Make your chatbot use interjections (Oh wow!)

New to Science Says? This is a 3min practical summary of a scientific study 🎓 Join 31,402 marketers who use science, not flawed opinions 📈 Subscribe here.

This insight is brought to you by… KeepCart

Prevent discount coupon abuse that eats into your ecommerce profits.

Overpaid commissions to affiliates and influencers add up fast, eating up to 15% of your margins. Get rid of the headache and revenue losses with KeepCart. 

Try it for free for 14 days to find out which of your coupons have leaked on sites like Honey, CapitalOne, RetailMeNot.

Cancel anytime, no commitments.

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

📝 Intro

Imagine you're searching for the perfect anniversary gift for your partner. 

After aimlessly browsing through tens of product pages, you decide to give the ecommerce site’s AI chatbot a chance. You type, "Help me find a special gift for my girlfriend for our anniversary." 

You are expecting a generic response with a couple of product links, but instead, you're surprised when it says, "Aww, that’s fantastic! How about this fruity perfume?"

It feels good - like a chatty agent who understands you, rather than AI.

What triggered the emotion?

🔥 Just out -> We’ve partnered with The Wharton School, University of Pennsylvania, to create the Wharton Blueprint for Effective AI Chatbots. 

A practical science-based guide for tweaking your customer service AI chatbots (or any other chatbot), so that users will be more satisfied, like you more, and be more likely to buy from you (Bonus: you can even feed it as instructions to the AI powering your chatbot 😉).

It’s free to view and download, just grab your copy here. And if you like it, we go even deeper in our own Science Says Playbook of AI Best Practices.

Want hundreds more insights like these? Explore all Science Says insights here.

People prefer chatbots that use interjections - and are more likely to buy

Topics: AI | Website/App | Ecommerce | Customer Experience
For: B2C 
Research date: August 2024
Universities: International University of Japan, Queensland University of Technology, University of Strathclyde

📈 Recommendation

Make your AI chatbot use interjections like “Hmm,” “Oh,”, “Wow”,  “Uh huh,” or “Aww” to mimic human reactions to a conversation (e.g. surprise, comprehension). This makes the chatbot feel more human and engaging. 

People will feel heard, enjoy the interaction more, and be more likely to trust its suggestions and make a purchase.

🎓 Findings

• People prefer interacting with chatbots that use interjections (vs a chatbot that doesn’t). They are also more likely to accept suggestions and buy.

• In 4 experiments, researchers found that when a chatbot used interjections (vs. not), people:

  • Were 49.7% more satisfied (i.e. enjoyed the interaction more) and thought the chatbot performed 8.2% better

  • Found the chatbot 32.5% more human-like, and felt 17.5% better understood by it

  • Even when the outcome was not positive (e.g. The chatbot denied returns), people were 18% more likely to purchase the alternative suggested by the chatbot and 54.8% more likely to continue shopping with the company

🧠 Why it works

• When chatbots have human-like traits (e.g. conversational or empathetic tone of voice), we perceive them as more human.

• Chatbots that use interjections mimic our ability to convey thoughts and feelings, not only plain information.

• This makes us feel listened to and makes our experience more positive and enjoyable.

• That improves our satisfaction and makes us more loyal and willing to make purchases.

🏆 Why B2B SaaS Leaders Trust dofollow.com

Companies like Surfshark and Pitch don’t leave rankings to chance—they invest in high-authority backlinks for real growth. Google ranks trust, and the right links build it fast.

Why dofollow.com?

✔ DR 70-90+ backlinks from the best sites in your vertical

✔ Dedicated strategist who understands your audience

✔ Real-time reporting & free link replacements

✔ No lock-in contracts—just results



This announcement was sponsored. Want your brand here? Click here.

✋ Limitations

• The study looked at text-based chatbots (e.g. Facebook Messenger chatbot and website-integrated chatbot). It did not look at bots interacting through audio (e.g. Siri), or video (e.g. Avatars) where the tone of voice or facial expressions could impact people’s perceptions too.

• The experiments were run with American participants, other cultures might prefer different language - such as more formality.

• The study did not consider how the effect differs across different buying stages (e.g. information search, post-purchase) or contexts (e.g. retail, hospitality, warranty claims).

🏢 Companies using this

• Companies are increasingly giving their chatbots conversational and empathetic tones of voice (e.g. use of emojis) that mimic how humans interact with one another.

• However, interjections are not widely used, companies seem to limit the “humanization” of chatbots to a friendly tone of voice, avatars, and names.

  • Zalando’s shopping assistance chatbot uses a helpful tone of voice, similar to that of a human customer assistance. Considering the informal and fun context, interactions could be improved by the use of interjections.

  • Canva’s AI assistant provides friendly and relevant comments (e.g. That’s awesome!) when you reference your project, mimicking a human response. Using interjections could further enhance this positive effect.

  • Personal AI assistant Pi, when mentioning a relevant milestone, responds with “wow!”, however, it doesn’t seem to be using other types of interjections.

Social scheduling tool Buffer answers people with a friendly tone and uses emojis. The interaction could be enhanced by adding relevant interjections.

⚡ Steps to implement

• Consider whether chatbots using interjections are appropriate for your business context. If the task has high stakes for customers, and/or high professionalism is expected (e.g. requesting insurance help), interjections might come across as unprofessional and backfire. In such contexts, more machine-like interactions might work best.

• If appropriate, make your chatbot use relevant interjections that support the main message and build rapport (e.g. “Oh no! That sounds frustrating” for complaints) to show empathy and boost customer satisfaction.

• Highlight the benefits of your chatbot (e.g. immediate assistance) to increase customer satisfaction even further.

• Give your chatbot other human-like traits (e.g. name, use of emojis in chats) to make your chatbot feel even more human. People will feel closer to it and reduce unethical behavior (e.g. reporting false reasons for returns), which is more common when chatbots feel less human.

• Train your AI to mirror interjections customers use themselves to personalize interactions further.

• Avoid overusing or forcing them, as it can make conversations feel unnatural and inauthentic.

🔍 Study type

Online experiments.

📖 Research

Wow! Interjections Improve Chatbot Performance: The Mediating Role of Anthropomorphism and Perceived Listening. Communication Research (August, 2024).

🏫 Researchers

• Ben Sheehan. International University of Japan

• Hyun Seung Jin. Queensland University of Technology

• Brett Martin. Queensland University of Technology

• Hyoje Jay Kim. University of Strathclyde Glasgow

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.

Here is how else Science Says can help your marketing:

• 📈 Join the Science Says Platform to unlock all 250+ 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 on-demand evidence to make better decisions. My team of PhDs and I regularly help leading brands in FMCG (e.g. Mars), retail, and tech. Reach out here.

🎓 It took 3 of us 11.5 hours to accurately turn this 27-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 👇