Prompt intent classification
Updated 2026-09-15
Prompt intent classification is the step that sorts a tracked prompt into a class — brand, discovery, comparison, pricing, switching — before metrics are computed. It turns one flat average into segmented rates that describe different commercial problems.
Why the segments matter
Branded prompts almost always score near the ceiling: the brand name is in the question. Discovery prompts almost always score lowest, because the engine is choosing who to name. Reporting one blended number therefore hides the only figure that reflects new demand. Classification is applied per prompt at measurement time and kept stable across cycles, so a shifting mix cannot be mistaken for a shifting rate.
Illustrative segmentation (example numbers)
Blended mention rate 0.23 splits into brand 1.00 and discovery 0.04. The blended number looks like modest visibility; the split shows that everything measured is recognition and nothing is discovery.
Common mistakes
- Reporting a blended rate that is really a brand-prompt average.
- Reclassifying prompts between cycles, which changes the segments under the trend line.
- Removing branded prompts entirely — they are a useful control for whether the measurement itself is working.
Frequently asked questions
- How many intent classes are enough?
- Five covers most B2B and consumer categories: discovery, comparison, use-case or integration, pricing and switching, with brand prompts kept as a small control group.
- Can classification be automated?
- Largely, since brand and competitor names in the prompt text carry most of the signal — but the classes should be reviewed once when the prompt set is built, because the segmentation is what every later report depends on.
