What is a prior-art search?
By the ScoutIP team · Last updated July 26, 2026
A prior-art search is a structured search for everything already made public that is relevant to an invention - granted patents, published patent applications, and sometimes non-patent literature - so you can judge how new the idea actually is. "Prior art" is the body of knowledge that existed before a given date; if your invention is already fully described in it, it is not novel and cannot be patented.
In short: a prior-art search answers one question - has someone already disclosed this? - with evidence rather than a hunch.
Why a prior-art search matters
- Before filing a patent. A patent is only granted for something new and non-obvious (EPC Article 54, 35 U.S.C. 102). Finding close prior art early tells you whether a filing is worth the cost, or how to narrow the claims to what is genuinely new.
- Before investing or building. Prior art shows how crowded a technical field is and who else is working in it.
- To challenge a patent. The same search, run against an existing patent, can surface art that undermines its validity.
Novelty vs non-obviousness
A prior-art search informs two separate legal questions:
- Novelty: is the exact combination of features already disclosed in a single document?
- Non-obviousness (inventive step): even if no single document shows everything, would combining a few known documents be an obvious step for a skilled person?
See novelty vs freedom-to-operate for a related distinction that is often confused with novelty.
Keyword search vs semantic search
Traditional patent search matches keywords, so a patent that describes the same idea in different words is missed - a well-documented weakness, because inventors and examiners rarely choose the same vocabulary. A semantic search turns the invention into a meaning representation (an embedding) and finds patents describing the same concept regardless of wording.
| Keyword search | Semantic search | |
|---|---|---|
| Matches on | Exact words and synonyms you supply | Meaning of the whole description |
| Misses | Same idea, different words | Fewer conceptual matches |
| Effort | Manual query crafting | Plain-language description |
ScoutIP searches by meaning across a corpus of roughly 164 million patents from 106 patent offices, spanning all eras and languages.
How to run a prior-art search
- Describe the invention in plain language - what problem it solves and how, feature by feature.
- Search broadly, then narrow - cast a wide semantic net, then focus on the closest technical areas.
- Read the closest art against your feature list - a prior-art matrix makes this legible at a glance.
- Decide - file, narrow the claims, pivot, or take the strongest references to a patent attorney.
Frequently asked questions
How much does a prior-art search cost?
It ranges from free self-service semantic tools to several thousand euros for a professional attorney-run search. A software-based first pass (like ScoutIP) lets you screen an idea in minutes for a fraction of that, then reserve paid attorney time for the ideas that survive.
How long does a prior-art search take?
A manual professional search can take days to weeks. A semantic search engine returns the closest prior art in minutes - ScoutIP typically produces a mapped report in under 10 minutes.
Does a prior-art search guarantee my patent will be granted?
No. A search reduces uncertainty by surfacing the closest known art, but patent offices run their own examination and may cite references a search missed. Treat it as the best available evidence, not a guarantee.
Can I do a prior-art search myself?
Yes, for a first pass. Free tools like Google Patents and semantic engines let you screen an idea. For a filing decision, confirm the results with a qualified patent attorney.
Run one
Describe your invention in plain language and let the search find and map the closest prior art for you.