This is a different question from the one about coverage most teams ask. Testing whether your terms and classifications were wide enough asks whether the strategy reached the documents inside the collections it searched. This guide asks which collections it searched at all.
Key takeaways
- A result set shows what was found, but may not show what was outside the search. Distinguish the intended scope from a failure to retrieve material within it.
- Papers, preprints, standards, repositories, and product documentation can contain relevant disclosures not present in the patent documents retrieved. Check which of these sources the search actually covered.
- Check the document universe before you interpret the result count. A conclusion must reflect both the sources searched and the limits of the evidence found.
A search has a shape
Every search has a boundary. A result list may not make it obvious which collections were queried, so inspect the source coverage and search record rather than inferring the boundary from the patents returned.
Two distinctions help you inspect that boundary.
Publication timing differs. A paper, preprint, manual, or code release may become public before a corresponding patent publication, or may have no corresponding patent document. Check the actual dates and disclosures; do not assume that one source type is always earlier.
Classification coverage depends on the collection. Some patent-search systems include non-patent literature and assign classifications to it. WIPO’s PATENTSCOPE training material, for example, describes IPC classification for NPL. That does not mean all relevant NPL is indexed or classified. Check the collection and fields used by your search.
The useful distinction is between a patent-only search and a search that includes specified non-patent sources. A tool’s name does not tell you which one you ran.
Four universes outside the patent record
- Academic and technical literature. Journal articles, conference papers, preprints, and theses. Their references and bibliographic records can provide additional search routes.
- Standards. Published standards and some technical drafts can contain detailed disclosures. Check the standards body, the version, public availability, and whether the search service includes that material.
- Code and repositories. In software and machine learning, a substantial amount of technical work is released as code with accompanying documentation. Whether a particular repository counts as a public disclosure is a separate question, and the guide on using a document as prior art covers it. A search that never reaches repositories never gets to ask.
- Product and industry documentation. Manuals, datasheets, white papers, and technical blog posts. Not academic, and still capable of constituting a public disclosure in the legal sense if the material was genuinely available to the public.
These are source groups to check, not collections that every patent-search service necessarily excludes. Some are available through integrated databases; others need a separate route. Record what was searched in this run.
Where this matters most
The useful sources depend on the subject. The examples below are prompts for planning coverage, not a ranking of industries by how much their prior art lies outside patents.
| Subject | Non-patent sources worth checking |
|---|---|
| Software and machine learning | Conference papers, preprints, technical documentation, and relevant public code releases |
| Biotechnology and life sciences | Journal articles, conference abstracts, theses, and relevant technical or product documentation |
| Telecommunications and electronics | Published standards, publicly accessible technical contributions, papers, and product specifications |
| Materials and chemistry | Research papers, theses, supplier datasheets, and technical manuals |
| Established mechanical engineering | Product manuals, catalogues, standards, technical papers, and evidence of relevant public use |
A patent-only search excludes these source groups unless they were reached through another recorded route. Whether that omission materially affects the assessment depends on the subject and the decision the search is intended to support.
Look beyond the result count. Use a Novelty Search to inspect the references returned, then check what you can establish about the sources searched.
How to check your own run
Start with the existing record. The known-document check may also require a targeted lookup or a rerun.
- Read the source list, not the result count. Which collections were queried? If the run does not say, that is the finding. A search whose sources cannot be named cannot be scoped.
- Ask how a relevant non-patent document could be found. Check which collections and fields were available to the strategy. A classification route may retrieve selected classified NPL, but it cannot establish coverage of papers, standards, or repositories that were never in its searchable collection.
- Test with a document you already know. First confirm that the document belongs within the intended scope and is available in the selected collection, for example through an identifier lookup. Then check whether the substantive search strategy retrieves it without adding its title or identifier as a shortcut. A miss is a retrieval failure to investigate; it does not by itself identify whether the cause is collection coverage, indexing, query design, or filters. Finding one document does not prove completeness.
- Write down what was not searched. State the excluded source groups and the reason for each material limitation. A patent-only search can serve a defined purpose, but naming that scope does not make every conclusion drawn from it justified.
What this does not tell you
Adding source groups can improve the evidence available, but it does not guarantee that every relevant disclosure has been found. Record the sources and remaining limits rather than describing the search as complete.
It also does not resolve whether a given document qualifies as prior art at all. A repository commit, a conference slide deck, or an internal manual that became publicly accessible can each raise questions about date and availability that the retrieval itself does not settle. Those are covered separately in the guide on whether a document can be used as prior art.
And when a conclusion needs to be defended, the question of what a particular disclosure establishes is a legal assessment, not a retrieval one.
How to inspect search coverage 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.
For a coverage review, inspect the available search routes and references, then distinguish what you can confirm was searched from what remains unverified. A cited paper does not establish coverage of standards, repositories, or product documentation. Record material gaps and arrange separate searches where needed; the interface shown below does not establish coverage of any particular non-patent collection.
This article is general information about search method. It is not legal advice. Whether a particular disclosure constitutes prior art depends on the applicable law, the relevant date, and the jurisdiction, and should be assessed with qualified professional advice.
