The meaning of patentable is narrower than ordinary use suggests and broader than most people expect in one specific way: it applies after grant as well as before.
In ordinary use the word means roughly "worth patenting" or "the sort of thing that gets patented."
In law it means capable of being validly patented — that every statutory condition is met, and would still be met if someone spent real money trying to prove otherwise.
That second sense is the operative one, because a patent that issues on an unpatentable invention is not safe. It is merely unchallenged.
The definition
Patentable, adjective: satisfying all conditions of patentability under Title 35 of the United States Code.
Four conditions. All must hold simultaneously.
| Condition | Statute | The question |
|---|---|---|
| Eligible subject matter | 35 U.S.C. §101 | Is this the kind of thing patents cover? |
| Novelty | 35 U.S.C. §102 | Has anyone disclosed it before? |
| Non-obviousness | 35 U.S.C. §103 | Would it have been obvious to a skilled person? |
| Adequate disclosure | 35 U.S.C. §112 | Is it described and claimed properly? |
Failing one is enough. A brilliantly non-obvious invention that is an abstract idea is unpatentable. A perfectly eligible machine already sold last year is unpatentable. The conditions do not trade off against each other, and strength in three does not compensate for failure in the fourth.
Utility sits inside §101 alongside subject matter. The bar is very low — some specific, substantial and credible use — and almost nothing fails it except inoperable inventions such as perpetual motion machines.
Patentable, patented, pending, unpatentable
Four words routinely used as if interchangeable. They are not, and the distinctions carry money.
| Term | What it means | Who says so |
|---|---|---|
| Patentable | Could validly be patented | An opinion, until tested |
| Patent pending | An application is on file | A fact about filing status |
| Patented | A patent has issued | A fact about the register |
| Unpatentable | At least one condition fails | An examiner, the PTAB, or a court |
Patent pending says nothing about patentability. It means an application exists. Anyone can file on anything; the USPTO does not screen for merit before according a filing date. See patent pending for what the status does and does not buy.
Patented does not settle patentability either, which is the point most often missed. A granted patent enjoys a statutory presumption of validity under §282, but a presumption is a starting position in an argument, not a conclusion.
Invalid and unpatentable mean the same thing in different forums. The Patent Office cancels claims it finds unpatentable. A court holds claims invalid. The underlying analysis is the same set of four conditions, applied at different standards of proof.
Patentable subject matter is only one condition
The narrowest and most confused sense of the word. Patentable subject matter refers specifically to §101 — whether the invention is a process, machine, article of manufacture or composition of matter, and whether a judicial exception applies.
The four categories are read broadly. Almost every physical invention fits one. Arguing that something is not a machine or a process is rarely productive.
The judicial exceptions are where §101 actually bites. Abstract ideas, laws of nature and natural phenomena are excluded, on the reasoning that they are the basic tools of technological work.
| Exception | Excludes | Field most affected |
|---|---|---|
| Abstract ideas | Mathematical concepts, fundamental economic practices, methods of organising human activity, mental processes | Software, fintech, business methods |
| Laws of nature | Natural correlations and relationships | Diagnostics, personalised medicine |
| Natural phenomena | Naturally occurring substances as found | Biotechnology, isolated compounds |
Software is the hardest case. Under the Alice two-step, a claim directed to an abstract idea survives only if its elements amount to significantly more than the exception. Generic computer implementation does not qualify. A specific improvement in how a machine operates does.
The single most useful drafting distinction: a claim reciting what is achieved is far more vulnerable than one reciting how it is achieved.
What can be patented works through the Alice test and the statutory categories in full.
Novelty and non-obviousness: where inventions actually fail
Subject matter gets the attention; §102 and §103 do the killing.
Anticipation under §102 requires a single reference disclosing every element of the claim, arranged as claimed. If one document does that, the claim is unpatentable. Nothing else matters.
Prior art is broader than inventors expect. Patents, published applications, journal articles, conference papers, product manuals, public demonstrations, sales, theses in a single library — anywhere in the world, in any language, whether or not anyone actually read it.
Your own disclosure counts against you. The US grace period under §102(b)(1) gives twelve months from your own public disclosure. Most other jurisdictions give nothing, so a conference talk or a sales offer usually destroys foreign rights outright while US rights are still alive.
Obviousness under §103 asks whether the differences between the invention and the prior art would have been obvious to a person of ordinary skill at the time. Multiple references can be combined, provided there was a reason to combine them and a reasonable expectation of success.
| Secondary consideration | What it tends to show |
|---|---|
| Commercial success | The market valued the claimed feature |
| Long-felt but unmet need | Others wanted the solution and did not find it |
| Failure of others | Skilled practitioners tried and could not |
| 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 tie to the claimed features rather than to price, branding or distribution. Commercial success driven by a marketing budget proves nothing about obviousness.
