In short. Exa Search API returns ranked web results whose contents options select page text, query-relevant highlights, or generated summaries.

Definition: As of 17 October 2026, Exa Search API is a web search interface that accepts a natural language query and returns ranked results with optional page content. Its contents options select extracted text, relevant excerpts called highlights, or generated summaries.

Key facts

Exa's Search API guide identifies query as the only required request field and auto as the default search mode.
The Search reference defines result fields including title, url, publishedDate, author, and id.
Exa's Highlights guide describes query-relevant passages returned in results[].highlights.
The Contents API guide documents extraction from supplied URLs, including JavaScript-rendered pages and PDFs.

How does a search request work?

Exa's Search API guide shows an authenticated HTTP POST request to https://api.exa.ai/search. This example follows its request structure and asks for highlights:

curl -s -X POST "https://api.exa.ai/search" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $EXA_API_KEY" \
  -d '{
    "query": "Exa Search API documentation",
    "type": "auto",
    "contents": { "highlights": true }
  }'

EXA_API_KEY holds the authentication key. The response contains a results array ranked by relevance. The guide documents a default of up to 10 results; numResults can request up to 100, with fewer returned when fewer relevant pages are available.

Which retrieved content fields matter?

The Search response schema gives each result a title and URL. It describes publishedDate as an estimated creation date obtained by parsing HTML, author as available when present, and id as a temporary document identifier usable with /contents.

The content fields have different purposes:

text contains extracted page text. Exa's Contents API guide describes this output as the clean page body in Markdown; text.maxCharacters limits its length.
highlights contains an array of extracted passages. The Highlights guide says selection runs against the query on each request. highlights.maxCharacters sets a fixed per-page excerpt limit when needed.
summary contains a generated overview. The Contents API guide documents a language model call for each page and a summary.query option to guide the summary. A summary.schema option requests structured fields, returned as a JSON string.

For a website review, url identifies the returned page, while text or highlights exposes the content supplied to the calling application.

How does content retrieval handle existing URLs and freshness?

The Contents API guide documents POST /contents for known URLs or document IDs, supplied in ids. Its text, highlights, and summary options sit at the request's top level. Search places those options inside contents.

The same guide defines maxAgeHours: omitting it allows cached content and fetches pages when needed; a positive value sets the acceptable cache age in hours. Setting it to 0 requests fresh content, while -1 uses only cached content. The /contents response includes statuses for checking each URL's success or failure.

FAQ

Are highlights generated summaries?

Exa's Contents API guide describes highlights as passages copied from the page. If you need a generated overview, select summary.

Can I control which domains appear?

The Search reference documents includeDomains and excludeDomains. You can supply hostnames, path prefixes, or wildcard subdomains to filter results.

Sources

Check how a page reads to answer engines. Open the GEO score analyzer