Skip to Main Content

How to boost your concept testing survey with AI-generated ideas

Discover how we pitted AI and human ideas against each other to learn about the value of generative AI in concept testing.

Person on laptop with a phone next to them showing the AI app

The Rise of AI in Market Research

At this point, artificial intelligence (AI) technology is evolving by the day. AI can now help you become an artist, a better writer, and even create music using the voices of famous singers

Market research can definitely benefit a lot from generative AI (AI that can be used to create new content) - especially now that tools like ChatGPT are available to the broader public. Insights pros can now use this technology to improve their reports, create better data-driven stories, and generate new concepts.

Read more: 4 Ways to Use AI and Improve the Innovation Process

Uses of AI in Concept Testing 

The concept testing aspect of market research is becoming more manageable with the rise of AI.

One of the most significant benefits of using AI in concept testing is speed. 

Instead of spending hours and days brainstorming new ideas, teams can now use tools like ChatGPT to generate creative and unique concepts in a few seconds. By generating ideas quicker, you can launch your testing sooner and get to insight faster.

Another great thing about AI is the quantity of ideas it can generate. 

While not every idea AI generates is a winner, you have a greater range of options to choose from. You can even use those less compelling AI-generated ideas and build off them as a starting point. So, you never have to be scared of having a blank page because AI can speed up the ideation and brainstorming process.

AI is perfect for testing concepts because it’s flexible and easy to operate. 

With the right prompts, researchers can control what kinds of ideas AI will generate. They can even direct the machine to write a concept according to their own research standards. 

AI v Humans: The Results 

As innovation geeks, we wanted to see how well AI can perform in real concept testing conditions (and fool around with ChatGPT a little bit). 

Can AI generate compelling enough ideas for new products? Can humans come up with ideas which are as unique and interesting? And who would win this battle of best idea generation - AI or humans?

Study #1: Milkshake flavors

We asked ChatGPT to come up with 10 new flavors for milkshakes. Then, we asked our internal team to do the same. The goal was to create flavors that had never existed before so that respondents could judge not only on flavor but also creativity.

We then inputted all flavors into an Upsiide idea screening tool and surveyed 200 people in the US and Canada. Respondents swiped right if they liked a flavor and left if they didn’t like it; then, they were asked to choose a favorite.

The result? Humans won. 

Chocolate + Baileys (a human-generated idea) came to the top with an Idea Score of 68. 

But AI came just one point behind the winning idea: White Chocolate and Raspberry Truffle was a close runner-up, scoring 67 in Idea Score.

And what’s cool is that US consumers especially loved the idea of White Chocolate and Raspberry Truffle. So here’s the market for ChatGPT in case it wants to open a milkshake business (hey, you never know). 

AI also did well capturing the attention of a small but dedicated group of people. Spicy Mango Habanero, an AI idea, appeared in the Niche quadrant - meaning that there is a group of consumers who would choose this flavor over all others.

Preview of TikTok video on a phone

Study #2: Dipping sauce flavors

Okay, after testing different milkshakes, we kind of got hooked on testing more food flavors. This time, we screened new flavors for dipping sauces.

Smoked Crispy Bacon Crunch (a human idea) topped the charts with an Idea Score of 66 - which we think makes sense. In our previous research on burger flavors, we discovered that North American consumers really love flavors that convey a crispy texture or contain bacon.

Maple Bourbon (a ChatGPT creation) came second, scoring 55. But we think if this idea was developed a little more, taking into context that consumers love flavors with different textures, it could have had a chance against Smoked Crispy Bacon Crunch. 

As a matter of fact, we went back to ChatGPT and asked it to incorporate some texture-related wording in this flavor idea, and it gave us some interesting options:

  • Maple Bourbon Crisp Delight

  • Spicy Bacon Maple Bourbon

  • Salted Pretzel Maple Bourbon

Who knows, maybe one of these flavors would have beaten the winner.

Preview of TikTok video

Study #3: Credit card rewards for Gen Z

For this last study, we thought AI needed a different kind of challenge. So, we entered the finance world and asked ChatGPT to develop some credit card reward ideas specifically targeted at Gen Z. 

When we looked at the ideas submitted by our team and AI, we found many of the same ones. While our team has tons of experience working with finance clients, ChatGPT doesn’t. 

So it was so cool (and sort of creepy) to see how well AI knows banking and what rewards consumers might like. And it was evident in the results too. 

2 ideas in the top 3 belong to AI. 

“Discounts on food delivery services, such as Uber Eats or DoorDash” and “Discounts on streaming services, such as Netflix or Spotify” scored 72 and 74 in the Idea Score, respectively. But the winner still became a human idea, “Cashback rewards on purchases from online retailers (e.g. Amazon)”. 

AI again did well in finding a rewards idea that would be attractive to a small niche of Zoomers - it was “rewards points for making peer-to-peer transactions, such as splitting bills with friends or paying rent”. 

Preview of TikTok video

Key takeaways and why you should leverage AI in concept testing

After pitting AI and human ideas against each other, we took away these key learnings:

1. AI can generate ideas just as creative and unique as human ones. 

And with a bit of fixing and developing here and there, AI actually has tons of potential to compete against human ideas. 

2. Concepts created by AI perform well with niches of consumers, which are small in size but remain committed to that one idea. 

If presented by the right brand or with the right messaging, these concepts can perform even better than human ones

3. AI can significantly speed up the concept testing process, making it more efficient and allowing teams to explore more ideas in less time.

Less time brainstorming = more time focusing on analysis and strategy.

Ready to launch your first concept testing survey on Upsiide using AI?