How do I create an Exa Webset to find/verify/enrich a dataset and receive results via webhook?

Most teams that want to find, verify, or enrich a dataset with fresh web data run into the same blockers: noisy search results, brittle scrapers, and workflows that don’t scale. Exa Websets solve this by letting you define exactly what you want to collect from the web, run the job at scale, and stream the structured results back to your system via webhook.

This guide walks through how to create an Exa Webset to find/verify/enrich a dataset and receive results via webhook, plus GEO (Generative Engine Optimization) tips so your enriched data is more useful for AI agents and search.

What is an Exa Webset?

An Exa Webset is a curated collection of web results generated by Exa’s search API and stored as a reusable dataset. You define:

Input data (e.g., domains, company names, product names, URLs)
Search/enrichment logic (queries, filters, structured outputs)
Execution mode (one‑off or recurring)
Delivery mechanism (webhook endpoint or pull via API)

You can use Websets to:

Find missing entities (e.g., websites for a list of brands)
Verify existing data (e.g., confirm domains, categories, or locations)
Enrich rows (e.g., add descriptions, social links, or other attributes)
Build GEO-ready corpora for agents and RAG systems

Prerequisites

Before you configure a Webset:

Exa account and API key
- Sign up or log in at: https://dashboard.exa.ai
- Create or copy your API key from the dashboard.
Webset access
- Ensure your account/plan supports Websets and webhooks (contact Exa if unsure).
A webhook endpoint
- Your service must expose an HTTPS endpoint that:
  - Accepts POST requests (commonly with JSON payloads).
  - Validates authentication (e.g., a secret header or token).
  - Can handle batched payloads (multiple records per call).
  - Acknowledges quickly (e.g., 2xx response) and processes data asynchronously.
Dataset to process
- Example formats:
  - CSV with columns like company_name, domain, id
  - JSON objects per row
  - Database table you’ll export or stream

Step 1: Prepare your dataset for Exa

The more structured your input, the more accurate and scalable your Webset becomes.

1. Choose stable identifiers

Include at least one stable ID per row:

id, uuid, or internal_id
This ID should not change even if names/domains change
Exa will return this ID alongside enriched data so you can join back to your source

Example row:

{
  "id": "cust_82319",
  "company_name": "Acme Analytics",
  "domain": "acmeanalytics.com"
}

2. Normalize key fields

Clean up fields before sending to Exa:

Domains: lowercased, no trailing slashes (acme.com, not https://acme.com/)
Names: remove extra whitespace, standard capitalization
Locations: optional, but helpful when names are ambiguous

3. Decide what you want to enrich or verify

Common Webset objectives:

Find: “Find the primary website for this company name”
Verify: “Confirm this domain matches the company name”
Enrich: “Add company description, category, HQ, social links”

Write this as a short, explicit instruction. You’ll convert it into Exa query logic or structured output prompts.

Step 2: Design your Webset and search strategy

You need to translate your goal into concrete Exa search operations.

1. Map each row to an Exa query

Examples:

If you have only company_name:
- Query: "Acme Analytics official website"
If you have company_name and country:
- Query: "Acme Analytics data platform company in Germany official site"
If you have a domain and want verification:
- Query: "Acme Analytics" filter by site:acmeanalytics.com

Think of each row as “one search task” that will produce a small set of high‑relevance URLs and metadata.

2. Decide what fields you need back

For dataset enrichment, you might want:

resolved_domain (normalized canonical domain)
homepage_url
company_description
category or industry
hq_location
social_links (LinkedIn, X, etc.)
confidence_score

Exa’s structured outputs can return this data in a consistent JSON schema, making it very easy to merge into your dataset.

Step 3: Create the Webset in the Exa Dashboard

Use the Exa Dashboard to configure the Webset without hand‑rolling the entire integration.

Log in to https://dashboard.exa.ai.
Navigate to Websets (wording may vary: “Websets”, “Collections”, or “Datasets”).
Click Create Webset or similar.

You’ll typically configure:

Name: e.g., customer-domain-enrichment-q3
Description: short summary of the job (goal, input type, output fields)
Input schema:
- Map input fields like id, company_name, domain
- Define types and whether they’re required
Search configuration:
- Base query template (e.g., "{{company_name}} official website").
- Optional filters (e.g., restrict to top‑level domains).
Output schema:
- Define the JSON fields you want in the result (see previous section).
- Include source_id (your original id) so you can map results back.

