Patent Search Fundamentals: A Complete Methodology for Prior Art Research

Patent searching is a technical discipline that looks simple from the outside but contains considerable depth. Many beginners assume that "entering keywords into a search box" constitutes a search — in reality, that is only the starting point. The factors that truly determine search quality are the logic behind keyword selection, the ability to use classification codes effectively, a thorough understanding of database characteristics, and the accumulated experience of filtering and analyzing results.

This guide covers patent search from methodology through practice, systematically introducing the complete knowledge base. Whether you are an inventor conducting a novelty search, a corporate IP team performing competitor analysis, or an attorney writing a search report, you will find practical guidance here.

Types of Patent Searches

Before starting a search, you must clarify the purpose, because different objectives require different strategies and different levels of thoroughness.

Search Type	Purpose	Search Depth	Typical Context
Novelty search	Determine whether prior art exists for an invention	High — must be as comprehensive as possible	Pre-filing assessment
Inventive step search	Identify the closest prior art	High	Supporting patent drafting
Freedom-to-operate (FTO)	Confirm freedom to implement a technology	Very high — must cover all active patents	Pre-product launch
Competitor monitoring	Track competitor technology developments	Medium — continuously updated	Strategic planning
Technology trend analysis	Understand development paths in a field	Medium — focus on statistical patterns	R&D project initiation
Invalidity search	Find prior art to invalidate a target patent	Very high	Patent litigation

Key principle: A novelty search aims to be "as comprehensive as possible," while an FTO analysis aims to "miss no active patents" — the priorities differ, and strategies should differ accordingly.

Keyword Strategy: From Single Terms to Search Queries

Keyword searching is the most intuitive starting point, but it is also where mistakes are most commonly made.

The Four-Dimensional Keyword Expansion Method

A complete keyword strategy requires expansion across four dimensions:

1. Synonyms and near-synonyms The same technical feature may be described in multiple ways. For example, "display screen" in patent text may appear as "LCD," "liquid crystal display," "display panel," "display device," "screen," or "monitor." Missing any one expression may cause relevant patents to be overlooked.

2. Broader and narrower concepts "Solid-state battery" is a narrower concept; "energy storage device" is broader. "Deep learning" is narrower; "machine learning" is broader. Searches should cover a reasonable range of hierarchical levels.

3. Functional descriptions When a technical feature cannot be accurately described by noun terms, functional language serves as an important supplement. Phrases like "a sealing structure for preventing liquid leakage" appear frequently in early patents.

4. Cross-language expansion Patent data is inherently multilingual. CNIPA databases are primarily in Chinese, USPTO and EPO in English, and JPO in Japanese. Both Chinese and English keywords must be considered for the target technical domain.

Correct Use of Boolean Operators

Operator	Meaning	Use Case	Example
AND	Both terms present	Narrow search scope	solid-state battery AND solid electrolyte
OR	Either term present	Expand synonyms	solid-state battery OR all-solid-state battery
NOT	Exclude term	Filter irrelevant results	battery AND lithium NOT primary lithium battery
Quotation marks " "	Exact phrase	Fixed collocations	"deep neural network"
Truncation *	Word stem expansion	Cover word form variations (mainly English)	encrypt* → encrypt/encrypted/encryption
Proximity operator	Positional relationship	Limit distance between words	W/3 (within 3 words)

From Keywords to Search Queries

A complete search query combines keywords and operators into a logical expression. Using "facial recognition access control system" as an example:

(facial recognition OR face recognition OR face detection) AND (access control OR door control OR entry system) AND (system OR apparatus OR device)

IPC and CPC Classification Codes: Precision Instruments for Search

Pure keyword searching has a fundamental limitation: patents covering the same technical solution may use entirely different language. Classification code searching breaks through language barriers by directly locating technical subjects.

