Using generative AI for UX research — practitioner case study

What the article covers

Tania Ostanina updates her popular AI-for-UX method for 2025, incorporating ChatGPT Enterprise’s Deep Research functionality. The article includes a real-world case study with anonymized data, showing the complete workflow from feeding data to harvesting insights.

Context

This is the second iteration of Ostanina’s approach, following an original article that became a “sleeper hit.” The update was presented at the HCID 2025 conference and reflects meaningful changes in available tools, particularly the expanded ability to process large documents, JSON files, screenshots, and web links.

Key takeaway

Ostanina’s method follows three steps: Feed (load context including data, documents, and discussion guides), Instruct (use a structured “Prompt Deck” approach), and Harvest (extract and apply the outputs). The critical addition in 2025 is the role of the Discussion Guide as the anchor document for AI analysis. By feeding the AI both the raw data and the research plan, including objectives, hypotheses, participant information, and interview structure, the AI produces outputs that are directly aligned with the research questions. The case study demonstrates this concretely: the prompting method references the discussion guide’s structure, ensuring AI-generated themes map back to predefined research objectives rather than producing generic summaries.

Who should read this

Researchers who want a specific, tested prompting method for qualitative analysis, particularly those using ChatGPT Enterprise or similar tools with Deep Research capabilities, and anyone who found value in Ostanina’s original method and wants the updated version.