Exa
Day 0 after signup
Up to 94% fewer tokens per search
How Exa cuts input tokens up to 16x without losing accuracy.
Why it works
- Start with a familiar problem, then introduce the feature as a direct solution.
- Explain the benefit with a specific benchmark and follow with several practical ways to try the feature.
How it is built
- Personal greeting
- Search cost problem
- Feature and benefit
- Ways to try it
- Playground call to action
- Team sign-off
Email copy
Hi [First name], When your agent searches the web, most of the cost comes from reading the results. Built-in search tools usually load entire pages into context, including navigation, ads, and text that has nothing to do with the question. With Exa Highlights, each result comes back as a few relevant sentences. In our benchmarks, answers were just as accurate with up to 16x fewer tokens. A few ways to try it: Ask your AI assistant to use Exa Highlights in search Add "contents": {"highlights": true} to your Exa API calls (docs) Want to see the difference? Test Exa vs Anthropic/OpenAI search Try Highlights in the playground Happy building! - Exa team Exa.ai Docs Careers 3180 18th Street, San Francisco, CA 94110 Unsubscribe Privacy Policy