Methodology
We measure a live activity with a human, we don’t ask about it afterwards.
We observe a real, unscripted conversation between a real buyer and an AI agent, and track every dimension of that exchange. Our respondents engage in an activity increasingly familiar to over a billion consumers*, an activity with a real decision behind it, which surfaces things no research designer would think to look for. At scale, that produces patterns which help us truly understand the shape of an AI purchase conversation, and its implications for your brand.
* Source: DataReportal Digital 2026 report
What Nova is
Nova is a research instrument, built to give a high definition picture of every level of the AI purchase conversation.
Nova is the AI purchase conversation.
Nova facilitates an already mainstream activity, measuring buyers before, during and after they enter the conversation, and measuring every action and decision the AI takes along the way.
How a session runs
Six stages, in order, every time.
-
Participants are recruited because they are intending to buy that product, and Nova supports them in making the right choice. Anyone whose situation doesn't genuinely match the study is screened out.
-
Every respondent gives Nova an opening prompt, exactly what they’d typically type into an LLM in this context.
-
Before the conversation starts, participants answer a few questions about their situation. This gives us further quality checks, and enables us to segment the data to ensure we can slice our buyers across the most meaningful cohorts.
Nova is blind to this intake, it only sees the opening prompt.
-
With only the buyer’s opening prompt to guide it, Nova initiates a dynamic and complete conversation to support the participant with their intended purchase. Nova follows the typical system prompt for the LLM platform and model underneath it.
-
Every message, every moment impacting the decision, every instance a brand is mentioned and how it’s received. Every search run by the AI, every search that makes it to the conversation, every way the AI describes a brand, and every hyperlink the buyer actually clicks. It’s all tracked, and personal or sensitive details are automatically stripped before anything is stored.
-
After the conversation, participants answer a short set of questions about their experience. We build these questions dynamically to capture the most valuable information for that product and that purchase journey. This further enhances our segmentation capabilities.
Let's talk.
Tell us what you need to know. We'll go find out.