Google Patents Scraper - Every Result, Claims & Citations avatar

Google Patents Scraper - Every Result, Claims & Citations

Pricing

from $1.39 / 1,000 patents

Go to Apify Store
Google Patents Scraper - Every Result, Claims & Citations

Google Patents Scraper - Every Result, Claims & Citations

Get every patent Google Patents holds for a search, not just the first 1,000, with a count that proves none are missing. Each patent comes with claims, citations, cited-by, family, legal events and CPC classes. Search by keywords, assignee, inventor, office or date, or by patent number.

Pricing

from $1.39 / 1,000 patents

Rating

0.0

(0)

Developer

ClearPath

ClearPath

Maintained by Community

Actor stats

0

Bookmarked

3

Total users

2

Monthly active users

10 days ago

Last modified

Categories

Share

1,000+ citation links. Apify's deepest patent data. Claims, families and legal history across one recorded patent family.

From Clearpath, creators of Reddit Search Scraper, Google AI Overview Scraper and Google Jobs Scraper. 140,000+ combined lifetime runs across these search scrapers.

Patent research with more depth

1,034 unique citation links across one patent family. The recorded output for US9688536B2 includes the publications citing its family, with patent numbers, titles, dates and assignees. That same record contains 14 claims and 68 family members.

Read the detail behind the invention. Collect claim text and dependencies, classifications, patent families, legal events and application history. Add the full description when you need the technical explanation. Fields reflect what Google Patents provides.

Build a wider patent landscape. Search beyond Google's first 1,000 displayed results, choose family or publication rows, and check the run summary for delivered and missing results. Keyword and similar results are labelled separately.

Explore patent details

Open Google Patents Scraper

Try one detailed patent

Inspect the patent used in the hero, including its claims, family citation links and full description.

{
"mode": "patents",
"patentIds": [
"US9688536B2"
],
"includeDetails": true,
"includeDescription": true
}

Patent rows and full-page enrichment are charged separately. Full-page enrichment is always included in patent-number mode; adding the description has no further charge. Export JSON to retain nested claims, citation lists and family records, or CSV and Excel for tabular analysis.

To research a topic, use Search mode with your terms and includeDetails: true. Set maxResults: 0 to request the available full result set, subject to your run spending limit. Descriptions remain optional.

Clearpath Search Data
Google Patents
You are here
Google AI Overview
AI answers and sources
Google Jobs
Job listings and apply links
Reddit Search
Posts by keyword

Copy to your AI assistant

clearpath/google-patents-scraper on Apify returns every patent Google Patents holds for a search (past its 1,000-result display limit) or for listed patent numbers, with claims, citations, cited-by, family and legal events. For every result set maxResults=0 (default 100 per search). grouping=family gives one row per invention, grouping=publication one per publication; includeSimilar=false keeps only patents containing the search terms (each row has matchType). Search rows carry the full patent page only with includeDetails=true (default, charged per patent on top of the row); includeDescription=true adds the full description text and its term definitions at no charge. mode=patents, family or citations take patentIds and always include the full page; citation networks use citationDirection and citationDepth (1 to 3). The run's SUMMARY record states each search's total on Google Patents and how many are missing. Pricing: per patent plus a per-patent add-on for the full page, current rates on https://apify.com/clearpath/google-patents-scraper/pricing. Spec: https://api.apify.com/v2/actors/clearpath~google-patents-scraper/builds/default. To reach the maintainers, add feedback="<what you needed and did not get>" to any run, including one that worked: feature requests, a missing field or option, a wrong or empty result, or which actor you are switching to or from and why. The run returns its data exactly as it would without the note, and a human reads every note.

