Google Patents Scraper

Search Google Patents and pull complete patent records — title, abstract, claims, classifications (CPC/IPC), inventors, assignees, priority/grant dates, family jurisdictions, citation graph, and PDF link — all from the public Google Patents endpoints.

No login, no API key, no captcha-solving service required.

What you can do

• Search patents by free-text query (same syntax as the Google Patents site search).
• Lookup a single patent by publication number (e.g. US11704715B2, EP3789012A1, WO2020012345A1).
• Search by inventor name.
• Search by assignee / company.
• Search by CPC or IPC classification code (e.g. H04L9/06, G06N3/00).
• Walk the citation graph — list every patent a given patent cites (backward) and every patent that cites it (forward).
• Fetch by URL — paste any patents.google.com URL (search result, patent page).

Output fields

Every record contains as many of these fields as Google Patents exposes for the patent:

Field	Type	Description
publicationNumber	string	Canonical publication number (e.g. US11704715B2).
country	string	Filing-jurisdiction prefix (US, EP, WO, JP, ...).
title	string	Patent title.
abstract	string	Full abstract paragraph.
description	string	Specification text (truncated at 50k chars).
claims	string[]	Each claim as a separate string.
claimCount	integer	Number of claims.
inventors	string[]	All listed inventors.
primaryInventor	string	First listed inventor.
assigneesOriginal	string[]	Original assignees at filing.
assigneesCurrent	string[]	Current assignees after re-assignment.
primaryAssignee	string	First assignee.
priorityDate	string (YYYY-MM-DD)	Earliest priority date.
filingDate	string
publicationDate	string
grantDate	string	When granted.
expirationDate	string	When applicable.
applicationNumber	string
legalStatus	string	Active / Expired / Pending — when shown.
classifications	string[]	CPC + IPC codes.
backwardCitations	object[]	Patents this one cites (publication number, title, dates, assignee).
forwardCitations	object[]	Patents citing this one.
backwardCitationCount	integer
forwardCitationCount	integer
familyJurisdictions	object[]	Per-country status of family members.
language	string	Language of the rendered metadata.
pdfUrl	string	Public Google Cloud Storage PDF URL.
thumbnailUrl	string	Front-page figure image URL.
figureUrls	string[]	All embedded figures (search-only).
patentUrl	string	Canonical patents.google.com URL.
recordType	string	patent or citation.
siteName	string	Google Patents.
scrapedAt	string (ISO-8601)

Modes & input

Mode	What it does	Required input
searchPatents	Free-text search across all of Google Patents.	query
byPatent	Fetch one patent (or many via publicationNumbers).	publicationNumber or publicationNumbers
byInventor	All patents naming a specific inventor.	inventor
byAssignee	All patents assigned to a company.	assignee
byClassification	All patents under a CPC / IPC code.	classification
byCitation	Forward and/or backward citations of a patent.	publicationNumber, citationDirection
byUrl	Fetch any patents.google.com URL.	urls

Filters (all modes)

• country — restrict to a specific patent office (US, EP, WO, JP, CN, ...).
• language — translation locale for rendered metadata.
• status — granted only / applications only / any.
• sortBy — relevance / newest / oldest.
• filingDateFrom, filingDateTo, publicationDateFrom, publicationDateTo — YYYY-MM-DD bounds.
• fetchFullDetails — for search modes, also fetch each patent's full HTML for abstract/claims/citations (1 extra request per patent).
• maxItems — hard cap.
• useProxy / proxyConfiguration / autoEscalateOnBlock — proxy controls. Direct fetches usually work; the actor auto-escalates to Apify proxy if Google starts returning 403/429.

Daily test prefill

mode=searchPatents, query="quantum computing", maxItems=10 — produces ≥10 records on every run.

Reliability

• HTTP-only via curl_cffi (Chrome 131 fingerprint). No browser, no Playwright.
• Retries on 429/5xx with exponential backoff.
• Auto-escalates to Apify proxy on the first 403/429.
• Recursive omit-empty before every record push (no null / "" / [] fields).

FAQ

Q: Does this scrape USPTO directly?
No — Google Patents already aggregates USPTO + EPO + WIPO + JPO + CNIPA + KIPO + 80+ other patent offices. Querying Google Patents is broader and faster.

Q: How many patents can I get per run?
maxItems is capped at 1000. Beyond that, split into multiple runs with date-range filters.

Q: Does fetchFullDetails slow things down?
Yes — adds ~1 second per patent. Leave it off if you only need the search snippet (title, dates, primary inventor/assignee, thumbnail).

Q: PDF URLs — do they work directly?
Yes. Patent PDFs are public on patentimages.storage.googleapis.com with no auth or hotlinking restriction.

Q: Citation traversal — recursive?
No, single hop. Provide one or more publication numbers, get their direct citations. Run again with the new IDs to walk further.

Q: Why is my non-English query returning English results?
Google Patents detects query language automatically. Set language=de (or another) to bias the rendered metadata.