IPC Classification Structure

The International Patent Classification (IPC) uses a hierarchical tree structure:

H (Electricity section)
└─ H01 (Basic Electric Elements)
   └─ H01M (Processes or means for direct conversion of chemical energy into electrical energy)
      └─ H01M 10/00 (Secondary cells; Manufacture thereof)
         └─ H01M 10/05 (Accumulators with non-aqueous electrolyte)

Core IPC classifications for common technical domains:

Technical Domain	Core IPC Codes
AI/Machine learning	G06N 3/00, G06N 20/00
Computer vision	G06V 10/00, G06V 40/00
Natural language processing	G06F 40/00
Semiconductor integrated circuits	H01L 21/00, H01L 27/00
Lithium-ion batteries	H01M 10/05, H01M 4/00
Photovoltaic power generation	H02S 10/00, H01L 31/00
Wireless communications	H04W 72/00, H04L 27/00
Medical diagnostics	A61B 5/00, A61B 6/00

How to Find the Right Classification Code

Method 1: Search for a known relevant patent in CNIPA or Espacenet, check its IPC classification code, then use that code as the basis for expanded searching.

Method 2: Use the WIPO IPC classification table (ipcpub.wipo.int) to search for corresponding classification codes by keyword.

Method 3: Use EPO's CPC classification browser (worldwide.espacenet.com/classification). CPC is more granular than IPC, suitable for precisely locating technical sub-domains.

Relationship Between CPC and IPC

CPC (Cooperative Patent Classification) was jointly developed by EPO and USPTO as a more detailed extension of IPC, with approximately 260,000 classification groups (IPC has about 70,000). CPC is only used in EPO (Espacenet) and USPTO data, while CNIPA uses its own classification system (IPC-compatible with domestic extensions).

Major Patent Databases: Features and Search Tips

The world's major patent databases each have distinct strengths. Understanding their advantages and limitations enables selection of the right platform for each search objective.

WIPO Patentscope

Data scale: Over 83 million patent documents from multiple patent offices — the preferred platform for searching PCT applications.

Core features:

• Interface available in 9 languages
• CLIR (Cross-Lingual Information Retrieval): search multilingual patent databases in any input language
• Command-line search supports complete Boolean logic and field restrictions
• PCT applications viewable in full text for free

Suitable for: PCT international application searches, cross-language searches, patent data from developing countries.

Search tip: Patentscope advanced search supports field restrictions such as TI: (title), AB: (abstract), and CL: (claims). Using these in combination greatly improves precision. Example: TI:solid electrolyte AND CL:lithium.

EPO Espacenet

Data scale: Over 110 million patent documents from 97 countries — one of the broadest-coverage free patent databases available.

Core features:

• Full-text search (including machine-translated full text)
• CPC classification browsing and searching
• Patent family lookup (finding related applications for the same invention across countries)
• Legal status lookup (INPADOC)
• Espacenet OPS API for developers

Suitable for: European patent searches, global patent family analysis, legal status tracking.

Search tip: "Smart search" accepts natural language queries and automatically searches titles and abstracts. For professional searches, switch to "Advanced search" for precise field and date range restrictions.

CNIPA Patent Search and Analysis System

Data scale: Full Chinese patent database; as of 2025, over 59 million records covering invention publications, invention grants, utility models, and design patents.

Access: pss-system.cnipa.gov.cn (registration required, free)

Core features:

• Three search modes: basic, advanced, command-line
• Full-text search (including specification full text)
• Real-time legal status updates
• Patent analysis tools (application trends, applicant analysis)
• Drawing downloads

Suitable for: Chinese patent prior art searches, domestic applicant competitive analysis, CNIPA legal status verification.

Search tip: Command-line syntax is more efficient than GUI searching. Basic syntax: TI=keyword AND AB=keyword AND IPC=H01M. For Chinese word segmentation issues, use double quotation marks for exact phrase matching.