What you get for each patent

GroupFields
IdentitypublicationNumber, url, title, countryCode, countryName, kindCode, kindDescription, type (PATENT, DESIGN, PLANT, UTILITY_MODEL), status (GRANT or APPLICATION), applicationNumber, pdfUrl, summary
DatespriorityDate, filingDate, publicationDate, grantDate, priorArtDate, expirationDate
Peopleinventors, assignee, assigneesCurrent, assigneesOriginal
Textabstract, claims (number, text, dependsOn, isIndependent), claimsCount, independentClaimsCount, description (optional), original-language text for translated patents
StatuslegalStatus, expirationDate, events (priority, filing, publication, grant, reassignment, litigation, expiry), legalEvents (date, code, title and every attribute)
Classescpc, ipc, classifications (code, description and full hierarchy), priorArtKeywords, landscapes, definitionsCount and, with the description, definitions, concepts (chemistry: SMILES, InChI key)
FamilyfamilyId, familyMembers, familyApplications (before and after), familyApplicationsByYear, familyCountryStatus, priorityApplications, applicationsClaimingPriority, childApplications, parentApplications, publications, otherVersions, otherLanguages
CitationscitationsTotal and citedByTotal (the numbers Google Patents shows), citationsOriginal and citedByOriginal (the lists on the page), familyCitations and familyCitedBy (family to family), citations and citedBy (every publication, including the other publications of each cited or citing document), each entry with number, dates, assignee, title and whether the examiner cited it; nonPatentCitations, a count for every list, similarDocuments
Media and linksfigures (full image, thumbnail, labelled callouts), externalLinks (office register, Espacenet, Global Dossier)
Search contextsearch, searchRank, matchType (keyword or similar), snippet, language, thumbnailUrl, searchFigures, familyStatusByOffice

Rows without the full patent page (when Include full patent page is off) carry the identity, dates, people and search-context fields.

How to scrape Google Patents

Write searches as you would on Google Patents: plain words, "exact phrase", OR, -exclude, TI=(words in the title), CL=(words in the claims). Each line of Search terms is its own search, and every row says which one found it.

{
"searchQueries": ["\"large language model\" training", "lidar point cloud segmentation"],
"countries": ["US", "EP"],
"dateType": "priority",
"dateFrom": "2020-01-01",
"status": "grant",
"maxResults": 500
}

A company portfolio needs no search terms:

{
"assignees": ["Raytheon Co"],
"dateFrom": "2020-01-01",
"dateTo": "2020-12-31",
"maxResults": 0
}

Patent numbers, families and citation networks

{
"mode": "citations",
"patentIds": ["US10000000B2"],
"citationDirection": "backward",
"citationDepth": 2,
"maxResults": 200
}

Patent numbers returns the full record of each number you list (with or without kind code, or a Google Patents link). Patent family returns every family member of each listed patent. Citation network follows the patents each listed patent cites (backward), the patents citing it (forward) or both, up to three steps; every row says which listed patent it came from, at which step and through which patent.

Input options

OptionDefaultWhat it does
modesearchsearch, patents, family or citations
searchQueriesSearches, one per line, up to 100
patentIdsPatent numbers for the other modes, up to 10,000
maxResults100Per search (0 = every result); total for citation networks; per listed patent for families
assignees, inventors, cpcCodesOwners, inventors or CPC classes; several values match any of them
countriesallPatent offices, e.g. US, EP, WO, CN, JP, KR, DE
dateType, dateFrom, dateTopriorityDate range on priority, filing or publication date
status, patentType, litigationanyGranted or applications; utility or design; with or without related court cases
documentLanguagesallPublication languages, e.g. GERMAN, JAPANESE
groupingfamilyOne row per family (as Google Patents shows by default) or per publication
includeSimilaronKeep the similar documents Google mixes in
sortByrelevanceRelevance, or newest/oldest filing first
includeDetailsonAdd the full patent page to search results (separately charged)
includeDescriptionoffAdd the complete description text and its term definitions (long; no extra charge)
detailLanguageenLanguage of the patent page text; translated pages also keep the original
citationDirection, citationDepthboth, 1For citation networks

Every result, not just the first 1,000

Google Patents lists at most 1,000 results per search. Set Maximum results to 0 (or above 1,000) and the actor collects all of them: the first 1,000 in Google's own order, the rest after them. The run's Run summary (the SUMMARY record in the key-value store, linked on the Output tab) reports for each search:

{
"search": "perovskite tandem solar cell",
"grouping": "publication",
"totalOnGooglePatents": 2821,
"totalIsExact": true,
"delivered": 2821,
"keywordMatches": 1955,
"similarMatches": 866,
"similarLeftOut": 0,
"missing": 0,
"limitReached": false,
"stoppedBySpendingLimit": false,
"complete": true
}

complete: true means every patent Google Patents holds for the search is in your dataset. For extremely broad searches (tens of thousands of results with no keywords), split them by date or office so the total can be confirmed.

