How to Create a Patent Landscape with AI
A patent landscape helps you understand how a technology field is developing: which technical approaches appear most often, which organizations are active, and which areas deserve closer review. This guide shows how an AI Skill turns a defined research question and patent evidence into that structured view—without treating a list of search results as a finished landscape.
This real partial run uses sulfide-electrolyte all-solid-state batteries as a public example. It produced a provisional working query, comparable company counts, overlapping route signals, and representative-family checks. Precision approval, human taxonomy review, and full-pool tagging were not completed, so the Sample is a landscape basis rather than the final report.
Sulfide-electrolyte all-solid-state batteries
The provisional working query returned 343 DOCDB-collapsed groups and four overlapping route signals. Toyota has the largest independently queried company subset, but neither volume nor a query hit proves technical leadership.
| Landscape view | Current finding | What to review next |
|---|---|---|
| Working population | 343 DOCDB-collapsed groups under the provisional TAC query | Approve query precision and freeze the denominator before using population-level statistics |
| Technical routes | Manufacturing 184; interfaces 128; composition 113; cell architecture 72 | Review overlaps and approve a human taxonomy before treating these as field structure |
| Company activity | Toyota 108; LGES 27; Hyundai/Kia 21; GM 10; Samsung SDI 5 | Verify entity coverage and representative families under the same query |
| Representative patents | Toyota US11201332B2; GM US11217826B2; Samsung US12341154B2; CATL US11699812B2 | Check claims and disclosures before assigning route or strategy labels |
Define the field and decision, then create a traceable landscape basis.
What a patent landscape should explain
A landscape is more than a patent count. It defines a technical boundary, normalizes families and applicants, separates complete metrics from representative records, and connects observed patterns to decisions.
A useful report normally combines an activity view, leading organizations, jurisdictions, legal-status context, a reviewed technology taxonomy, representative families, and explicit implications for R&D or portfolio work. Every view needs a declared date basis, counting rule, and evidence cutoff.
The difficult part is preserving the chain from query and screening decisions to taxonomy labels, patent evidence, interpretation, and the next action. Until human tagging is returned, provisional branches should remain a research map rather than a final description of the field.
How these patent intelligence Skills work together
These Skills form a connected research path, but they are not a mandatory sequence. Start with the task you need: build the field view, examine change over time, compare companies, monitor a selected competitor, or combine patent evidence with a broader project assessment.
How the Skill builds a defensible patent landscape
The Skill moves through four linked stages. Each stage produces an artifact that constrains the next, so a broad search cannot silently become a finished landscape.
1. Define the decision and scope
It records the decision the landscape must support, the technical boundary, inclusions and exclusions, jurisdictions, dates, family rule, and audience. This prevents the research question from changing after results appear.
Output: an approved scope and counting contract.
2. Build and test the search
It creates versioned search branches, checks precision and near misses, and documents de-noising decisions. Search Patents operates here as an internal retrieval stage; it does not create the final landscape conclusion.
Output: a reproducible query and candidate population.
3. Structure the evidence
It normalizes applicants and families, proposes a technology taxonomy, packages records for human review, and keeps provisional query signals separate from approved tags.
Output: comparable statistics, a reviewable taxonomy, and tagged evidence packages.
4. Explain the field
It connects activity patterns, organizations, routes, representative patents, counterevidence, and limitations to the original decision. Unsupported white-space or leadership claims remain out of the report.
Output: an evidence-backed landscape with bounded implications and next actions.
Patent analysis backed by connected records
A landscape depends on current patent records linked across families, normalized applicants, classifications, claims, legal status, and citations. Patsnap provides those connected intelligence layers; the Skill preserves the method that turns them into an auditable analysis rather than a search-results summary.
Database foundationPatsnap links global patent bibliography, families, applicants, classifications, claims, legal status, citations, scientific literature, and enterprise signals. That connected database foundation makes the evidence easier to retrieve, reconcile, and audit.
Patsnap OpenPatsnap Open provides integration paths through APIs and MCP Servers so selected patent data and research tools can be used inside AI and enterprise workflows. The records support analysis; qualified reviewers still own legal interpretation and current-register checks.
Prepare, install, and run
What you need to provide
- The decision the landscape must support
- Technology scope, inclusions, and exclusions
- Jurisdictions, time period, and intended audience
What the Skill keeps consistent
- Patent-family counting rule
- Applicant and subsidiary treatment
- Evidence cutoff and review scope
Once the scope is defined, use a prompt like the one below. The Skill will build and test the search, organize the evidence, and stop at any required human review checkpoint.
Create a patent landscape for sulfide-electrolyte all-solid-state batteries used in EV traction cells. Use DOCDB family collapse, disclose the evidence cutoff, validate query precision and near misses, compare selected applicants under the same query, and stop truthfully at the human tagging checkpoint if reviewed tags are not returned.Before using the result: Approve query precision and the technology taxonomy, then complete the required record tagging and representative-family review before treating the result as a completed landscape.
How to use the partial landscape above
The Sample gives you a reproducible working population, comparable company subsets, provisional route signals, and representative patents. Use it to approve or revise the scope and decide where human taxonomy and family review should begin.
Do not conclude yet: route shares, white space, exhaustive company rankings, or legal position. Those require approved query precision, a reviewed taxonomy, full-pool tagging, and deeper claim and status analysis.
Continue from the landscape to deeper evidence
Explore connected patent research tools when a route, company, family, or claim needs deeper verification.
Frequently asked questions
Is a patent landscape just a search-results summary?
No. It also requires defined counting, entity normalization, taxonomy, evidence review, and bounded interpretation.
Why can route counts exceed the population?
The route searches are overlapping labels; one family may contribute to several technical routes.
Can sparse patent activity prove white space?
No. It is a hypothesis requiring broader search, engineering need, feasibility, claims, and market validation.
Disclosure: This article describes a Patsnap product and links to Patsnap Open. Skill output supports research and does not replace qualified technical, commercial, scientific, regulatory, or legal review.
