Prior art is any evidence that your invention was already known to the public before your filing date. Under 35 U.S.C. 102, that means anything patented, described in a printed publication, in public use, on sale, or otherwise available to the public anywhere in the world, in any language, at any time before the day you file. Not before you invented it. Before you file. The United States has run on first-inventor-to-file since March 2013, and the calendar decides the case.
Designing for patentability means knowing what the record already holds and building your invention so that what is new about it can be claimed. That work happens before you file, not after a rejection arrives.
The Statutory Definition, in Plain Language
Section 102(a)(1) bars a patent if the claimed invention was:
- patented,
- described in a printed publication,
- in public use,
- on sale, or
- otherwise available to the public
before the effective filing date. Section 102(a)(2) adds a second category: an invention described in a US patent or published application filed by someone else before your filing date, even though that document did not publish until after you filed.
That second category is the one that catches people. A competitor filed 14 months ago. Their application publishes four months from now. It is prior art against you as of their filing date, and no search on earth can find it today. Every prior art search carries this blind spot, and anyone who tells you a search is exhaustive is selling something.
“Otherwise available to the public” is a catch-all with teeth. A YouTube video, a Reddit thread, a trade show booth, a doctoral thesis shelved in a German university library, a product listing that went live for six hours before being taken down. None of these need to be found by anyone. They need only to have been accessible.
What Counts, and What Surprises People
| Source | Prior art? | Notes |
|---|---|---|
| Issued US patent | Yes | As of its priority date, not its issue date |
| Published US application | Yes | Even if abandoned and never granted |
| Foreign patent in any language | Yes | Chinese utility models count in full |
| Journal article, textbook, manual | Yes | Printed publication, no circulation minimum |
| Product sold anywhere | Yes | On-sale bar |
| Trade show display | Yes | Public use and often public disclosure |
| Kickstarter or Indiegogo page | Yes | Public disclosure the day it goes live |
| YouTube video, blog post, forum thread | Yes | Publicly accessible |
| Your own published thesis | Yes | And it is prior art against you |
| Confidential disclosure under signed NDA | No | Not available to the public |
| Private prototype in your garage | No | Not public until you show it |
| A conversation with your brother-in-law | Depends | No confidentiality obligation means it is a public disclosure |
The last row is not a joke. A disclosure with no confidentiality obligation attached is a public disclosure. Whether anyone would find it later is not the legal test.
The Effective Filing Date Is the Whole Ballgame
Everything turns on one date. Your effective filing date is the day you filed the application in question, or the day you filed an earlier application that supports the same subject matter. A provisional filed in March 2025 gives a non-provisional filed in February 2026 an effective filing date of March 2025, but only for the subject matter the provisional described in enough detail to enable it.
That last clause does the damage. Inventors file a five-page provisional in March, redesign the product in July, then file a full application in February claiming the July version. The July features get February’s date, not March’s. Anything published between March and February is prior art against those features. What survives is decided by what the provisional described on the page, which is why a provisional patent application is worth exactly as much as the disclosure inside it.
The One-Year Grace Period, and What It Does Not Cover
Section 102(b)(1) gives you a limited exception. A disclosure made one year or less before your effective filing date does not count as prior art against you if it was made by you, by a co-inventor, or by someone who obtained the subject matter from you.
Four things this does not do.
It does not protect you against independent third-party disclosures. If someone else publishes the same thing during your grace period without having gotten it from you, that publication is prior art. Your grace period covers your disclosures, not theirs.
It does not exist in most of the world. Europe, China, and much of Asia apply absolute novelty. Publish before filing and you have destroyed your foreign rights on day one, permanently, with no cure. If international markets matter at all, treat the grace period as if it does not exist.
It does not stop the clock on someone else’s filing. Disclose publicly in June, and a competitor who sees it and files in August has an earlier filing date than you if you wait until October. You may have grounds to attack their filing as derived from you, and that is an expensive fight you would rather not have.
It does not restart. The one year runs from the first disclosure, not the most recent one.
The practical rule: file before you disclose. A provisional application costs $130 in USPTO fees at small entity rates and buys twelve months of “patent pending” status while you talk to buyers. The alternative is a legal argument about dates conducted years later with money on the table.
The On-Sale Bar, Including Secret Sales
An invention is barred if it was “on sale” before the effective filing date. Three details make this harsher than inventors expect.
An offer is enough. No sale needs to close. A commercial offer for sale of a product embodying the invention starts the clock.
Secret sales count. The Supreme Court settled this in 2019. A confidential sale to a single distributor under a nondisclosure agreement triggers the bar even though the public learned nothing. The theory is that you cannot commercially exploit an invention and then take a patent term on top of it.
One customer is a sale. There is no volume threshold. Selling three units at a farmers market in 2024 puts you on the clock.
