Google Patents Records, Claims & Citation Graph

Read Google Patents without cookies, logins or an API key. Search the worldwide patent index by keyword, inventor, assignee or CPC class, or open known publications by number or URL. Return structured records with claims, legal status, legal events, citations, cited-by, family members, classifications, figures and PDF links.

Modes

Mode	What it does
searchPatents	Keyword search with office, language, status, type, date, sort, grouping and litigation filters.
byPatent	Full record for one or more publication numbers.
byInventor	Portfolio search for one or more inventor names.
byAssignee	Portfolio search for one or more companies or applicants.
byClassification	CPC class or prefix search. Use searchPatents to combine a keyword query with a class.
byCitation	Walk backward citations, forward cited-by links, or both, up to three hops.
byFamily	One row per family member or family publication.
byUrl	Paste patent pages or Google Patents search URLs; all search filters in the URL are kept.

Input reference

Field	Type	Default	Description
mode	string	searchPatents	searchPatents, byPatent, byInventor, byAssignee, byClassification, byCitation, byFamily or byUrl.
searchQueries	array	["quantum computing"]	Keyword queries for searchPatents.
publicationNumbers	array	[]	IDs for byPatent, byCitation and byFamily.
inventors	array	[]	Names for byInventor, or an optional search filter.
assignees	array	[]	Companies for byAssignee, or an optional search filter.
classifications	array	[]	CPC codes for byClassification, or an optional search filter.
urls	array	[]	Patent or search URLs for byUrl.
countryCodes	array	[]	Two-letter office codes; empty means all offices.
documentLanguages	array	[]	Original publication languages (16 values).
detailLanguage	string	en	Detail-page language.
status	string	ANY	ANY, GRANT or APPLICATION.
patentType	string	ANY	ANY, PATENT or DESIGN.
dateType	string	priority	priority, filing or publication.
dateFrom	string	""	Earliest date, inclusive, YYYY-MM-DD.
dateTo	string	""	Latest date, inclusive, YYYY-MM-DD.
sortBy	string	relevance	relevance, new or old.
grouping	string	family	family or publication.
includeSimilar	boolean	true	Keep Google's similar documents or drop them.
includeNonPatentLiterature	boolean	false	Also return Google Scholar rows.
litigation	string	ANY	ANY, YES or NO.
citationDirection	string	both	backward, forward or both.
citationDepth	integer	1	Citation hops, 1-3.
fetchFullDetails	boolean	false	Fetch each search hit's patent page.
maxItems	integer	25	Hard cap across all queries; prefill is 3.
maxItemsPerQuery	integer	0	Optional per-query cap; 0 means global only.

Filters

• Multiple values: inventor and assignee arrays run one search per value and merge (Google's XHR keeps only the last repeat value); CPC arrays are ORed in one query; office and language arrays are comma-joined. Different filter types are combined with AND.
• Patent office: countryCodes accepts US, EP, WO, JP, CN, KR, DE, GB, FR, CA, AU, IN and more.
• Document language: 16 original-publication languages. This is a language filter, not a translation setting.
• Grant status: any, granted, or applications.
• Type: any, patent, or design.
• Date: priority, filing or publication date, from and to.
• Sort: relevance, newest, or oldest.
• Grouping: one row per family (Google's default) or one row per publication.
• Similar documents: keep them with matchType=similar, or drop them.
• Litigation: with related litigation, without known litigation, or any.
• Non-patent literature: optionally add Google Scholar rows for each keyword query.

Output

Patent rows include publicationNumber, title, abstract, claims, claimCount, inventors, assigneesOriginal, assigneesCurrent, primaryAssignee, priorityDate, filingDate, publicationDate, grantDate, priorArtDate, expirationDate where the source has one, applicationNumber, legalStatus, legalEvents, cpcCodes, classificationDetails, priorArtKeywords, familyId, familyMembers, familyApplications, familyCountryStatus, priorityApplications, publications, alsoPublishedAs, backwardCitations, forwardCitations, familyCitations, familyCitedBy, nonPatentCitations, similarDocuments, figures, thumbnailUrl, pdfUrl, externalLinks, patentUrl, matchType, searchRank, searchQuery, recordType and scrapedAt.

