This guide covers what has to be in the record, who checks it and in what order, and how to decide how much independent review each search needs.
Key takeaways
- A reviewable search lets a colleague reconstruct its strategy and inspect its limits. Missing records restrict what can be checked and should be identified explicitly.
- Where practical, the reviewer forms an initial view before opening the output. This provides an independent comparison point; a report-first review can still challenge the work.
- Not every search needs the same level of independent review. Define the checks and escalation criteria before the output is used to support a decision.
Make the review checkpoint explicit
Two kinds of information need to survive the move from search to report.
Why this strategy was used. Keep the subject definition, the important alternatives considered, and the reasons for changing or retaining a route. Cheaply generating several variants does not by itself record why one was chosen.
Where the work remains limited. Document missing sources, unverified dates, uncertain comparisons, and the follow-up needed. Do not assume that either the person who ran the search or the generated report will preserve these details without an explicit process.
Assign a review checkpoint before the result is used for a consequential decision.
Part one: what the record has to hold
Make the strategy reconstructable
"A comprehensive worldwide search was conducted" is not enough to reconstruct a strategy. Record what was actually searched and the limits that affect the intended decision.
Record the purpose and input version, search date, relevant date limits, databases and coverage, search terms and their relationships, classifications, filters, and material changes between iterations. State whether non-patent literature was searched and where. For AI-assisted work, also retain the visible prompts or instructions, material user edits, run identifiers, and dated outputs available in the workflow. Note any settings or history that cannot be exported.
The standard is not that a rerun must return an identical list. Indexes, rankings, and collections can change. A colleague should be able to reconstruct the recorded strategy, compare a rerun with a dated snapshot, and investigate material differences.
Keep the reasoning behind the selections
Where the report selects a reference as a starting point for an inventive-step assessment, explain that choice. Give reasons tied to the claimed subject matter and the assessment framework, rather than treating the highest-ranked or most similar result as automatically decisive.
Record material alternatives and why they were not used. Not every search needs one designated closest reference, and a different starting point is not automatically an error. Keep the selection reasons visible whether they came from a person or a tool.
Distinguish not found from not there
"Not explicitly disclosed" does not rule out implicit disclosure. "Not identified in the material reviewed" records a limited observation. A finding that a feature is not disclosed requires a reasoned reading of the relevant reference, including what a skilled reader would derive directly and unambiguously.
Use wording that states what was checked. Do not turn an unsuccessful text search, an absent snippet, or a missing citation into a finding that the full reference lacks the feature.
Write the uncertainty down
Record limitations that could change the decision, such as an unresolved implicit-disclosure question, an unsearched source, or a date that could not be verified.
Where an issue has been resolved, keep the evidence and reasoning. The absence of an uncertainty section does not prove that a report hid a problem; the question is whether the record accurately shows the work and its remaining limits.
Give the reviewer more than a summary. Start with an invention description, then use the search strategy and feature comparisons as material for a colleague's review.
Part two: who checks it, and in what order
A useful starting sequence
Where practical, read the invention disclosure and proposed claims before opening the search output.
Form an initial view of the subject and sketch possible search routes: terms, classifications, and sources. Then compare that view with the output. For a report-first review, still return to the original input and independently test material choices and findings.
The purpose is to give the reviewer a comparison point that was not copied from the output. The sequence is a practical safeguard, not proof that the review is independent or correct.
What independence means in practice
For an independent review, use someone other than the person who ran the search, with enough technical and patent-search competence to challenge it. The inventor can clarify the technical input but should not be the sole independent reviewer of that interpretation. Complementary experience can help; a different background is not a substitute for relevant expertise.
Two people can still miss the same issue. The purpose is to test material choices and evidence, not to assume that the second reviewer must find something different.
Part three: how much independent review is needed
Independent review takes time. Use a tiered approach based on the consequences of an incorrect result, the uncertainty involved, and the resources available.
| Search type | Review | Why |
|---|---|---|
| Core inventions | Independent review, always | The conclusion shapes filing strategy and portfolio position |
| Important improvements | Sampled review | Balances cost against unchecked work accumulating |
| Core inventions with a genuinely borderline call | Independent review, and consider a separate external search | Internal experience alone is not resolving it |
| Routine improvements with limited decision impact | Basic operator checks; independent review or sampling as the risk warrants | Lower impact may justify lighter independent review, but unresolved issues still need escalation |
This table is an example to adapt, not a standard to adopt. The right tiers depend on your filing volume, your field, and what a wrong call costs you.
One line is worth holding whatever you choose: running no review at all on core inventions should be a decision someone made, not something that happened because the work got fast.
What the reviewer writes down
The output of a review is a record of the checks performed, material findings, and follow-up decisions. Include disagreements where they arise; agreement after substantive checks is also a valid outcome.
A different understanding of the subject may require the input to be clarified. A different search route may reveal a coverage gap or simply an alternative way to reach the same evidence. A different comparison finding calls for the underlying passage and reasoning to be checked.
Record what action each material finding requires. "Reviewed and approved" can be a sign-off, but it is not a substitute for the supporting review record.
The questions a reviewer should be able to answer
Before the reviewed search is used, the reviewer should be able to answer these questions from the record. Consulting the source documents is part of the review, not evidence that it is unfinished.
- How did you establish an independent understanding of the search subject?
- Which alternative routes did you consider, and did any reveal a material gap?
- Which findings did you challenge or verify, and what was the outcome?
- What in this record could you not verify?
- Which unresolved limitation could most affect the decision, and what follow-up is needed?
The last question connects the review to the next action. Record the answer even when the appropriate action is to accept a stated limitation rather than run another search.
How a reviewable search takes shape 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 can follow that work from the original input to the search routes, source documents, and comparison rationale. Where practical, form an initial view of the invention before reading the output, then record the checks performed and any follow-up needed. The workflow provides material for review; the team supplies the independent reviewer and records any context the output does not retain.
This article is general information about search and analysis method. It is not legal advice. Patentability and infringement questions depend on the claims, the applicable law, the relevant date, and the jurisdiction, and should be assessed with qualified professional advice.
