---
path: docs/owl-ai/prompting-tips
title: "Prompting Tips"
desc: "Write effective prompts for Owl AI."
date: "Apr 2, 2026"
---

# Prompting Tips

Get the best results from Owl AI by following these tips.

## Be Specific About Your Data

Reference layer names and column names explicitly. Instead of "show me the big cities," say "filter the `us_cities` layer to show features where `population` is greater than 500,000."

## Specify the Output Format

Tell Owl AI what you want as a result: a new layer on the map, a table, a chart, or a number. For example: "Calculate the total area of parks in km² and show it as a number" versus "Highlight all parks on the map."

## Use Spatial Language

Owl AI understands spatial terms like "within," "near," "adjacent to," "overlapping," "inside," and distance units (meters, kilometers, miles, feet). Use them: "Show all restaurants within 1 km of subway stations."

## Iterate

If the first result isn't right, refine your prompt. Owl AI maintains conversation context within a session, so you can say "only include results from the downtown district" to narrow down a previous query.

## Complex Workflows

Chain multiple steps: "First, buffer all highways by 100 meters. Then, find all residential parcels that intersect those buffers. Finally, calculate the total assessed value of those parcels."

## What to Avoid

- Ambiguous pronouns without context ("Show me those things")
- Requests without spatial context ("What's the best location?" — best for what?)
- Expecting Owl AI to know external data it doesn't have access to