Commercial Patent Database Comparison

Platform	Data Scale	AI Features	Price	Target Users
PatSnap	170M+	Semantic search, technology landscape	Enterprise custom	Large enterprises/law firms
Incopat	150M+	Chinese semantic search, patent maps	~RMB 30K/year	Domestic enterprises
Derwent Innovation	100M+	Standardized titles, enhanced abstracts	High	Multinational enterprises/law firms
Orbit Intelligence	150M+	Family analysis, value assessment	High	Large enterprises
Lens.org	300M+	Basic AI assistance	Free	Academic research
Google Patents	120M+	Similarity search	Free	Entry-level users

Patent Families and Legal Status Searching

Why Search Patent Families

An invention is typically patented in multiple countries, and these applications together form a "patent family." Understanding patent families is important for several reasons:

• FTO analysis: A Chinese patent may have been invalidated, but its US family member may still be active and represent an infringement risk
• Tracing inventors: Family members often contain more detailed technical descriptions (later national phase applications typically add more embodiments)
• Assessing scope of protection: Claims scope may differ across jurisdictions; understanding the family provides a complete view of protection boundaries

How to search: In Espacenet, after finding a target patent, click the "Patent family" tab to view all family members and their legal status.

Key Legal Status Categories

Legal Status	Meaning	FTO Impact
Pending	Application under examination, not yet granted	Low risk, but monitor
Active/Granted	Granted, within protection period	High risk — review claims scope
Lapsed/Abandoned	Lapsed due to non-payment of fees or voluntarily abandoned	No risk
Expired	Exceeded maximum protection period (20 years for inventions)	No risk — technology enters public domain
Invalidated	Declared invalid through invalidity proceedings	No risk

Standards for Writing Search Reports

After completing a search, the process and results are typically documented in a formal search report. The following is the standard structure:

Standard Search Report Structure

Section 1: Search Objective Clearly state the specific goal (novelty search / FTO analysis / competitor monitoring) and technical scope.

Section 2: Databases Searched List all databases used and the search dates; describe the coverage of each database.

Section 3: Search Strategy Document in detail the combination of keywords, classification codes, and Boolean logic. This section is critical for reproducibility.

Example search queries:
Databases: CNIPA + Espacenet
Query 1: TI=(face recognition OR facial recognition) AND IPC=G06V40/16 AND filing year: 2020–2024
Results: 234 documents

Query 2: AB=(face recognition OR facial recognition) AND CPC=G06V40/16
Results: 1,567 documents

Section 4: Results Summary List the number of hits for each query and the number of relevant documents after filtering.

Section 5: Closest Prior Art Analysis Provide detailed analysis of the most relevant documents, describing their similarities and differences from the target invention.

Section 6: Conclusion Based on the search results, state a clear conclusion (e.g., "No prior art documents have been found that would destroy novelty" or "X highly relevant prior art documents have been identified").

AI-Assisted Search: Emerging Trends

Between 2024 and 2025, AI-assisted patent search tools matured rapidly, primarily in three areas:

Semantic search: Overcomes the language limitations of keyword searching by finding patents that describe the same technical solution with different vocabulary, using semantic similarity. PatSnap, Incopat, and other platforms have integrated semantic search capabilities.

Cross-language search: Input Chinese and automatically match English, Japanese, and German patents; input English and automatically match Chinese patents. For global prior art searches, efficiency gains are significant.

Intelligent summarization and classification: AI automatically scores search results for relevance and extracts summaries, compressing the work of manually reviewing 1,000 patents down to focused reading of the most critical documents.

It is worth noting that recall and precision rates of AI search tools vary significantly across platforms and technical domains. For important searches, combining AI semantic search with traditional keyword + classification code searching — using both methods to cross-validate — is strongly recommended.

---

Action Checklist: Seven-Step Novelty Search Process

• 1. Clarify the search objective and technical subject; identify the core technical features of the claims to be protected
• 2. Develop a four-dimensional keyword list (synonyms, hierarchical concepts, functional descriptions, cross-language)
• 3. Identify 2–3 corresponding IPC/CPC classification codes
• 4. Complete domestic patent search in CNIPA (covering the last 10 years as a priority)
• 5. Complete international patent search in Espacenet or Patentscope
• 6. Filter hits by relevance; closely read the top 20–50 most relevant documents
• 7. Compile a search report documenting strategy, results, and conclusions