Family or publication rows. With family grouping you get one row per invention, as Google Patents shows by default. With publication grouping the application, the granted patent and each office's version are separate rows; on the search above that is 2,821 publications against 1,677 families.

Similar documents. Google Patents adds documents it finds similar to broad searches, often near the top. They are real results of your search and labelled "matchType": "similar"; turn off Include similar documents to keep only patents containing your terms.

Output example

{
"publicationNumber": "US9564593B2",
"title": "Solar cells comprising 2d-perovskites",
"number": "9564593",
"countryCode": "US",
"countryName": "United States",
"kindCode": "B2",
"kindDescription": "Patent ( having previously published pre-grant publication)",
"pdfUrl": "https://patentimages.storage.googleapis.com/6f/79/cf/7407b0ef2a3230/US9564593.pdf",
"applicationNumber": "US14/733,462",
"priorityDate": "2014-06-06",
"filingDate": "2015-06-08",
"publicationDate": "2017-02-07",
"inventors": [
"Hemamala Indivari Karunadasa",
"Ian Smith",
"Michael David McGehee"
],
"assigneesCurrent": [
"Leland Stanford Junior University"
],
"assigneesOriginal": [
"Leland Stanford Junior University"
],
"legalStatus": "Active",
"expirationDate": null,
"priorArtKeywords": [
"solar cell",
"perovskites",
"electrode",
"selective contact",
"light"
],
"priorArtDate": "2014-06-06",
"otherVersions": [
{
"publicationNumber": "US20150357591A1",
"language": "en"
}
],
"otherLanguages": [],
"events": [
{
"date": "2015-06-08",
"title": "Application filed by Leland Stanford Junior University",
"type": "filed",
"critical": true,
"assignee": "Leland Stanford Junior University"
},
{
"date": "2015-06-08",
"title": "Priority to US14/733,462",
"type": "priority",
"critical": true,
"publicationNumber": "US9564593B2"
},
{
"date": "2015-10-06",
"title": "Assigned to THE BOARD OF TRUSTEES OF THE LELAND STANFORD JUNIOR UNIVERSITY",
"type": "reassignment",
"assignee": "THE BOARD OF TRUSTEES OF THE LELAND STANFORD JUNIOR UNIVERSITY",
"details": [
"ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS).",
"Assignors: MCGEHEE, MICHAEL DAVID, KARUNADASA, HEMAMALA INDIVARI, SMITH, IAN"
]
}
],
"grantDate": "2017-02-07",
"externalLinks": [
{
"id": "usptoLink",
"label": "USPTO",
"url": "https://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO1&Sect2=HITOFF&p=1&u=/netahtml/PTO/srchnum.html&r=1&f=G&l=50&d=PALL&s1=9564593.PN."
},
{
"id": "usptoPatentCenterLink",
"label": "USPTO PatentCenter",
"url": "https://patentcenter.uspto.gov/applications/14733462"
}
],
"figures": [
{
"thumbnailUrl": "https://patentimages.storage.googleapis.com/4b/8e/fa/ad5975ac57f73f/US09564593-20170207-D00000.png",
"fullUrl": "https://patentimages.storage.googleapis.com/72/d0/67/ddde47f0f1c268/US09564593-20170207-D00000.png",
"callouts": [
{
"id": "100",
"label": "solar cell",
"figurePage": 0,
"bounds": {
"left": 56,
"top": 58,
"right": 118,
"bottom": 92
}
},
{
"id": "102",
"label": "electrode",
"figurePage": 0,
"bounds": {
"left": 899,
"top": 866,
"right": 964,
"bottom": 902
}
}
]
}
],
"classifications": [
{
"code": "H10K85/50",
"description": "Organic perovskites; Hybrid organic-inorganic perovskites [HOIP], e.g. CH3NH3PbI3",
"isFirst": true,
"hierarchy": [
{
"code": "H",
"description": "ELECTRICITY"
},
{
"code": "H10",
"description": "SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR"
},
{
"code": "H10K",
"description": "ORGANIC ELECTRIC SOLID-STATE DEVICES"
},
{
"code": "H10K85/00",
