How do I improve my alert's accuracy?

Improving the accuracy of your alerts is an iterative process of refining your instructions to the AI. If you're getting too many irrelevant results (false positives) or missing expected matches (false negatives), here are the best ways to improve performance.
1. Review the Search History
This is the single most important step. Go to your alert's detail page and look at the Search History.
For "No Match" entries: Read the AI's reasoning. It will often tell you what it found and why it didn't meet your condition. This is your best clue for how to adjust your condition.
For "Match Found" entries: If a match was incorrect, review the source URL and the summary. This helps you understand how the AI misinterpreted the content, allowing you to add more specific negative constraints to your condition.
2. Make Your Condition More Specific
The more specific your instruction, the better the AI can perform.
Add Keywords: Include more specific terms that would only appear in a relevant result.
Use Quotes for Exact Phrases: If you need a specific phrase like
"announced a partnership", put it in double quotes.Add Negative Constraints: Tell the AI what not to look for. Example:
"Find new AI startups, but exclude any that are focused on healthcare."
3. Adjust the Timeframe
Ensure your Search Timeframe is appropriate. A timeframe that is too broad ("Last 90 Days") for a news-based alert can bring in old, irrelevant articles. For most breaking events, "Last 7 days" is the most effective setting.
4. Use Custom AI Insights to Verify
Add Custom AI Insights to ask targeted, clarifying questions. This forces the AI to look for specific data points before it can confirm a match.
Example Condition:
"A company announced a new Head of Sales."Good Custom Insight:
"What is the full name of the new Head of Sales?"Even Better Custom Insight:
"Does the article explicitly state that this person's title is 'Head of Sales' or 'VP of Sales'?"
By using a combination of these techniques, you can fine-tune your alerts to deliver highly accurate and relevant results.