What can be patented has a statutory answer and a practical one, and the gap between them is where most applications fail.
The statutory answer is in 35 U.S.C. 101: any new and useful process, machine, article of manufacture, or composition of matter, or any new and useful improvement thereof.
The practical answer is that meeting the category is easy and the three additional requirements — novelty, non-obviousness and adequate disclosure — are where inventions die.
The four statutory categories
| Category | Covers | Examples |
|---|---|---|
| Process | A series of steps or acts | Manufacturing methods, treatment methods, data processing methods |
| Machine | A device with interacting parts | Engines, instruments, apparatus, systems |
| Article of manufacture | A made thing | Tools, containers, components, fabricated items |
| Composition of matter | Chemical combinations | Compounds, alloys, formulations, materials |
Almost every physical invention fits one of these. The categories have been read broadly for a century, and arguing that something is not a machine or a process is rarely the productive line of attack.
Improvements count. You do not have to invent something from nothing — a new and useful improvement to an existing machine is squarely within section 101.
What matters far more is the judicial exceptions, which are not in the statute at all.
The three judicial exceptions
Abstract ideas, laws of nature and natural phenomena cannot be patented, on the reasoning that they are the basic tools of scientific and technological work and monopolising them would impede rather than promote innovation.
| Exception | What it excludes | Where it bites |
|---|---|---|
| Abstract ideas | Mathematical concepts, fundamental economic practices, methods of organising human activity, mental processes | Software, business methods, fintech |
| Laws of nature | Natural relationships and correlations | Diagnostics, personalised medicine |
| Natural phenomena | Naturally occurring substances as found | Biotechnology, isolated compounds |
These exceptions are the single largest source of rejection in the affected fields, and they are the reason two applications of similar technical merit can have entirely different prospects depending on subject matter.
The Alice two-step
The framework from Alice Corp. v. CLS Bank (2014), applied to every eligibility question.
Step one. Is the claim directed to a judicial exception — an abstract idea, law of nature or natural phenomenon? If not, the claim is eligible and the analysis ends.
Step two. If it is, do the claim elements, individually and as an ordered combination, amount to significantly more than the exception itself? Generic computer implementation does not. A specific technical improvement does.
What tends to pass:
- A claim to a specific improvement in how a computer operates
- A claim solving a technical problem rooted in technology, not business
- A claim requiring a particular machine or transforming a particular article
- A claim with a specific, non-conventional arrangement of components
What tends to fail:
- Automating a task humans previously did manually
- Collecting, analysing and displaying information, without more
- A fundamental economic practice implemented on a computer
- A claim reciting a result rather than how the result is achieved
The last point is the most useful drafting guidance available. A claim that says what is achieved is far more vulnerable than one that says how.
Novelty: section 102
The invention must be new. If a single prior art reference discloses every element of a claim, arranged as claimed, the claim is anticipated and cannot be patented.
Prior art is broader than most inventors expect. It includes patents, published applications, journal articles, conference papers, product manuals, public demonstrations, sales, and anything otherwise available to the public — anywhere in the world, in any language, whether or not anyone read it.
Your own disclosure counts against you outside the grace period. Under 102(b)(1) the US gives twelve months from your own public disclosure to file. Most other countries give nothing, so a conference presentation destroys European and Chinese rights immediately.
The on-sale bar catches people. A commercial offer for sale triggers the bar even where the invention itself was never disclosed in the transaction.
See prior art for what qualifies and how searches work.
Non-obviousness: section 103
The requirement that defeats most applications.
An invention is unpatentable if the differences between it and the prior art are such that the invention as a whole would have been obvious to a person of ordinary skill in the art at the time.
Several references can be combined, provided a skilled person would have had reason to combine them with a reasonable expectation of success. After KSR v. Teleflex (2007), that reason can come from market forces, design incentives or common sense rather than an explicit teaching — which made obviousness rejections substantially easier to make.
Secondary considerations can rebut it:
| Consideration | What it shows |
|---|---|
| Commercial success | The market valued the specific feature |
| Long-felt but unmet need | Others tried and failed |
| Failure of others | The solution was not obvious to skilled practitioners |
| Unexpected results | The outcome would not have been predicted |
| Copying by competitors | The solution was recognised as valuable |
| Industry praise | Contemporaneous recognition of significance |
A nexus is required. The evidence must connect to the claimed features rather than to marketing, price or brand.