"description": "Organic materials used in the body or electrodes of devices covered by this subclass"
},
{
"code": "H10K85/50",
"description": "Organic perovskites; Hybrid organic-inorganic perovskites [HOIP], e.g. CH3NH3PbI3"
}
],
"isCpc": true
}
],
"cpc": [
"H10K85/50",
"H01L51/0032",
"H01L31/0256",
"H01L51/0045",
"H01L51/4226"
],
"ipc": [],
"definitionsCount": 95,
"landscapes": [
{
"name": "Chemical & Material Sciences",
"type": "AREA"
},
{
"name": "Engineering & Computer Science",
"type": "AREA"
}
],
"concepts": [
{
"id": "125000000962",
"name": "organic group",
"domain": "Chemical group",
"similarity": 0,
"sections": [
"claims",
"description"
],
"count": 13
}
],
"abstract": "A solar cell includes a light-absorbing layer comprising a 2d-perovskite.",
"abstractOriginal": null,
"translatedFrom": null,
"claims": [
{
"number": 1,
"text": "1. A solar cell including:\na first electrode, wherein the first electrode is optically transparent for at least some wavelengths of light;\na second electrode; and\na light-absorbing layer disposed between the first electrode and the second electrode, wherein the light-absorbing layer comprises a mixture of different 2d-perovskites each having inorganic layers and each having the form:\n(A)2(B)n−1[MnX3n+1]\nwhere:\nA=RNH3 +, R=organic group\nB═R′NH3 +, R′=organic group\nM=metal\nX=anion\nn denotes the number of M-X sheets in each inorganic layer, and\nwherein for each of the 2d-perovskites in the mixture, n is at least 2, and further\nwherein the different 2d-perovskites vary in n.",
"isIndependent": true
},
{
"number": 2,
"text": "2. The solar cell of claim 1 wherein for at least one of the 2d-perovskites in the mixture, R═C6H5(CH2)2.",
"dependsOn": [
1
],
"isIndependent": false
}
],
"claimsCount": 17,
"independentClaimsCount": 1,
"priorityApplications": [
{
"applicationNumber": "US14/733,462",
"publicationNumber": "US9564593B2",
"language": "en",
"priorityDate": "2014-06-06",
"filingDate": "2015-06-08",
"title": "Solar cells comprising 2d-perovskites"
}
],
"applicationsClaimingPriority": [
{
"applicationNumber": "US201462008704P",
"priorityDate": "2014-06-06",
"filingDate": "2014-06-06"
}
],
"childApplications": [],
"parentApplications": [],
"publications": [
{
"publicationNumber": "US20150357591A1",
"language": "en",
"publicationDate": "2015-12-10"
},
{
"publicationNumber": "US9564593B2",
"language": "en",
"publicationDate": "2017-02-07",
"isThisPatent": true
}
],
"familyId": "54770299",
"familyApplications": [
{
"applicationNumber": "US14/733,462",
"publicationNumber": "US9564593B2",
"language": "en",
"status": "Active",
"priorityDate": "2014-06-06",
"filingDate": "2015-06-08",
"title": "Solar cells comprising 2d-perovskites"
}
],
"familyApplicationsBefore": [],
"familyApplicationsAfter": [],
"familyApplicationsByYear": [
{
"year": 2015,
"applications": [
{
"filingDate": "2015-06-08",
"countryCode": "US",
"applicationNumber": "US14/733,462",
"publicationNumber": "US9564593B2",
"legalStatus": "Active",
"legalStatusCategory": "active",
"isThisApplication": true
}
]
}
],
"familyCountryStatus": [
{
"countryCode": "US",
"count": 1,
"publicationNumber": "US9564593B2",
"language": "en",
"isThisCountry": true
}
],
"familyMembers": [
{
"publicationNumber": "US20150357591A1",
"countryCode": "US",
"publicationDate": "2015-12-10",
"language": "en"
}
],
"familyMembersCount": 1,
"citations": [
{
"publicationNumber": "WO2014045021A1",
"language": "en",
"examinerCited": true,
"priorityDate": "2012-09-18",
"publicationDate": "2014-03-27",
"assignee": "Isis Innovation Limited",
"title": "Optoelectronic device"
}
],
"citationsOriginal": [
