When AI-written messages backfire

This insight is brought to you by… Modash

100+ ChatGPT prompts to supercharge your productivity.

This free guide from HubSpot is your fast track to AI mastery:

• Productivity Guide: 21 best practices to 10x your efficiency with AI

• Industry-Specific Use Cases: 15+ real-world applications

• Prompt Powerhouse: 100+ ready-to-use prompts

• Challenge Buster: Overcome common AI hurdles

Plus, in-depth sections on email writing, content creation, customer support, and data analysis.

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

📝 Context

Topic: AI | Messaging & Copy
For: B2C
Research date: October 2024
Universities: New York Institute of Technology, West Virginia University

You’re launching a long-awaited update to your B2B prospecting software, and you want to announce it with an email campaign. Prepping for the launch is taking up a big chunk of your time and you decide to use AI to help you. You’re excited so you prompt the AI to write a cheery email and send it announcing the incoming launch.

You expected to get exciting feedback, but 

a week passes and no one, even those that had been asking for that update for months, replies. The reason is that they probably spotted that your message was AI-written and reacted negatively to it.

P.S.: Remember to always be careful when you use AI, people are ok using it themselves, but see others using it more negatively.

📈 Recommendation

Avoid using AI to write emotional messages (e.g. expressing pride for a new product launch, sadness for a loss). If you must use AI (e.g. to help with grammar) make sure the text feels human (e.g. avoid using AI-specific language) and avoid mentioning it was generated by AI.

People feel disgusted and respond negatively to emotional messages written by AI, and are less likely to recommend the brand.

🎓 Findings

• When people know (e.g. it is explicitly said) that an emotional message is written by AI (vs a human), they react more negatively, are less likely to talk positively about the brand, and feel morally disgusted (e.g. felt it was dishonest and fake).

• Across 6 experiments, with over 2,200 participants across different categories (e.g. retail, personal care), researchers found that when people learned that emotional messages were written by AI (vs human), people:

  • Were 24.6% less likely to recommend the brand, and reported 58.4% more moral disgust on a brand’s follow-up emotional email (e.g. sending an email saying “I’m inspired by your commitment to weight loss” after purchasing a diet plan)

  • Reported 64% more moral disgust when a communication about a sales’ manager returnto work was emotional, compared to only 15.9% more when the message was factual

  • Reported 2x more disgust and shared 23.6% less word of mouth on a sympathetic message after a founder’s death.

  • Reported 140.9% more moral disgust when a message was written by AI, but signed by a person (vs. written and signed by a human).

🧠 Why it works

• We value authenticity and honesty in our interactions with companies.

• So we are more likely to interact with and trust messages that feel genuine and emotionally resonant.

• However, when we know a brand has used AI to write a message, it doesn’t feel genuine.

• This makes us think the company has violated our moral norms, so we feel emotional disgust.

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

✋ Limitations

• The study only looked at text communication. It’s likely, but wasn’t tested that the effect is the same for communications done in other formats like images, audio, or video (fun fact: we are doing research with Fox School of Business and The Wharton School on just this).

• The study focused on people from the UK and US, people from other countries might have a different perception of AI-generated messages

👀 Real-life example

Ed-tech company infrastructure issued an apology statement after a system outage that disrupted several universities across North America. 

❌ Issue: People felt the message sounded AI-written. A Reddit post about it had overwhelmingly negative comments.

✅ Solution: Always try to write apology messages yourself, as they will always sound more genuine. However, AI can be used as a support (e.g. correcting grammar) but with these guidelines in mind:

• Firstly, ensure it has enough context and is prompted to keep your tone and message. 

• Secondly, the message could thank people for their patience and understanding, as this increases satisfaction more than a simple apology.

• The message should also use “I” instead of “we”. People will react more positively.

• Finally, make it more personal by signing the message, to further improve perceptions.

🔍 Study type

Online experiments.

📖 Research

The AI-authorship effect: Understanding authenticity, moral disgust, and consumer responses to AI-generated marketing communications. Journal of Business Research (October 2024)

🏫 Researchers

• Colleen P. Kirk. New York Institute of Technology.

• Julian Givi. West Virginia University. 

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 mini-insight

Check your knowledge from previous insights (for paid Platform members only). Read the full insight here if you are already a member.

🎓 Insight: Using AI bots can turn customers away if used in the wrong situations

📈 Recommendation: Avoid fully replacing humans with AI if warmth is important to the service you’re providing (e.g. scheduling appointments at a spa, ordering room service in a hotel). If you do, people will be more likely to stop using your services.

You can use machines and AI to provide services if customers mainly look for efficiency and practicality (e.g. grocery check-out).

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 10.5 hours to accurately turn this 14-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 👇