To get Google Patents search results through an API, one option is a third-party service such as SerpApi. To analyze public patent datasets with SQL, use BigQuery. Choose according to the required data source, search inputs, and whether the task calls for record retrieval or SQL analysis.
For applications that need patent retrieval, Patsnap Open offers text-query and semantic search APIs over its patent database, with additional patent-data operations for follow-up research.
Which patent data access method should you use?
Need to browse and read patents manually?
→ Use Google Patents or a patent research application such as Patsnap Eureka. The API routes below are for programmatic access.
Need to analyze a defined patent dataset with SQL?
→ Use public patent tables in BigQuery.
Need your application to retrieve patents from user queries?
→ Choose a patent search API according to the source and query type:
- Google Patents search results: consider a third-party service such as SerpApi.
- Search a patent database within your application: consider Patsnap Open, choosing standard text-query or semantic search according to the input.
Semantic search is a retrieval capability within a patent search API. Check the supported query methods and filters; keyword and semantic search can be complementary.
Retrieve Google Patents results with SerpApi
SerpApi’s Google Patents API is a third-party service for retrieving Google Patents search results. Its documented search uses engine=google_patents and q for the query. Choose this route when Google Patents is the required search source; use SerpApi’s documentation for authentication, response fields, and pagination.
For a first request, use a familiar patent topic and inspect the returned records before building your application’s screening or display logic.
Access public patent datasets with BigQuery
The Google patent datasets repository links to the patent dataset in Google Cloud and explains its use for SQL analysis. Its collection includes tables contributed by government, research, and private organizations. The repository is now archived, so use it to locate the project and understand the examples, then follow current Cloud documentation for implementation.
Google documents access to BigQuery public datasets through the Cloud console, command-line tools, REST API, and client libraries. A query runs in your Google Cloud project. Public access to a dataset does not remove query-processing costs or your organization’s access controls.
Get started with the BigQuery route
Use a small, inspectable query before designing a larger pipeline:
- Locate the dataset. Follow the dataset link in the Google repository and inspect the available tables in the Cloud console.
- Select a project. Confirm that your account can run queries and access the dataset. Review billing requirements for the workload you intend to run.
- Inspect the schema. Identify the fields needed for your research question and how repeated values are represented.
- Query a bounded population. Select the required fields and restrict the scope. Review the query’s estimated processing before running it.
- Move the checked query into your application. Use the BigQuery REST API or an appropriate client library, retaining the query definition and dataset reference.
For example, an application intended to summarize filing activity needs a defined date basis and counting unit before it needs generated commentary. Decide whether the output represents publications, applications, or families, then make the query and labels consistent with that choice.
Check whether the data fits the application
Before committing to a data model, inspect a few records representative of the output you want to build.
| Requirement | Question to resolve |
|---|---|
| Document identity | Which identifier will let a reviewer return to the source patent? |
| Text depth | Does the selected table contain the text needed for your analysis? |
| Dates | Does the field mean filing, publication, priority, or another event? |
| Multiple values | How will you retain multiple applicants, classifications, or language versions when present? |
| Currency | When was the table modified, and what does the dataset say about updates? |
| Reuse | Do the dataset’s terms support your intended use and redistribution? |
Store the dataset reference, query version, and extraction time with each output so later changes can be investigated.
Search patent data with Patsnap Open APIs
When your application needs patent search, choose the query method around the user's input. A researcher may supply known terms and field constraints, or describe a technical problem whose terminology is still being explored. Patsnap Open supports both routes.
For conventional queries, the Patsnap text-query search API searches Patsnap's patent database using a query_text request. Its documented options include application- and family-based result grouping; returned fields include patent identifiers, titles, dates, and applicant or assignee information.
The Patsnap Semantic Search API accepts technical text and returns a patent list. Its documented request includes a required text field, with optional search controls; the response includes patent identifiers and bibliographic fields. Use this route when the input is a technical description rather than a prepared search expression.
For example, a user describing “a sealing mechanism that tolerates repeated thermal expansion” may start with semantic retrieval, then use terms found in relevant documents to develop a more explicit query. To try this with your application, choose one representative user question, submit it through the documented search operation, and inspect the returned candidates before adding generated analysis.
The broader Patsnap API catalog also includes patent basic and legal data operations. This makes Patsnap relevant to applications that need to move from finding patents to retrieving further records for review. Select the operations and fields required by that workflow.
Keep the responsibilities clear:
- Request: the text query or technical description, with controls supported by the selected operation.
- Source response: the returned patent candidates and fields.
- Application output: your screening, annotations, or AI-assisted explanation of relevance.
A similarity-ordered result list provides candidates for examination. Your application should connect each explanation to the document information that supports it.
Follow Patsnap’s first-request guide to configure authentication, then use the selected operation reference to send your query. Keep the API key in the application’s secure configuration and check the response against the fields your feature needs.
Make your first Patsnap API request
Use the authentication and request guide to try the patent-search operation you selected, then check the returned records against your application’s needs.
Frequently asked questions
Is the BigQuery API a Google Patents search API?
The BigQuery API operates on BigQuery data and query jobs. For public patent tables, it provides a programmatic route to SQL results. Treat it as a data-query interface when designing your application.
Is SerpApi’s Google Patents API provided by Google?
It is provided by SerpApi, a third-party service. The product name describes the search source it accesses; authentication and service conditions come from SerpApi.
Can patent records be used in an AI research application?
Yes, subject to the source’s terms and the records available. A practical design passes selected records to the analysis step, retains their identifiers, and distinguishes source text from generated interpretation.
What should the first prototype demonstrate?
That the chosen route returns the fields your application needs, that a reviewer can trace an output to its records, and that the query scope matches the intended task. Test those points before expanding the dataset or automating interpretation.
Sources and disclosure
Published by Patsnap. Access routes and product descriptions are based on the linked documentation, reviewed September 11, 2026. The application scenarios are illustrative; the APIs and BigQuery workflow described here were not executed for this article.