How to Run and Review a Novelty Search in Five Steps

Step 1: Define what must be searched

Start with the disclosure, drawings, and any conversation with the inventor. The first task is not to generate keywords. It is to agree on what the invention is.

Capture four things:

• the technical problem;
• the proposed technical solution;
• the features that make the solution work; and
• the technical effect attributed to those features.

Separate essential features from preferred embodiments and implementation detail when planning the search. When checking a proposed claim, retain every limitation and the relationships between them, including any feature you initially treated as optional. Translate broad benefit language into the structure or operation that produces the benefit. "Easier to use" alone does not identify what to search.

The question for the reviewer is simple: if the inventor reads the feature set, would they recognize the technical idea being searched? If not, stop here. More queries will only make the wrong search look more complete.

Step 2: Build several routes into the literature

Translate each core feature into more than one retrieval route. Depending on the subject, that may include:

• direct, broader, and narrower terms;
• functional and structural wording;
• relevant IPC or CPC classifications;
• citations, families, applicants, or inventors; and
• non-patent literature where the technology calls for it.

This is not a contest to build the longest Boolean string. Each route should serve a clear purpose and target a specific technical concept.

The USPTO's multi-step search guidance similarly moves from alternative description terms to document review, classifications, cited references, foreign patents, and non-patent literature. It also recommends recording the databases, date, terms, and documents retrieved.

The reviewer should be able to ask: which feature does this route test, and what kind of miss is it intended to prevent?

Step 3: Search, learn, and loop back

Keep the contribution of each route identifiable as you run, combine, or narrow the searches. Review useful documents for terminology, classifications, citations, and adjacent technical fields that were not visible in the original disclosure.

Good searching changes the strategy. A newly discovered term may send you back to Step 2. A document that reveals a misunderstood mechanism may send you back to Step 1. Record those changes instead of preserving only the final query.

Do not use a hit count as a quality score. A large set can contain mostly noise; a small set can reflect an overly narrow query or limited source coverage. What matters is whether the routes are technically justified and whether the resulting documents can be examined.

Step 4: Compare features without collapsing the evidence

Compare each selected reference with the defined feature set. Keep the source passage and the reasoning visible. A useful comparison answers:

• Which feature is being compared?
• Where does the document disclose the relevant subject matter?
• Is the feature expressly stated or necessarily implied, and what supports that reading?
• Is any part of the feature missing from the identified disclosure, or does the comparison remain uncertain?

Keep each document separate during a novelty review. Under the EPO Guidelines on novelty, the question is whether a prior-art disclosure makes the claimed combination available. Partial matches across separate documents, or unrelated embodiments within one document, should not be assembled into a disclosure of the whole combination. Whether a feature is necessarily implied requires its own explanation; inventive step is a separate assessment.

The table is a review aid, not a patentability opinion. Claims, relevant dates, applicable law, and the content of the source documents still require professional judgment.

Step 5: Report the path, not only the result

A report should let another person reconstruct the analysis. At minimum, it should show:

• what technical subject and feature set were searched;
• which routes, sources, classifications, filters, and exclusions were used, distinguishing the search date from the relevant date assumptions;
• which documents were selected and why;
• how each important feature was compared with the source evidence; and
• what the search did not cover or could not resolve.

The conclusion should be proportionate to that record. A reviewer needs to see uncertainty and remaining work, not only a favorable or unfavorable label.

How the five checkpoints appear in Eureka

The Eureka Novelty Search workflow starts from a natural-language invention description. The description becomes a technical summary and an extracted feature set. Those features feed search elements and multiple routes. Selected references can then be compared feature by feature and included in a report with supporting documents.

Reviewers should still inspect the extracted features, search routes, source documents, and comparison rationale. A product status such as Disclosed is not, by itself, a legal conclusion about patentability.