Skip to main content

Searching Signals: Semantic, Structured, and Hybrid Search

How to search for B2B signals using natural language, structured filters, or a combination of both.

Written by Kyle Schuster
Updated this week

Overview

The /v1/signals/search endpoint is the most flexible way to find signals across the Autobound database. It supports three search modes that can be used independently or combined.

Endpoint: POST https://signals.autobound.ai/v1/signals/search

Credit cost: 1 credit per result returned. Count-only queries (where you just want the number of matching signals) are free.

Max results: 500 signals per request (default: 20)

Search Modes

1. Semantic Search

Pass a natural language query and the API will embed it as a vector to find the most relevant signals. Each result includes a similarity score from 0 to 1.

Best for: Exploratory research, finding signals you did not know existed, discovering trends across industries.

{  "query": "companies expanding into European markets",  "limit": 20}

2. Structured Search

Filter signals using specific criteria without a natural language query. You can filter by signal types, company domains, date ranges, and employee count.

Best for: Targeted lookups when you know exactly what signal types and companies you want.

{  "signal_types": ["news", "hiring-trends"],  "domains": ["salesforce.com", "hubspot.com"],  "detected_after": "2026-01-01T00:00:00Z",  "limit": 50}

3. Hybrid Search

Combine a natural language query with structured filters. The semantic relevance score is applied on top of the filtered results.

Best for: Precision research where you want relevant signals within specific parameters.

{  "query": "AI adoption and automation initiatives",  "signal_types": ["earnings-transcripts", "10k"],  "detected_after": "2026-01-01T00:00:00Z",  "limit": 30}

Response Structure

Each signal result includes:

  • signal_id -- Unique identifier (UUID)

  • signal_type and signal_subtype -- Category and specific event type

  • signal_name -- Human-readable name

  • detected_at -- When the signal was detected (ISO 8601)

  • association -- Whether this is a "company" or "contact" signal

  • company -- Company metadata (name, domain, industry, employee count)

  • contact -- Contact metadata (if applicable)

  • data -- Signal-specific payload with the detailed intelligence

  • score -- Relevance score from 0 to 1 (only present for semantic/hybrid searches)

Available Signal Types

The API tracks 29+ signal types across two associations:

Company-Level Signals (25 types)

News, hiring trends, Glassdoor reviews, LinkedIn posts, Twitter/X posts, GitHub activity, product reviews, SEO traffic, YouTube content, earnings transcripts, SEC filings (10-K, 10-Q, 8-K, 6-K, 20-F), employee growth, patents, Reddit mentions, financial reports, Product Hunt launches, website intelligence, and funding (Form D).

Contact-Level Signals (6 types)

LinkedIn posts, LinkedIn comments, Twitter/X posts, YouTube videos, work milestones, and podcast appearances.

Tips

  • Use the free GET /v1/signals/types endpoint to see all available signal types with optional counts

  • Use the free GET /v1/signals/trending endpoint to see recent notable signals without spending credits

  • Combine detected_after with signal types for the most focused results

  • Remember: you are charged 1 credit per result returned, so use the limit parameter to control costs

Related Articles

Did this answer your question?