{
"publicationNumber": "WO2014045021A1",
"language": "en",
"examinerCited": true,
"priorityDate": "2012-09-18",
"publicationDate": "2014-03-27",
"assignee": "Isis Innovation Limited",
"title": "Optoelectronic device"
}
],
"familyCitations": [],
"citedBy": [
{
"publicationNumber": "US10535791B2",
"language": "en",
"priorityDate": "2014-12-03",
"publicationDate": "2020-01-14",
"assignee": "The Board Of Trustees Of The Leland Stanford Junior University",
"title": "2-terminal metal halide semiconductor/C-silicon multijunction solar cell with tunnel junction"
}
],
"citedByOriginal": [
{
"publicationNumber": "WO2018219155A1",
"language": "en",
"examinerCited": true,
"priorityDate": "2017-06-02",
"publicationDate": "2018-12-06",
"assignee": "颜步一",
"title": "Perovskite solar cell module and manufacturing method thereof"
}
],
"familyCitedBy": [
{
"publicationNumber": "GB201513366D0",
"language": "en",
"examinerCited": true,
"priorityDate": "2015-07-29",
"publicationDate": "2015-09-09",
"assignee": "Univ Ulster",
"title": "Photovoltaic device"
}
],
"nonPatentCitations": [
{
"text": "Lee et al. \"Efficient hybrid solar cells based on meso-superstructured organometal halide perovskites\". Science 338, 643-647 (2012).",
"examinerCited": true,
"url": "http://scholar.google.com/scholar?q=\"Efficient+hybrid+solar+cells+based+on+meso-superstructured+organometal+halide+perovskites\""
}
],
"citationsCount": 1,
"citationsOriginalCount": 1,
"familyCitationsCount": 0,
"citedByCount": 6,
"citedByOriginalCount": 5,
"familyCitedByCount": 16,
"nonPatentCitationsCount": 3,
"citationsTotal": 1,
"citedByTotal": 21,
"similarDocuments": [
{
"publicationNumber": "US20150357591A1",
"language": "en",
"publicationDate": "2015-12-10",
"title": "Solar Cells Comprising 2d-Perovskites",
"isPatent": true
}
],
"legalEvents": [
{
"date": "2015-10-06",
"code": "AS",
"title": "Assignment",
"attributes": [
{
"label": "Owner name",
"value": "THE BOARD OF TRUSTEES OF THE LELAND STANFORD JUNIO"
},
{
"label": "Free format text",
"value": "ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:KARUNADASA, HEMAMALA INDIVARI;SMITH, IAN;MCGEHEE, MICHAEL DAVID;SIGNING DATES FROM 20150928 TO 20151001;REEL/FRAME:036735/0539"
}
]
},
{
"date": "2017-01-18",
"code": "STCF",
"title": "Information on status: patent grant",
"attributes": [
{
"label": "Free format text",
"value": "PATENTED CASE"
}
]
}
],
"snippet": "A solar cell includes a light-absorbing layer comprising a 2d-perovskite.",
"matchType": "keyword",
"searchRank": 809,
"search": "perovskite tandem solar cell",
"language": "en",
"thumbnailUrl": "https://patentimages.storage.googleapis.com/61/1c/10/a7bee0846e42e9/US09564593-20170207-D00000.png",
"searchFigures": [
{
"thumbnailUrl": "https://patentimages.storage.googleapis.com/4b/8e/fa/ad5975ac57f73f/US09564593-20170207-D00000.png",
"fullUrl": "https://patentimages.storage.googleapis.com/72/d0/67/ddde47f0f1c268/US09564593-20170207-D00000.png"
}
],
"familyStatusByOffice": [
{
"countryCode": "US",
"state": "ACTIVE",
"stateLabel": "Active"
}
],
"detailLevel": "full",
"mode": "search",
"url": "https://patents.google.com/patent/US9564593B2/en",
"status": "GRANT",
"type": "PATENT",
"assignee": "Leland Stanford Junior University",
"summary": "US9564593B2: Solar cells comprising 2d-perovskites. Granted 2017-02-07. Assignee: Leland Stanford Junior University. Inventors: Hemamala Indivari Karunadasa, Ian Smith, Michael David McGehee. Priority date 2014-06-06. Legal status: Active. 17 claims, 1 patent citation, cited by 21.",
"scrapedAt": "2026-10-08T05:58:12Z"
}