Adequate disclosure: section 112
Three separate requirements, frequently confused.
Enablement. The specification must teach a skilled person how to make and use the full scope of what is claimed, without undue experimentation. A claim broader than what the description supports fails.
Written description. The specification must show the inventor possessed the claimed invention at the time of filing. This bites hardest where claims are broadened during prosecution beyond what was originally described — new matter cannot be added.
Definiteness. The claims must particularly point out and distinctly claim the subject matter, so that a skilled person can determine the boundaries. Terms of degree without a standard for measuring them are the usual failure.
Worked examples
A software invention that fails
Claim: A method of allocating advertising budget, comprising receiving campaign data, calculating an optimal allocation using a weighted formula, and displaying the result on a display device.
| Step | Analysis |
|---|---|
| §101 step one | Directed to an abstract idea — a fundamental economic practice combined with a mathematical calculation |
| §101 step two | "Display device" is generic computing. No technical improvement. Fails |
The problem is that the claim recites what is achieved, and the only non-abstract element is a conventional computer doing conventional things.
The same subject matter, drafted to pass
Claim: A method of reducing memory consumption in a campaign allocation system, comprising partitioning the campaign dataset into fixed-size blocks based on a cardinality threshold, processing each block in a single pass while maintaining a bounded accumulator, and merging block results using a specified reconciliation step.
| Step | Analysis |
|---|---|
| §101 step one | Arguably directed to a specific improvement in computer operation, not an abstract idea |
| §101 step two | If reached: the partitioning and bounded accumulator are a specific technical arrangement, not conventional. Better prospects |
What changed. The claim is now about how the computation is performed and what technical problem that solves — memory consumption — rather than about the business outcome.
The underlying invention may be identical. Eligibility frequently turns on what the claim is directed to rather than what was invented, which is an uncomfortable feature of the current framework and a real one.
A mechanical invention
Claim: A valve assembly comprising a housing, a spring-biased poppet, and a secondary bypass channel positioned to equalise pressure across the poppet during transition.
| Requirement | Analysis |
|---|---|
| §101 | A machine. No exception in play. Passes easily |
| §102 | Turns on whether any single reference discloses all three elements |
| §103 | The real battleground — is the bypass channel an obvious addition? |
| §112 | Straightforward if the description supports the claim |
Mechanical inventions rarely have eligibility problems. The fight is almost always obviousness, which is why prior art searching matters more than subject matter analysis in those fields.
A diagnostic method
Claim: A method of diagnosing a condition, comprising measuring the level of biomarker X in a patient sample and correlating an elevated level with the condition.
| Step | Analysis |
|---|---|
| §101 step one | Directed to a law of nature — the natural correlation between the biomarker and the condition |
| §101 step two | Measuring and correlating are conventional. Fails |
This is the Mayo problem, and it is why diagnostic method claims are difficult. A newly discovered natural correlation is a discovery rather than an invention, however valuable and however hard it was to find.
What can survive is a claim to a specific new detection technique, or to a method of treatment based on the correlation — treating rather than merely diagnosing.
Patentability is not freedom to operate
A different question entirely, and conflating them is expensive.
| Patentability | Freedom to operate | |
|---|---|---|
| Asks | Can I patent this? | Can I sell this? |
| Looks at | All prior art, including expired patents | In-force claims only |
| Expired patents | Relevant | Irrelevant |
| Answer can be | Yes while FTO is no | No while FTO is yes |
You can have a perfectly patentable improvement that you cannot legally build, because it falls within someone else's earlier claim. See freedom to operate.
Assessing your own invention
- Is it in a statutory category? Almost certainly yes if it is a physical thing or a series of steps.
- Is a judicial exception in play? Software, business methods and diagnostics need this assessed first, because it can be fatal.
- Has anything disclosed it already? Run a prior art search. This is where most inventions fail and it is the cheapest thing to check.
- Would it have been obvious? Harder to assess yourself, because hindsight makes everything look obvious.
- Can you describe it well enough to enable it? If you cannot explain how to build it, you cannot claim it.
- Have you disclosed it publicly? If so, check the twelve-month US grace period, and assume foreign rights are already gone.
- File before disclosing anything further — see patent pending for what a provisional does and does not buy.