What does not trigger it: experimental use, where the sale or use is genuinely for testing the invention rather than commercial gain. That exception is narrow, fact-heavy, and rarely as available as the inventor believes. Documented testing under a written agreement stands a chance. Selling units and calling it market research does not.
Anticipation Versus Obviousness
Two different rejections, two different standards, and inventors conflate them constantly.
Anticipation, section 102. A single prior art reference discloses every element of your claim, arranged as you claim it. One document, all elements. If a reference is missing one element of your claim, it does not anticipate. This is why claim drafting matters: a claim with seven elements is harder to anticipate than a claim with three.
Obviousness, section 103. Your claimed invention is not identical to anything in the record, but the difference between your invention and the prior art would have been obvious to a person of ordinary skill in the field at the time. Examiners may combine two or more references. This is the rejection that kills most applications.
The framework comes from a 1966 Supreme Court case and asks four questions: what does the prior art teach, what is the difference between it and your claim, what is the level of ordinary skill in the field, and are there objective indicators pointing toward non-obviousness. That last category includes commercial success, a long-felt need others failed to solve, and evidence that people skilled in the field tried and failed.
A 2007 Supreme Court decision widened what examiners can combine. Combining known elements according to known methods to yield predictable results is obvious. If your invention takes the handle from reference A and the hinge from reference B and the result behaves the way anyone would expect, expect a 103 rejection.
Designing for patentability means engineering toward the opposite: an interaction between elements that produces a result the prior art would not predict. That is a design decision made at the concept stage, and it is far cheaper than arguing about it during prosecution, where each office action response runs $1,500 to $4,000.
Prior Art You Create Yourself
The most common source of fatal prior art is the inventor.
- A crowdfunding page live for two weeks in 2023.
- A booth at a regional trade show with a working sample on the table.
- A LinkedIn post with photos of the prototype.
- Twelve units sold to a local retailer to test demand.
- A pitch to a manufacturer with no NDA signed.
- A product page put up “just to see if anyone clicks.”
Every one of these is a disclosure event with a date. Before you file, write down every date on which your invention was shown, described, offered, or sold to anyone outside a confidentiality obligation. Bring that list to whoever prepares your application. Lying by omission here does not protect you. It produces a patent that gets invalidated when it matters, which is worse than no patent at all.
Locking down confidentiality before you talk to companies is a separate discipline, covered in the protections to lock in before submitting an invention.
What a Search Can and Cannot Tell You
A prior art search finds documents. It cannot find:
- Applications filed in the last 18 months that have not published.
- Public uses and sales that were never written down anywhere.
- Foreign-language art that no indexing system has captured well.
- Non-patent literature outside the indexed journals.
What it can do is identify the closest published references and tell you which features of your invention are already in the record. That is the input to a real design decision. If three of your five features are in the art and two are not, you now know where to concentrate the disclosure and where the claims have to live. Running that search yourself starts with a structured Google Patents search, and it is worth doing before you pay anyone.
When the search does turn up a close reference, the next question is whether it blocks you or merely narrows you, and those are different outcomes with different responses. What to do when your search finds something close walks through the claim comparison.
Sequence: Search, Design, File, Disclose
The order that keeps you out of trouble:
- Search before you invest. Free databases first, then a professional search. The USPTO’s patent basics material covers what the office is looking for.
- Design toward the gaps. Once you know what the record holds, adjust the invention so the novel interaction is real and describable.
- File before you show anyone. A $399 patent search followed by a provisional puts a date on record for a few hundred dollars in fees.
- Then disclose. Trade shows, crowdfunding, buyer meetings, all of it. After the filing date, not before.
Inventors who run this in the wrong order do not usually find out for two years. The full cost sequence for each step is broken out in the inventor cost breakdown, and the search sits at the front of it because it is the cheapest place to learn bad news.
FAQ
Does my own product count as prior art against me?
Yes, after the grace period. Your own public disclosure starts a twelve-month clock in the United States and destroys foreign rights immediately in absolute-novelty countries. Selling, showing, or publishing before filing is the single most common self-inflicted wound in this process.
Can something unpatented be prior art?
Yes. A printed publication, a public demonstration, a product on a shelf, a manual, a video. Nothing has to be patented to block a patent. The test is whether it was available to the public before your effective filing date.
What if the prior art is in Chinese and nobody has translated it?
It counts. Prior art has no language requirement and no geographic limit. A Chinese utility model from 2015 describing your mechanism is prior art in a US examination, and examiners search foreign collections routinely.
How close does prior art have to be to stop me?
For anticipation, one reference has to disclose every element of your claim. For obviousness, two or more references can be combined if the combination would have been predictable to someone skilled in the field. The second standard is much easier for an examiner to meet, and it is where most rejections come from.
If I find damaging prior art, is the idea dead?
Not usually. It means the broad version is dead. The response is to identify what the reference does not disclose, determine whether that difference produces a result the art would not predict, and claim there. That is narrower protection, and narrow protection on a real point of novelty is worth more than broad claims that fall over on first review.