Real row from a full-page search run (US9564593B2). Long lists are shortened to their first entries: the row holds 17 claims, 6 citedBy, 5 citedByOriginal, 16 familyCitedBy, 3 nonPatentCitations, 8 events, 5 legalEvents, 3 figures, 18 classifications, 23 similarDocuments, 6 externalLinks, 7 landscapes, 140 concepts, 2 applicationsClaimingPriority, 3 searchFigures, 18 cpc. The full description and the 95 term definitions are added when Include description and definitions is on. The platform stores rows of up to 9 MB; in the rare case a patent's row would exceed that (a very long description with the full text switched on), its longest fields are shortened and flagged, for example descriptionTruncated: true.

Pricing

You pay per patent returned, plus an add-on for each patent that carries its full page.

EventWhen it is chargedFree planBronzeSilverGold and above
PatentEach patent written to the dataset$1.99 / 1,000$1.79 / 1,000$1.69 / 1,000$1.39 / 1,000
Full patent pageEach patent with its full page (search results with Include full patent page on, and every patent in the number, family and citation modes)$2.99 / 1,000$2.69 / 1,000$2.49 / 1,000$2.09 / 1,000

On the Free plan, 100 search results with their full pages cost about $0.50 ($0.199 + $0.299); 1,000 results without the page cost about $1.99. Each run also has a start charge of $0.005 per GB of run memory ($0.02 at the default 4 GB) and a storage charge of $0.01 per 1,000 rows. The description and definitions add no charge. Set a spending limit on the run and the actor stops cleanly when it is reached, keeping everything already collected.

FAQ

Why do I get more results than Google Patents shows? Google Patents stops listing at 1,000 results per search; the patents beyond that are still in its index. The actor collects them too, and the run summary shows the total.

Why does the result count differ from "About N results" on Google Patents? That figure is an estimate. With family grouping it also counts publications that collapse into one family. When the run summary says totalIsExact: true, its totalOnGooglePatents is the exact number of rows the search yields in the grouping you chose.

What is a "similar" result? A document Google Patents shows for your search because it is close in meaning, even if it does not contain all your terms. Keep them for discovery, or switch Include similar documents off.

A search returned nothing. What now? Check the same search on Google Patents itself. If the run log says Google Patents could not read the search, a bracket or quote is unbalanced or an operator is misspelled. Narrow filters (office, dates, status, type) combine, so loosen one at a time.

Which offices are covered? Every patent office Google Patents covers, over 100, including USPTO, EPO, WIPO (PCT), China, Japan, Korea and Germany. Non-English patents come with Google's English translation and the original text.

Can I get claims, citations and family for a list of patent numbers? Yes: choose Patent numbers and paste the numbers. Each row has the full patent page.

Why does the citation count differ from the "Cited by" number on Google Patents? The number Google Patents shows beside "Patent citations" and "Cited by" adds the patents listed on the page to the family-to-family citations. That number is citationsTotal and citedByTotal. The citations and citedBy lists also include every other publication of each cited or citing patent, for example both its application and its grant, so their counts are usually higher than citationsOriginalCount and citedByOriginalCount.

How long does a run take? In our tests, a search of 1,677 patents with the full patent page for each finished in about two and a half minutes, and a search of 10,114 patents without the full page in about a minute. Turning on the description and definitions makes full-page runs slower, since each row is much larger.

Why is the abstract empty for some patents? Some publications have none, for example European granted patents (B1) and many Japanese grants. The claims and description are still there.

Does Google offer this data directly? Google publishes the underlying publication data as public patent datasets in BigQuery. Using them means writing SQL in a Google Cloud project, and they do not include Google Patents' search, ranking or similar-document results. This actor gives you the search itself and each patent page as data, with no setup.

Do I need a Google account or any credentials? No. Nothing to configure and nothing to log into.

How should I use the data? Patent documents are public records published by patent offices. Google notes that legal status and assignee data are its assumptions, not legal conclusions; confirm critical facts with the office register before relying on them.

Support

Something wrong or missing? Open an issue on the Issues tab, or email max@mapa.slmail.me. Reports that include the run ID get answered fastest.