and customer engagement when products are hard to evaluate.

• Across 4 experiments with 1,339 people, researchers found that when chatbots suggested products, people:

  • Had 44.7% higher purchase intentions for products that are hard to evaluate (e.g. formal office pants)

  • Had lower intentions to purchase for products that are easy to evaluate, and have small assortments (e.g. a supplement line with only a few vitamin options)

  • Had higher levels of engagement leading to higher purchase intentions for products that are easy to evaluate but have big assortments (e.g. 10+ options for multivitamins)

🧠 Why it works

• We find it hard to judge products that require comparisons between many options and have different elements to consider (e.g. fit, style, size).

• This makes us feel ambiguity and uncertainty in our choices.

• AI chatbots can help us find and structure the relevant information, which helps us feel more in control when we shop.

• On top of that, we interact and converse with chatbots the same way we do with other humans. 

• This makes us even more engaged and more likely to purchase what they suggest.

✋ Limitations

• The research looked at only a few product types (e.g. pants, books, multivitamins, and blazers), it’s unclear how the effect would be for services, or other kinds of more expensive products like luxury goods, or experiences (e.g. holidays, restaurants).

• The study did not consider how mistakes in recommendations (e.g. inaccurate information) might impact people’s likelihood to continue using the chatbot and accept its suggestions.

• It’s unknown how the environment where chatbots are implemented (e.g. in-store, website, apps) might change people’s likelihood to follow their recommendations.

• The design of the chatbot (e.g. human-like or machine-like) and how it speaks greatly impacts how people react to it and how likely they are to buy. This was not studied in this research, but we cover that extensively in our free Blueprint for Effective AI Chatbots we created in collaboration with Wharton.

Want to advertise to 30,000+ Science Says readers here? Click here.

👀 Real-life example

Asos is an online retailer selling items from all major brands. Their new arrivals page has 5k+ styles for women’s clothing alone.

❌ Issue: Thousands of items in different colors, sizes and styles, are overwhelming. People might find it hard to shop.

✅ Solution: They could improve their shoppers’ experience through some simple tweaks:

• Introducing a chatbot that would help people find the right style and compare options.

• Make sure that the AI clearly categorizes options by characteristics (e.g. open back, midi skirt) when suggesting multiple ones to reduce choice overload.

• For an extra nudge, make it use reassuring language (e.g. this fits well with long hair), instead of pressuring (e.g. only 3 left).

🔍 Study type

Online experiments.

📖 Research

Generative AI and customer engagement with the retailer: Does product type matter? Journal of Business Research (April 2026)

🏫 Researchers

• Lina Anggraini. University of Lille

• Nathalie T.M. Demoulin. University of Lille

• Gwarlann De Kerviler. University of Lille

Remember: This is a new scientific discovery. In the future it will probably be better understood and could even be proven wrong (that’s how science works). It may also not be generalizable to your situation. If it’s a risky change, always test it on a small scale before rolling it out widely.

🎁 Bonus: Trivia

Check your knowledge from previous insights (for paid Platform members only).

Here is how else Science Says can help your marketing:

• 📈 Join the Science Says Platform to unlock all 300+ insights, real-world case studies, and exclusive playbooks

• 📘 Boost your sales and profits with topic-specific Science-based Playbooks (e.g. Pricing, Ecommerce, SaaS, AI Best Practices)

• 🔬 Get science-based thought leadership content for your brand. We help market leaders in AI, tech, and across B2B enterprises, build authoritative thought leadership using a scientific perspective. Find out more.

🎓 It took me, a PhD analyst, and a marketing pro 11 hours to accurately turn this 18-page research paper into this 3min insight. 

If you enjoyed it please share it with a friend, or share it on LinkedIn and tag me (Thomas McKinlay), I’d love to engage and amplify! 

If this was forwarded by a friend you can subscribe below for $0 👇