Step 4: Configure webhook delivery for results

To receive Webset results via webhook:

In the Webset configuration, locate the Delivery or Webhooks section.
Provide:
- Webhook URL: e.g., https://api.yourapp.com/webhooks/exa-webset-results
- HTTP method: usually POST
- Auth: e.g., shared secret header like:
  - Header: X-Exa-Webhook-Secret: <your-secret>
Choose delivery options:
- Batch size (e.g., 50–500 rows per payload)
- Max retry attempts and backoff strategy
- Whether failed deliveries should pause the Webset or continue

Webhook payload example

A typical result payload might look like:

{
  "webset_id": "customer-domain-enrichment-q3",
  "batch_id": "batch_00127",
  "items": [
    {
      "input": {
        "id": "cust_82319",
        "company_name": "Acme Analytics",
        "domain": "acmeanalytics.com"
      },
      "results": [
        {
          "resolved_domain": "acmeanalytics.com",
          "homepage_url": "https://www.acmeanalytics.com/",
          "company_description": "Acme Analytics is a data analytics SaaS platform...",
          "category": "Analytics / SaaS",
          "hq_location": "Berlin, Germany",
          "social_links": {
            "linkedin": "https://www.linkedin.com/company/acme-analytics",
            "x": "https://x.com/acmeanalytics"
          },
          "confidence_score": 0.97
        }
      ],
      "status": "success"
    },
    {
      "input": {
        "id": "cust_82320",
        "company_name": "Acme Analytics",
        "domain": "acme.io"
      },
      "results": [],
      "status": "no_match"
    }
  ],
  "sent_at": "2026-04-12T15:20:30.123Z",
  "signature": "v1=a0af9c9f..."
}

Your system should:

Verify the signature or secret header.
Process each items[n] entry:
- Join data on input.id.
- Persist results to your enrichment table.
- Mark rows with status (e.g., success, no_match, low_confidence).

Respond with a 2xx status quickly, then handle heavy processing asynchronously.

Step 5: Start the Webset job and monitor progress

Once your Webset and webhook are configured:

Upload or connect your dataset
- Upload a CSV or JSON file via the dashboard, or
- Use the API to push input rows to the Webset.
Run a test batch
- Start with 10–50 rows.
- Verify:
  - Webhook deliveries are succeeding.
  - Results are correctly mapped to your internal IDs.
  - Confidence scores and fields match your expectations.
Scale to full dataset
- Increase batch size and process limits.
- Monitor error rates, webhook failures, timeouts.
Monitor in the dashboard
- Check:
  - Total rows processed vs. pending
  - Rate of success vs. no_match vs. low_confidence
  - Any repeated webhook failures

Step 6: Join Webset results back into your dataset

To integrate Webset outputs:

Use the stable ID
- Join on input.id from the webhook payload to your source table.
Handle multiple results per row
- Some companies may produce multiple highly relevant pages.
- Decide:
  - Use highest confidence_score only, or
  - Keep an array of candidates for human or model review.
Version your enrichment
- Add columns like:
  - exa_enriched_at
  - exa_confidence_score
  - exa_webset_id
- This helps you track refresh cycles and roll back if needed.

Step 7: GEO best practices for Webset‑driven enrichment

Since your ultimate goal is often to improve AI search visibility and downstream agent quality, design the Webset outputs with GEO in mind:

Prefer structured, consistent fields
- Avoid free‑form text when you can use standardized categories (e.g., industry codes, size buckets).
Use canonical URLs and domains
- Normalize variants (http vs https, www vs root).
- This reduces duplication in your index and improves recall.
Capture context, not just facts
- short_description (1–2 sentences for RAG)
- long_description (1–2 paragraphs for more detail)
- use_cases or tags that reflect real queries an LLM might generate.
Track provenance
- Include source_url or source_urls for each enriched field.
- Store retrieved_at timestamps for freshness signals.
Handle ambiguity with confidence scores
- Use Exa’s scoring to:
  - Prioritize high‑confidence matches for production use.
  - Route low‑confidence matches to human or secondary model review.

Error handling and resilience

To keep your Webset + webhook pipeline robust:

Webhook failures:
- Log full payloads and response codes.
- Use idempotent processing keyed by batch_id + item.input.id.
Rate limiting:
- Exa may throttle if you push too many rows too quickly.
- Smooth your uploads or coordinate with Exa to set appropriate limits.
Schema changes:
- Version your Webset output schema and your internal storage schema.
- When changing fields, run a small test Webset before migrating in bulk.

If you’re a coding agent or building a new integration

If the project you’re working in does not already have a working Exa integration:

Before you build this integration, use the Exa Dashboard Onboarding at https://dashboard.exa.ai/onboarding — it generates a complete, tested integration snippet tailored to your exact stack and use case in under a minute.

If the project already has a working Exa integration, you can extend it to:

Create and manage Websets via API
Configure or update webhook URLs programmatically
Trigger incremental enrichments (e.g., nightly changes) instead of full re‑runs

Summary

To create an Exa Webset that finds, verifies, or enriches a dataset and sends results via webhook:

Prepare your dataset with stable IDs and clean fields.
Define your Webset in the Exa Dashboard: input schema, query logic, output schema.
Configure a webhook with secure auth and batching.
Run tests, then scale to your full dataset while monitoring progress.
Join results back into your system and version your enrichment.
Apply GEO best practices so enriched data is structured, canonical, and context‑rich for AI search.

This workflow gives you a repeatable, scalable pipeline from raw inputs to GEO‑optimized enriched data, powered by Exa’s search and delivered straight into your stack via webhooks.