How to Search Prior Art Before Filing a Patent: A Practical Guide for Engineers

Why Prior Art Search Comes Before Everything Else

You built something. It works. Nobody you know has seen anything like it. That feeling is real, and it is also not evidence.

Prior art is the term patent examiners use for anything that already exists in public that relates to your idea: patents, published applications, academic papers, product manuals, GitHub repos, YouTube demos, even a forum post from 2011. If it was publicly available before your filing date, it counts. The examiner will search for it. You should search for it first.

For technical professionals, this step often gets skipped or rushed, because the instinct is to trust your own knowledge of the field. That instinct is usually wrong, not because you are not knowledgeable, but because patent prior art lives in places engineers do not normally look: expired patents, foreign filings, academic archives, and product documentation that never made it into a paper or a demo video.

A good prior art search will not tell you whether your idea is patentable. Only a patent attorney can walk through that analysis with you, and even then it is a judgment call, not a yes/no switch. What a search does is give you an honest picture of the landscape before you spend money finding out the hard way.

What You Are Actually Looking For

Prior art search has one job: find the closest things to your idea that already exist, so you can see what is actually new about it.

You are not trying to prove your idea is original. You are trying to find out how it is different from what came before, and whether that difference is the kind of thing that tends to matter in a patent analysis. That distinction matters, because most inventors search until they find nothing and stop. A better approach is to search until you find the closest match, then study exactly how your approach diverges from it.

Where to Search

Patent Databases

Start here, even though it is the least intuitive resource for someone from a technical background.

• Google Patents is the easiest entry point. It indexes patents from dozens of countries, includes full text search, and shows citation links (who cited this patent, and what this patent cited). Following citation chains is often more useful than keyword search.
• USPTO Patent Public Search is the official U.S. database. It is less friendly than Google Patents but has more precise classification search tools.
• Espacenet (from the European Patent Office) is worth checking because a meaningful amount of prior art originates outside the U.S. and never shows up in a U.S.-only search.
• WIPO PATENTSCOPE covers international applications filed under the PCT (Patent Cooperation Treaty), which is a filing route many companies use before deciding where to pursue protection.

Classification Codes, Not Just Keywords

This is the part most technical searchers skip, and it is the part that matters most.

Patents are organized under a system called CPC (Cooperative Patent Classification). Every patent gets tagged with one or more codes describing what it does, regardless of what words the inventor used to describe it. Someone in 2003 may have built the same mechanism you built and called it something completely different.

Find one relevant patent through a keyword search, look at its CPC codes, then search within that classification. This surfaces prior art that no keyword combination would have found, because inventors do not all describe the same function the same way.

Non-Patent Literature

A huge share of prior art is not a patent at all.

• Academic papers (Google Scholar, IEEE Xplore, arXiv for preprints)
• Product documentation and datasheets from companies in your space
• Open source repositories and their commit histories, which carry timestamps
• Trade publications and conference proceedings
• Old product manuals, sometimes only available through the Internet Archive

For software and hardware ideas especially, a lot of the relevant art lives in places search engines do not prioritize. A product that shipped in 2015 and was discontinued in 2017 may still count as prior art even though it barely shows up in a normal web search today.

Building a Search That Actually Works

Start Broad, Then Narrow by Function

Describe your idea in plain, functional terms before you search anything. Not "my invention" language, but what it does, mechanically or procedurally. Patents are written in dense technical language, but the underlying function is usually simple. Search that function, not your product name or marketing language.

Use Synonyms Aggressively

Engineers tend to search using the vocabulary of their specific subfield. Patent drafters, especially older ones, often used deliberately broad or unusual language to cover more ground. Search every synonym you can think of, then search the synonyms of those synonyms.

Follow the Citation Trail

Once you find one relevant patent, check two things: what it cites (earlier prior art the applicant or examiner found relevant) and what cites it (later filings that reference it, often because they are trying to distinguish themselves from it). This trail frequently leads to the most relevant results faster than keyword search alone.

Search by Assignee

If you know which companies work in your space, search patents assigned to them directly. Companies file defensively as well as offensively, meaning they patent things they may never ship, specifically to block competitors. Those filings are prior art too, even if the product never existed commercially.

Documenting What You Find

Keep a simple log as you search: date searched, terms used, databases checked, and what you found, including the near-misses. This record does two things.

First, it prevents you from re-running the same dead-end searches later. Second, if you eventually sit down with a patent attorney for a strategy session, this record saves real time and money. An attorney working from your search log can go straight to the substantive question, how your approach differs from the closest prior art, rather than starting the search from zero.

What a DIY Search Cannot Do

Be honest with yourself about the limits of a self-run search.

A professional patentability search, the kind a patent attorney or a specialized search firm conducts, uses paid databases with more complete coverage, more precise classification tools, and search logic built over years of practice. It also comes with legal judgment about what a found reference actually means for your specific claims, which is not something a keyword match can tell you on its own.

Your own search is not a substitute for that. It is preparation for it. The goal is to walk into a conversation with an attorney already knowing the landscape, so the conversation is about strategy rather than starting from a blank page.

A Reasonable Order of Operations

1. Write a plain-language, functional description of your idea, no marketing language.
2. Search Google Patents and Espacenet using that description and every synonym you can generate.
3. Identify the closest matching patents, and pull their CPC classification codes.
4. Search within those classifications for additional matches.
5. Follow citation chains forward and backward from your closest matches.
6. Search non-patent literature: academic papers, product documentation, open source history, trade publications.
7. Log everything: dates, terms, sources, and what you found.
8. Compare your idea against the closest prior art you found, specifically on what is different, not just whether something similar exists.
9. Bring that record into a conversation with a patent attorney before deciding how, or whether, to file.

None of this tells you whether your idea may qualify for a patent. That is a legal judgment that depends on specific claims, specific prior art, and specific rules that a patent attorney is trained to apply. What a careful search gives you is something just as valuable at this stage: a clear, honest map of what already exists, so your next decision is based on evidence instead of hope.