The Examiner Found Prior Art You Missed: How to Audit the Search

A later citation is an outcome, not a diagnosis

A search report looked favorable. During examination, a reference appears that the original search never surfaced.

The first reaction is often to blame the keywords. That may be correct, but it is only one possibility. The cited document may have been outside the original date or document scope. The claims may have changed. The database may not have indexed the relevant field or family member at the time. The document may have entered the result set and then been screened out.

The USPTO's guidance for applicants explicitly warns that a preliminary search may not be as complete as an examiner's search and that an examiner may rely on information not found in the preliminary work. That does not make private searching pointless. It means a later citation should trigger a disciplined audit rather than a confident one-line explanation.

Preserve the original search before rerunning it

Do not begin by editing the old query. First preserve the conditions that produced the original result:

• the disclosure and claim set available at the time;
• the search date and relevant date assumptions;
• the databases, searchable fields, jurisdictions, languages, and document types;
• the terms, classifications, citations, and other routes used;
• the filters and exclusions applied; and
• saved result lists and screening records showing which documents were retrieved, reviewed, or removed.

Also record the later-cited document as it actually appeared in the external review: publication number, relevant family member, cited passage or claim, publication date, and the issue for which it was cited.

Compare the later citation against both the original claim or feature set and the version under examination. Record any change in the question being assessed. If historical records are incomplete, mark the limits of the reconstruction rather than treating a new run as a reproduction.

Classify what the later citation reveals

First check the document’s legal relevance and historical availability, as set out below. A changed claim set, a justified scope boundary, or material unavailable at the time is not automatically a search failure. Where the evidence shows a retrieval or review gap, use one or more of these categories. Otherwise, record the changed circumstances or an unresolved cause.

The categories matter because each failure requires a different fix. Adding more synonyms will not repair an exclusion filter. Changing databases will not repair a misunderstood inventive point.

Check whether the document was retrievable at the time

Keep three dates distinct: the relevant date for the claimed subject matter, the original search date, and the rerun date. Then ask four questions:

• On what legal basis and for which claim or issue was the reference cited?
• Was the relied-on content, in the relevant document version or family member, publicly accessible and available in the searched source when the original search ran?
• Was it inside the original technical and document scope?
• Would the original searchable fields and filters have allowed it to appear?

Publication after the relevant date does not automatically make a patent reference unusable. EPC Article 54(3) and the U.S. AIA patent-document rules provide different routes for certain earlier-filed, later-published material. Check the applicable rule separately from whether the original search could have found the material.

A successful publication-number lookup confirms current database availability, not historical indexing or concept-based retrieval. Use saved results and screening records to distinguish a retrieval miss from a review-decision miss. Rerunning the original strategy today tests today’s retrieval; without historical evidence, it cannot establish which failure occurred originally.

Separate the baseline rerun from targeted correction

Reusing the original strategy and building a fresh strategy from the original disclosure are different tests. A fresh searcher or session can help assess independent retrieval if the expected reference is withheld; do not call a correction based on that reference blind.

Run three passes:

1. Baseline. Execute the original routes and settings without changes where possible. Record differences in the database, software, model, or settings that limit reproduction.
2. Correct. Make and log the changes supported by the diagnosis. The cited document may inform this pass; record that it did.
3. Stress-test. Check that the correction retrieves the cited document without overwhelming the result set with irrelevant material or tuning the search only to one publication. Where possible, also test relevant documents that were not used to make the correction.

In a public case, retrieval may lead to the applicant’s own publication or the examination file. Record that route rather than counting it as independent discovery of the cited prior art. Withholding a reference from the prompt alone does not establish a leakage-free test. Recovering a known document demonstrates retrieval under the tested conditions, not complete coverage or a proven historical cause.

The USPTO's examiner guidance illustrates why the record matters. MPEP 904 addresses the search and its updating; MPEP 719.05 provides the detailed search-history requirements, including search logic, databases, and dates. A private search has a different purpose and standard, but the documentation principle is useful: preserve enough detail to explain what was searched and what changed.

Turn one miss into a better search SOP

Close the audit with a short correction record:

• what the original search was intended to cover;
• what the later citation revealed;
• which gap, scope change, availability limit, or unresolved cause the evidence supports;
• what changed in the strategy or review process;
• whether the corrected search recovered the document, and under which conditions; and
• which limits or unresolved questions remain.

Then update the reusable part of the process. That may be an intake question, a required search route, a field-setting check, a second-review trigger, or a rule for documenting exclusions. Do not convert one unusual case into a universal checklist without testing whether the lesson transfers.

How Eureka can support the audit

Eureka Novelty Search supports review of search elements and routes, source-linked references, and comparisons with the extracted technical features. Confirm which patent and non-patent sources were actually included in the run. These views help reviewers distinguish the feature model, retrieval path, source evidence, and comparison rationale.

Here, a "blind rerun" describes a review method, not a named product feature. For a public case, start from the intended disclosure and record whether, and by what route, the reference is retrieved. The historical audit also needs the original search and screening records; source links alone do not recreate them.