AI-written reviews are more extreme

🪫 Out of office. Back in 2 weeks! 🔋

Diving through hundreds, thousands of research papers requires extreme focus and attention to detail.

To do that properly, we need to make sure we are always in tip-top shape. That includes taking a break to recharge your batteries when you have a chance to do so. As many of you are slowing down in August, we are seizing the moment for a 2-week newsletter break too.

Once we restart on Tuesday 1st September, expect a bunch of exciting things in the following weeks:

• Even better insights, tweaked based on your feedback

• A new FREE Playbook, we’re building with a partner (to be revealed) that I’m certain you’ll love [Hint: it’s related to this insight]

• A new and improved experience of our Platform with 300+ insights

See you in 2 weeks!

📝 Context

Topic: Reviews | AI
For: B2C
Research date: June 2026
Universities: LMU Munich 

You’ve read our insight and started replying thoughtfully to every review your travel agency gets on Tripadvisor. You’ve seen your review count increase. However, lately many feel different, as if they’re written by AI, not your actual customers. 

You read sentences like “It’s not just good - it’s mindblowing!” and wonder, what’s the effect of AI-written reviews?

Science says you might want to pay attention to this, and perhaps encourage people not to use it.

P.S.: People like personal touches. This works in reviews too. If you have reviews where people mention a little about themselves (e.g. names, city), highlight them. Readers will find them more useful.

📈 Recommendation

People are often tempted to use AI to write reviews. This can increase the number of reviews you receive but backfire by making the occasional negative review even worse.

If that happens to your business, discourage people from using AI (e.g. “Please write the review yourself, not with AI, other readers will appreciate your words much more!”)

You might get fewer and shorter reviews, but fewer extremely negative ones.

🎓 Findings

• People who use AI to write reviews produce more and longer ones, but they are also more extremely negative.

• Scientists analyzed 117,816 reviews and found that when reviewers could use AI (vs not, during 4-week ban of ChatGPT in Italy in April 2023):

  • The number of reviews increased by 2%

  • Review length increased by 9%

  • Extremely low reviews (1 or 2 stars) were more negative in content (e.g. “I don’t like the room” vs “I hate the room!!”)

  • The content sounds more subjective (e.g. “I didn’t like the breakfast) vs objective (e.g. “Breakfast was local and no continental option was available”), making negative reviews even more harmful.

🧠 Why it works

• We use AI to reduce effort and time spent writing reviews

• Because of this, we feel we’re more productive and efficient, so we are more likely to leave a review.

• But AI tends to write longer and more complex sentences than we do.

• And it reframes our writing to make our negative observations sound more extreme.

✋ Limitations

• The study looked at hotel reviews. The effect might vary for products and other types of services where writing a review feels less or more important (e.g. restaurant vs medical visit).

• AI adoption has increased since this data (April 2023). The effect might change with even more reviews being written by AI, or different models having different biases (e.g. reviews could now be more or less subjective and more or less negative).

• People may also become more experienced in using AI and more aware of its limits. This might change how likely they are to check and edit AI’s outputs, but we don’t know in which direction.

👀 Real-life example

Tripadvisor has a strict AI policy, whereby they monitor for AI-generated content, delete it, and ban the reviewer.

❌ Issue: Considering the businesses on Tripadvisor (e.g. hotels, restaurants) are prone to negative subjective reviews, they should discourage AI use right away.

✅ Solution: Simply understanding what makes people likely to use AI can help Tripadvisor use the right messaging to decrease AI usage:

• Those who know less about AI, are also more likely to use it because they see it as magical. This feeling can be weakened by explaining how it works (e.g. “AI is not magic and doesn’t read your mind. What you write with it reflects what it has already memorized from other people”)

• People see no issues in using AI themselves but are more critical of others using it. At the same time, we don’t want to look bad in the eyes of others so Tripadvisor could say something like “People can spot AI-written reviews, science shows they will be more critical of your words if you use it”.

• Overall, they should make sure to use assertive language (e.g. always, all, never, definitely) in their copy discouraging AI use. This will make the message, and them look more powerful, and make people less likely to go against it.

🔍 Study type

Market observation (analysis of reviews on 4909 hotels listed on Tripadvisor in Europe during a temporary Chatgpt ban in Italy).

📖 Research

The impact of generative artificial intelligence on consumer reviews: Insights from a quasi-experiment. International Journal of Research in Marketing (June 2026).

🏫 Researchers

• Weilong K.M. Gu, LMU Munich 

• Martin Spann, LMU Munich 

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 14 hours to accurately turn this 15-page research paper into this 3-min 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 👇