Can AI really plan your research? Expert insights from a blind authorship study

As research and design professionals, we know that a good research plan sets the foundation for actionable insights and project success. But with growing pressure to do more with less, many teams are exploring how AI might accelerate their planning process.

Recently, we put this idea to the test. Could AI — specifically ChatGPT and Gemini — generate a research plan good enough for experts to trust?
We asked both models to develop research plans exploring Gen Z’s willingness to sacrifice convenience for environmental benefits. Then, we gave these plans (without revealing their AI origins) to eight experienced researchers from the U.S, Canada, and Australia for their feedback.

Here’s what we learned about what AI does well, where it still falls short, and how human designers can future-proof their process in an AI-augmented world.

What exactly did ChatGPT and Gemini produce? Both AI models were given the same prompt, yet each crafted a unique research plan. Still, several common threads emerged:

• Core Structure: Each plan included research objectives, a set of research questions, methodology, data collection tools, analysis approach, ethical considerations, a timeline, and expected outcomes.
• Methodology: Both favored a mixed-methods approach, combining surveys for quantitative data with interviews for qualitative insights.
• Extras: ChatGPT’s plan included a list of hypotheses and a budget section. Gemini’s plan offered “Additional Considerations,” such as strategies for participant recruitment and suggestions for pilot testing — practical touches appreciated by some reviewers.

The plans also contained fully drafted interview guidelines and survey questions — enough to be put straight into practice, at least at a basic level. Yet, on closer inspection, the structure, depth, and level of detail varied in ways that became significant once human experts got involved.