Full text

Abstracts and source linksCopy Anchor Link

Search results can include an abstract, a source landing-page url, and a pdf_url. Any of these fields may be null when the provider does not supply them. An available PDF link does not guarantee download access or a particular license.

Request source-content extractionCopy Anchor Link

Set enrichment to "high" when creating a search to read each returned source in full. Deep search uses high by default:

{
  "text": "How does sleep affect memory consolidation?",
  "enrichment": "high"
}

For each source, the API fetches its landing page and, when there is one, its PDF, then converts them to text. The analysis (relevance rating, explanation, quoted evidence and paper type) is written from that text rather than from the abstract alone.

Retrieve the search using its job_id, then inspect each entry in the sources map. The parsed_source field contains the extracted text, up to about 12,000 characters. It can instead contain the abstract as a fallback, or be null, for example when a publisher blocks access. It is not a guarantee of the complete paper.

With high, the sources are also ranked again after they have been read. See How search works and the enrichment fields. Full-text analysis is priced per returned source; see Credits and pricing.

Filter for PDF availabilityCopy Anchor Link

You can set filters.open_access to true to require a non-empty PDF URL in source metadata. This filter does not verify access rights or the extracted content. See the filter reference.

Source-content extraction is an option on Create a search. There is no separate full-text download endpoint in the initial API.