See prior art for what qualifies and how searches are run.
The word after grant
This is where "patentable" stops being a pre-filing concept.
Examination is a limited search. An examiner has a finite number of hours, searches a defined set of databases, and does not know which claim will matter commercially in eight years. A challenger facing an infringement suit has a direct financial incentive, a litigation budget, and the specific claim in front of them. They will look harder.
Four routes exist to test patentability after issue.
| Route | Who decides | Grounds available | Standard of proof |
|---|---|---|---|
| Ex parte reexamination | USPTO examiner | §102 and §103 only, on patents and printed publications | Preponderance |
| Inter partes review | PTAB | §102 and §103 only, on patents and printed publications | Preponderance |
| Post-grant review | PTAB | Any ground, including §101 and §112 | Preponderance |
| District court litigation | Judge or jury | Any ground | Clear and convincing |
Post-grant review has a nine-month window from issue and is the only post-grant Office proceeding that can reach §101 and §112. After nine months the available grounds narrow sharply to prior art documents.
Ex parte reexamination turns on a substantial new question of patentability — the statutory phrase is the word itself. Anyone can request one, including the patent owner, and the fee is a fraction of an IPR.
The standards differ and it matters enormously. Before the PTAB a challenger needs a preponderance of the evidence. In district court they need clear and convincing evidence to overcome the presumption of validity. The same prior art can cancel a claim at the Board and fail in court.
Institution is not automatic. The PTAB decides whether to institute review at all, and that rate has moved substantially: roughly 65% in October 2024, falling to roughly 37% by February 2026. Quote it with the date attached — the figure is driven by discretionary practice rather than by the merits of petitions, and it has been volatile enough that a stale number is actively misleading.
For what happens when claims are challenged, see patent prosecution.
Reading the word in an office action
Most people first meet "unpatentable" in a rejection, phrased in a form that is precise rather than hostile.
Claims 1–12 are rejected under 35 U.S.C. 103 as being unpatentable over Smith in view of Tanaka.
Decode it in three parts.
| Part | Says |
|---|---|
| "under 35 U.S.C. 103" | The ground — obviousness, not novelty or eligibility |
| "unpatentable over Smith" | The primary reference the examiner is building from |
| "in view of Tanaka" | The secondary reference supplying the missing element |
Two references named means §103, not §102. Anticipation requires one reference disclosing everything. The moment an examiner combines documents, they have conceded that no single one anticipates, and the argument moves to whether combining them was obvious.
A rejection is an opening position. Most applications receive at least one. Amending claims, arguing the combination lacks motivation, or submitting secondary-consideration evidence are all normal responses, and the majority of applications that receive a first rejection still issue.
"Rejected" applies to claims; "invalid" applies to patents. Applications get rejections. Issued patents get invalidity holdings. Using them interchangeably signals unfamiliarity in any negotiation.
Unpatentable over your own patent
Obviousness-type double patenting is the rejection people find most surprising: claims can be unpatentable over an earlier patent naming the same inventor or owner, even though your own patent is not prior art against you under §102.
The rationale is term extension. Without the doctrine, an owner could file successive applications on obvious variants and stretch exclusivity indefinitely past the twenty-year term.
The fix is usually procedural. A terminal disclaimer surrenders any term of the later patent extending beyond the earlier one and ties the two together in common ownership. It costs a fee and a signature, and it resolves most such rejections — at the price of a shorter effective term and a permanent link between the patents that matters when either is later sold separately.
Patentable is jurisdiction-specific
The word carries different content in different offices, which is why an invention can be patentable in the US and not in Europe on identical facts.
| United States | European Patent Office | |
|---|---|---|
| Statutory test | Four categories plus judicial exceptions | Invention must have technical character |
| Software | Eligible if not an abstract idea under Alice | Excluded "as such"; allowed if it produces a further technical effect |
| Business methods | Very difficult | Excluded as such; non-technical features ignored for inventive step |
| Diagnostic methods | Mayo problem for natural correlations | Methods practised on the human body excluded; products and in-vitro methods allowed |
| Grace period | 12 months for the inventor's own disclosure | None in practice |
The grace period difference is the one that destroys rights. A US inventor who presents at a conference and files eleven months later has a valid US application and no European one. The invention was patentable when conceived and unpatentable in most of the world by the time it was filed.
File before disclosing. It is the only sequencing that keeps every option open, and it costs nothing extra to get right.
Worked example: three inventions, one question
Same question asked of three inventions. Real analysis, not a summary.