Rows that do not apply are omitted; the actor never writes null, empty-string or sentinel values. Citation rows add citationDirection, citationSeed, citationHop, citationDepth, citationParent, citedBy or citing. Family rows add familyId and sourceSeed. Non-patent-literature rows add authors, venue, urlHostname and scholarId.

Some data is genuinely missing upstream: for example, EP B1 documents and many Japanese grants have no abstract, not every office publishes claims, and an anticipated expiration date is shown mainly for US and Canadian patents. Those fields are omitted rather than guessed.

Examples

Keyword search with full details:

{
  "mode": "searchPatents",
  "searchQueries": ["solid-state battery"],
  "countryCodes": ["US", "EP"],
  "status": "GRANT",
  "dateFrom": "2020-01-01",
  "fetchFullDetails": true,
  "maxItems": 20
}

Company portfolio in one office:

{
  "mode": "byAssignee",
  "assignees": ["Siemens"],
  "countryCodes": ["DE", "EP"],
  "maxItems": 50
}

Citation graph two hops out:

{
  "mode": "byCitation",
  "publicationNumbers": ["US11704715B2"],
  "citationDirection": "both",
  "citationDepth": 2,
  "maxItems": 100
}

Paste a search URL:

{
  "mode": "byUrl",
  "urls": ["https://patents.google.com/?q=lidar&country=US&status=GRANT&after=filing:20200101"]
}

Limits

Google Patents shows about 1,000 results per query (100 pages). The actor pages through that window and tells you in the run status when it hits the source cap. To go deeper, split the search by date, office or CPC class. The run status also reports the delivered count. maxItems is a hard cap across all queries; maxItemsPerQuery optionally caps each query separately.

For keyword searches sorted newest or oldest, the actor reads the accessible result window and sorts it by publication date. Google's sorted endpoint otherwise ignores bare keyword terms. Non-patent-literature rows share the same maxItems budget and are written after patent rows.

• byFamily seeds are enriched with title, grant status and legal status from the patent page plus a publication lookup; non-seed family rows carry title/status only when Google lists them.
• Similar documents carry publicationDate when Google shows one. Non-patent-literature rows are text/venue/URL records with no structured date; a date may appear in the text.
• In citation runs of depth 2 or 3, every row has citationParent (the page it was read from). citedBy/citing remain seed-based for compatibility.

FAQ

Do I need a Google account or API key? No. The actor reads public Google Patents pages. No cookies, login or key are required.

Why did my search stop at about 1,000 rows? That is Google Patents' display cap for one query. Narrow the date range, choose fewer offices, or add a CPC class and run again.

Does it return claims for every patent? Only when the source page has them. US, EP, WO, CN and several other offices usually do; some applications, search-report documents and designs do not.

Why is the abstract empty on some records? Some documents have no abstract on Google Patents, including many EP B1 grants and Japanese grants. The field is omitted rather than filled with a placeholder.

What is the difference between family and publication grouping? Family grouping returns one row per invention, as Google shows by default. Publication grouping returns separate rows for the application, grant and each office's version.

Are citation walks recursive? Yes in byCitation mode: set citationDepth to 1, 2 or 3. Each row records the seed, hop and direction.

Do multiple inventors or CPC codes work in one search? Yes. Inventor/assignee arrays run one search per value and merge the results; CPC arrays are ORed in one query. Duplicate publications are removed.

Do the PDF and image links work without a referer? Yes. They point to public Google Cloud Storage objects.

Can I filter by original document language? Yes, with documentLanguages. It filters publications by language; it does not translate them. Use detailLanguage to choose the page language for detail extraction.

Does it use a proxy? The actor tries a direct request first and switches to the Apify proxy automatically if Google blocks it. There is no proxy input to configure.

Notes

Patent documents are public records. Legal status and assignee data are Google's assumptions, not legal conclusions; confirm critical facts with the relevant patent office. This actor is independent and not affiliated with Google.