Prior art search basics for inventors

A prior art search looks for anything made public before your filing date that shows your invention, or something close to it, already exists. For an inventor, the point is to learn early whether the idea is likely to be new, and to understand the field before spending money on an application.

This guide explains general concepts. It is not legal advice. Patent law differs between countries, and you should consult a registered patent attorney or patent agent before making decisions.

What counts as prior art

Prior art is broader than most people expect. In general it covers anything made available to the public before the relevant date, in any country and any language. That can include:

• Granted patents and published patent applications, including expired and abandoned ones
• Journal articles, conference papers and theses
• Books, manuals, product brochures and standards
• Websites, videos and online forum posts
• Products that were on sale or on public display

Your own earlier disclosure can count too. If you present the invention at a trade show or publish a paper before filing, that may be used against your own application. Some countries allow a grace period for an inventor's own disclosures and others allow very little or none, and the rules differ in detail. This is one of the first things to raise with a patent attorney, ideally before you show the idea to anyone outside a confidentiality agreement.

Novelty and obviousness, in general terms

Two tests matter most when prior art is weighed against an invention.

Novelty. An invention is not new if a single earlier disclosure shows every feature of what you claim. One document, all the features.

Obviousness. Even if no single document shows everything, a patent can be refused if the difference between your invention and what was known would have been obvious to a person with ordinary skill in the field. Examiners may combine more than one document to make this case. Many countries call this test inventive step.

This has a practical effect on searching. You are not only looking for an exact copy of your idea. You are looking for the closest things that exist, because the gap between them and your invention is what a patent would have to rest on. Finding close prior art is not a failure. It tells you, and your attorney, where the real invention lies.

A search method you can follow

1. Write the invention down as features

Describe the problem, the solution and the features that make the solution work. A list of five to ten features is a good working tool. It stops the search drifting, and it is what you will compare each result against.

2. Build a keyword list

For each feature, list synonyms, broader terms and narrower terms. Patent language is often more abstract than everyday speech. A screw may be a "fastener" or a "fastening means". A phone may be a "portable communication device". Include the vocabulary of neighbouring industries, since the same mechanism may have been invented for a different purpose. Add both British and American spellings.

3. Run keyword searches

Use a free public patent database. Several patent offices and the World Intellectual Property Organization run search tools that anyone can use without charge, and there are free search engines from commercial providers too. Coverage, search syntax and translation quality differ between them, so it is worth using more than one.

Start broad, then combine terms to narrow down. Skim titles, abstracts and drawings to decide what deserves a closer read.

4. Find the classification codes

Every patent document is sorted by patent offices into classes that describe its technical subject. Two schemes are widely used. The International Patent Classification is administered by the World Intellectual Property Organization. The Cooperative Patent Classification is managed jointly by the European Patent Office and the United States Patent and Trademark Office, and is more finely divided. Both are hierarchies, running from broad sections down to narrow groups.

When you find a relevant document, note its classification codes. Look up the definitions of those codes in the scheme, and you will often find a group that describes your field closely.

5. Search by classification

Searching within a class finds documents that use words you did not think of, including documents translated from other languages. This is the main reason professionals rely on classification. Combine a class with a few keywords to keep the list manageable.

6. Follow the citations

Each patent lists earlier documents cited against it. These are backward citations. Later documents that cite it are forward citations. Following both from a close result is one of the fastest ways to map a field, because examiners and applicants have already done part of the linking for you.

7. Look beyond patents

Search academic literature, product catalogues, standards and the general web. In fast-moving fields, much of the prior art never appears in a patent.

8. Keep a search log

Record which databases you used, the date, each query and which results you reviewed. Note why the close ones are close, feature by feature. A log stops you repeating work. It also gives your attorney a useful starting point.

Reading what you find

When judging whether a document is prior art against your idea, the whole document counts: description, drawings and claims alike. The claims tell you what the owner protected. The description tells you what was disclosed to the public, which is often much more. To understand the claims themselves, see how to read a patent claim.

Compare each close document against your feature list. A simple table with features down the side and documents across the top shows at a glance which features are already known and which are not.

Limits to keep in mind

• No search proves novelty. A search can find a document that shows your idea. It cannot prove that none exists.
• Recent applications are hidden. Patent applications are generally published around eighteen months after their earliest filing date. Until then, nobody outside the patent office can see them.
• Language and wording. Relevant documents may be in another language, or may describe the same idea in terms you did not try.
• Patentability is not freedom to operate. A prior art search asks whether your idea is new. Whether you can sell a product without infringing someone else's patent in force is a separate question, with a different kind of search.
• A self-search informs. It does not decide. Professional searchers and patent attorneys have training, tools and judgement that a first-time searcher lacks.

Where Prism fits

Prism Labs is building its own patent search product. Its corpus holds 9.8 million US patents and 109 million claims, each traceable to source text. The corpus is loaded and verified, and the search product is still being built. Prism's engine requires AI answers to cite stored evidence, an approach explained in why AI answers need sources. There is more on the intellectual property page, and early access is by waitlist.