A valve modification
The invention. A pressure relief valve with a secondary bypass channel positioned to equalise pressure across the poppet during transition, reducing chatter at low flow rates.
| Condition | Analysis | Result |
|---|---|---|
| §101 | A machine. No judicial exception in play | Passes |
| §102 | Search finds bypass channels in valves, none positioned for transition equalisation | Passes |
| §103 | The battleground. Is repositioning an existing feature obvious? | Arguable |
| §112 | Description supports the claim with dimensions and materials | Passes |
Verdict: patentable, with the fight on obviousness. Mechanical inventions almost never have eligibility problems, which is why prior art searching matters more than subject matter analysis in those fields.
Supporting evidence available. Two competitors tried spring redesigns and abandoned them. That is failure of others, and it goes directly to §103.
A parking app
The invention. A mobile application that predicts which parking spaces will be free, using historical occupancy data, and shows them on a map.
| Condition | Analysis | Result |
|---|---|---|
| §101 step one | Directed to collecting, analysing and displaying information — an abstract idea | Directed to an exception |
| §101 step two | Phone, server and map are generic computing. No technical improvement recited | Fails |
Verdict: unpatentable as claimed. The claim recites a result — predicting availability — rather than a technical mechanism.
What could change the answer. If the genuine innovation is a specific technique for compressing sparse sensor data so that prediction runs on-device within a memory bound, then claiming that — the partitioning scheme, the bounded accumulator, the reconciliation step — puts the claim on the improvement rather than on the business outcome. The underlying invention is identical. The patentability is not.
A biomarker correlation
The invention. The discovery that an elevated level of biomarker X in blood correlates with an early-stage condition, claimed as a method of measuring the level and correlating it with the condition.
| Condition | Analysis | Result |
|---|---|---|
| §101 step one | Directed to a law of nature — the natural correlation itself | Directed to an exception |
| §101 step two | Measuring and correlating are conventional steps | Fails |
Verdict: unpatentable. This is the Mayo problem. A newly discovered natural correlation is a discovery rather than an invention, however valuable and however expensive it was to find.
What survives. A claim to a specific novel detection technique, or to a method of treating the patient based on the correlation. Treating rather than merely diagnosing changes the analysis.
What patentable does not mean
It does not mean freedom to operate. Patentability asks whether you can get a patent. Freedom to operate asks whether you can sell the product without infringing someone else's in-force claims. Expired patents are prior art for patentability and irrelevant for clearance.
| Patentability | Freedom to operate | |
|---|---|---|
| Asks | Can I patent this? | Can I sell this? |
| Considers | All prior art, including expired | In-force claims only |
| Expired patents | Relevant | Irrelevant |
| Possible answer | Patentable but blocked | Unpatentable but clear |
You can hold a patentable improvement you are not permitted to build, because it falls inside an earlier claim. See freedom to operate.
It does not mean commercially worthwhile. Patentability has no market component at all. The clearest evidence is renewal behaviour, where owners vote with maintenance fees on patents they already own.
| Ipiry Patent Survival Curve v1.0 | Rate |
|---|---|
| Survive the 3.5-year fee (2022 cohort) | 85.8% |
| Survive the 7.5-year fee (2018 cohort) | 64.6% |
| Reach full term (2014 cohort) | 41.4% |
| Abandoned before full term | 58.6% |
Computed from 27,273,654 USPTO maintenance fee records covering 8,262,336 US utility patents. Every one of those abandoned patents was found patentable by an examiner. 58.6% were later judged not worth $8,280 — the third-stage large entity maintenance fee, 40% of that at small entity rates and 20% at micro.
Patentability is the entry condition, not the outcome. The full dataset is at the patent survival curve.
It does not mean granted. Roughly two thirds of applications eventually issue, but the route runs through at least one rejection for most of them. See how much does a patent cost for what that route costs.
Assessing patentability yourself
- Search the prior art first. This is where most inventions fail and it is the cheapest condition to test. Patents, published applications, journals, conference papers, product manuals, standards documents.
- Check what you have already disclosed. Presentations, sales, offers for sale, crowdfunding pages. If anything is public, the US twelve-month clock is running and foreign rights are probably already gone.
- Assess subject matter if you are in an affected field. Software, business methods and diagnostics need §101 addressed before anything else, because it can be fatal regardless of how novel the invention is.
- Ask whether the claim describes how or what. A claim reciting a result is vulnerable under §101 and hard to enforce. A claim reciting a mechanism is neither.
- Test enablement honestly. Could an engineer in the field build it from your description alone, without solving problems you have not addressed? If not, the disclosure is not yet sufficient.
- List your secondary considerations now. Failed competitor attempts, unexpected results, long-felt need. These are §103 evidence and they are far easier to document contemporaneously than to reconstruct in five years.
- Separate the patentability question from the value question. Ask whether the patent would be worth three maintenance fee payments before you commit to filing, not after the first one falls due.
- Get a patentability opinion before spending on a non-provisional if the answer to any of the above is uncertain. It is the cheapest step that can stop you spending the expensive ones.