AI Discovery
Describe the businesses you want, and AI Discovery searches the web, reads each website and drafts a listing for you to review and publish.
What is AI Discovery?
AI Discovery searches the web for businesses that match your description, visits each website, and extracts its details into a candidate listing. Instead of adding listings one by one, you describe what you're looking for, then choose which candidates to publish.
Example: You're building a restaurant directory for Chicago. Instead of searching and copying data by hand, you tell AI Discovery: "Find Italian restaurants in downtown Chicago." It returns candidates with their name, website, description, contact details, location and categories, ready to review and publish.
How It Works
1. Search
AI searches the web for businesses that match your description.
2. Extract
AI reads each business's website and fills in the details: name, website, description, logo, email, phone, contact name, social links, street address, city, state, postal code, country and categories, with a relevance score.
3. Review
You approve or reject each candidate. One click publishes only the candidates scoring at least 70 for relevance (and any you approved); lower scores wait for your approval. Published listings go live on your directory.
Credit Costs
AI Discovery uses credits from your account. A job needs at least 13 credits to start, and credits are used as each step runs:
| Action | Credits | Description |
|---|---|---|
| Discovery job | 10 | Once per job you start |
| Web search | 1 | Once per job |
| Website scrape | 1 | Per business found |
| AI extraction | 1 | Per business found |
| Enrichment | 2 | Per listing you enrich (one scrape and one extraction) |
Every plan includes credits (see pricing). Paid plans and credit packs are available now, in Settings > Billing.
Writing Effective Prompts
The quality of your results depends on your search prompt. Here are some examples:
Find Italian restaurants in downtown Chicago with outdoor seating
Find restaurants
Be specific about location, category, and unique attributes.
Find coworking spaces in Austin, TX that offer day passes
Coworking Austin
Include specific features or services you want to find.
Find pet-friendly hotels in San Francisco under $200/night
Hotels SF
Price ranges and filters help narrow results.
Review First vs. Automatic
Review First (default)
Every candidate waits for you. Approve or reject each one and create listings from the approved ones. One click publishes only the candidates scoring at least 70 for relevance (and any you approved); lower scores wait for your approval.
- Full control over what gets published
- Catch and fix errors before going live
- Recommended for new directories
Automatic
Candidates with a relevance score of 70 or more are approved for you, so on the review page you create listings from all of them in one click; lower scores wait for your approval. Good for large jobs from a clear description.
- Less reviewing
- Large batches published in one step
- Requires trust in AI accuracy
Best Practices
Start with Review First
Start in Review First mode so you can check what AI found before anything is published. Once you trust the results, try Automatic.
Be Specific in Prompts
The more specific your search prompt, the better the results. Include location, category, and any unique attributes you're looking for.
Run in Batches
Instead of one massive search, run multiple smaller, focused searches. This gives you better control and more relevant results.
Review Extracted Data
AI extraction is powerful but not perfect. Always review key fields like phone numbers and addresses before approving.
Use Enrichment Sparingly
Enrichment is great for filling in gaps, but it costs credits. Use it strategically on your most important listings.
Monitor Credit Usage
The AI Builder shows a warning when you have used 80% of your credits. A job only processes as many businesses as your credits cover.
Ready to try AI Discovery?
Start a discovery job from the AI Builder in any of your directories.