Three core principles for a patent search strategy

Principle 1: begin broadly, then control the noise

A patentability search looks for documents that could materially change an assessment. Therefore, an early query should usually favor reasonable breadth over an artificially clean result set. After reviewing representative results, the searcher can narrow the scope with better terminology, classifications, dates, fields, and exclusions.

Principle 2: cross-check through multiple paths

Do not rely on one query type. Keyword searching can catch language that classifications miss, while classification searching can retrieve documents that use unfamiliar wording. Citation and assignee searches create further paths into the same technical area. Cross-checking reduces omission risk, although it cannot eliminate it.

Principle 3: keep the process reviewable

Record the source, search date, query, filters, result count, and decisions made at each stage. This record helps colleagues review or repeat the work and explains why the search changed over time. It is a work product for quality control; its legal status and use in a proceeding depend on the applicable rules and professional advice.

Method 1: build the keyword strategy

Keyword searching is fundamental, but the same technical idea may appear under different terms. Start from the invention disclosure, then expand each core concept through technical, functional, commercial, and problem-oriented language.

Six useful sources of keywords

Build a keyword matrix

Create one row for each core technical concept. For a heat-dissipation concept, the matrix might contain the exact term, synonyms such as thermal management, broader terms such as heat transfer, narrower terms such as liquid cooling or heat pipe, and functional terms such as temperature control. Add language-specific equivalents only after a fluent reviewer or reliable domain source confirms them.

Before building that matrix, Eureka IP Drafting’s Invention Disclosure workflow can turn rough engineering notes into a structured disclosure, flag missing details, and highlight possible inventive points.² That structure can help a searcher identify concepts to test, but it does not replace inventor validation or professional review.

Use truncation, wildcards, and Boolean logic carefully

Truncation can capture word variants. For example, cool* may return cool, cooling, cooled, cooler, and coolant. A single-character wildcard may capture spelling variants. However, symbols and behavior differ by source, so check the relevant help page before running or documenting the query.

("heat dissipation" OR "thermal management" OR cooling)
AND (graphene OR "carbon material")
NOT (PCB OR "printed circuit")

Use OR between alternatives for the same concept, AND between distinct concepts, and exclusions only after confirming that they do not remove useful results. Field codes and the available form of NOT also vary by platform.

Method 2: add classifications to the patent search strategy

A keyword search can only use language the searcher anticipates. Classification searching offers another route when a document describes the same technology differently.

IPC, CPC, and Japan’s FI/F-term systems

• IPC: WIPO describes the International Patent Classification as a hierarchical, language-independent system of symbols for classifying patents and utility models by technical field.³
• CPC: The Cooperative Patent Classification extends the IPC and is jointly managed by the EPO and USPTO. Its hierarchy includes sections, classes, subclasses, groups, and subgroups.⁴
• FI and F-terms: The Japan Patent Office describes FI and F-terms as Japanese patent classification systems and provides official lists and a parallel IPC-FI-CPC viewer.⁵

The right system depends on the source, jurisdiction, technology, and search objective. Classification scope and assignments can change, so confirm the current definition rather than relying on an old label from a single document.

How to find useful classifications

1. Run a preliminary keyword search and identify representative documents.
2. Inspect the classifications assigned to those documents.
3. Open the official scheme and read the title, notes, hierarchy, and neighboring groups.
4. Test both the broader group and relevant subgroups against known documents.
5. Record why each code was included or rejected.

For example, a searcher investigating cooling arrangements for semiconductor devices might examine the H01L 23 hierarchy and then verify the current scope of a relevant subgroup in the official scheme. The exact code should follow the current classification definition and the technical facts, not the example alone.

Combine classifications with keywords

classification concept: semiconductor-device cooling
AND keyword concept: graphene OR "carbon nanotube"

This combination uses classification concepts to define a technical neighborhood and keywords to focus on a particular material, structure, function, or effect. Run several variations instead of treating one combined query as final.

Method 3: follow forward and backward citations

Citation searching starts from a relevant seed document and follows its recorded relationships. A backward-citation review examines earlier documents cited by the seed. A forward-citation review examines later documents that cite it.

What citation paths can reveal

• earlier documents that influenced the technical lineage;
• later documents that build on, distinguish, or discuss the seed;
• alternative terminology, classifications, inventors, applicants, and assignees; and
• new seed documents for another search round.

Google Patents provides patent full text, metadata, citations, similar documents, and links to available PDFs, while its search help explains keyword, phrase, metadata, and CPC searching.⁶ Espacenet likewise supports worldwide patent searching and document-level navigation.⁷ Citation lists can differ by source and may include applicant, examiner, or other references, so inspect the underlying document and context before drawing a conclusion.

Method 4: search applicants and assignees

An entity-focused search can reveal relevant portfolios that broad technical queries do not surface quickly. Build the entity list from evidence in the result set rather than relying only on market reputation.

Identify useful entity names

• applicants or assignees of strong seed documents;
• direct competitors known to develop comparable products;
• entities that appear repeatedly in a relevant classification or result set; and
• rightsholders whose portfolios may affect the specific technical area.

Normalize legal names, former names, subsidiaries, transliterations, and common spelling variants. Then combine the applicant or assignee field with a verified classification or technical concept. Field syntax differs across sources, so treat a pattern such as assignee:(name variants) AND classification:(verified code) as a conceptual template.

Coordinate multiple patent and literature sources

Coverage, update timing, search syntax, family handling, translations, and citation presentation differ across sources. Choose sources according to the relevant jurisdictions and evidence needs, then record the limits of each one.

WIPO states that PATENTSCOPE searches published PCT applications, participating national and regional patent collections, and non-patent literature, with criteria including keywords and IPC.⁸ The USPTO describes Patent Public Search as its web-based patent search application for patents and published patent applications.⁹ These examples are not a ranking or an exhaustive tool list.

Refine the search in documented rounds

Round 1: broad preliminary search

Combine the main concept groups with their most credible alternatives. Review representative results, terminology, languages, classifications, dates, and applicants. Do not judge the search only by the total result count.

Round 2: targeted deepening

Test promising classifications with separate keyword groups. Add terminology found in relevant documents, and search important applicants or assignees with name variants.

Round 3: focused gap checking

Use feature comparison to identify unresolved elements. Then run targeted queries, follow citations, and test alternative functions, structures, materials, and problem statements.

Round 4: non-patent literature review

Search the types of technical literature that matter for the field, such as journal articles, conference papers, standards, manuals, theses, regulatory materials, or product documentation. Confirm publication dates, public availability, and relevance to the applicable legal question.

Five common patent search strategy errors

1. Using keywords without classifications

Risk: unfamiliar terminology may hide relevant documents. Response: inspect classifications from representative results and test their official scope.

2. Missing synonyms and concept levels

Risk: the query reflects only the inventor’s preferred wording. Response: build a matrix of exact, synonymous, broader, narrower, functional, and problem-oriented terms.

3. Applying an unsuitable date range

Risk: arbitrary limits may exclude older foundational documents or recently published applications. Response: set dates from the relevant legal and technical context, then document the rationale.

4. Depending on one source

Risk: source-specific coverage and presentation gaps become search gaps. Response: cross-check sources appropriate to the target jurisdictions and technical field.

5. Ignoring non-English terminology

Risk: the search may miss documents whose title, abstract, claims, or full text use another language or translation. Response: add verified multilingual terms and use classifications, families, machine translations, and local-language review where appropriate.