Can you patent software or an app idea?
Software can be patented in the US; an idea for an app, on its own, cannot.
Under 35 U.S.C. 101 and the Supreme Court’s 2014 Alice decision, claims that only run an abstract idea or a business practice on a generic computer are rejected, while claims to a specific technical improvement in how a computer or system works can pass.
Whether a particular invention clears that line is a question for your attorney, and how the application is written decides a lot of it.
The rule: Section 101
A patent is available for any new and useful process, machine, manufacture or composition of matter. Courts have carved out three exceptions: laws of nature, natural phenomena and abstract ideas. Most software rejections are about the third.
Software is usually claimed as a method, a system or a computer readable medium, and the question is whether the claim is really directed to an abstract idea dressed in computer terms.
The Alice test
In Alice Corp. v. CLS Bank (2014) the Supreme Court set a two step test. First, is the claim directed to an abstract idea, such as a mathematical concept, a way of organizing human activity or a mental process?
If so, does it add an inventive concept that makes it more than the idea carried out on a computer?
The USPTO examines with its own version: Step 2A asks whether the claim recites an exception and, if it does, whether it integrates it into a practical application; Step 2B asks whether it adds significantly more.
What tends to pass
Claims to an improvement in how the computer itself works, or in another technology: a better data structure, a faster or more reliable way to store, compress or route data, a security technique, a rendering method, a control loop for a physical device.
The Federal Circuit held eligible claims to a self referential database table in Enfish (2016) and to rules based lip sync animation in McRO (2016).
Since December 2025 the USPTO’s examination manual also points examiners to Ex parte Desjardins, a precedential decision crediting an improvement to how a machine learning model is trained.
What tends to fail
Claims to a business process, a financial or marketing method, or a way of organizing information, carried out on generic computers. An app described by what it does for the user, with no explanation of how.
Collecting, analyzing and displaying data with no specific technical way of doing it.
Applying a known technique to a new field: in Recentive Analytics v. Fox (2025) the Federal Circuit rejected claims that applied generic machine learning to a new kind of data.
How the claims are framed
Around the technical problem and the technical solution, not the business result.
The application explains what was wrong with the earlier way of doing it, the specific steps or architecture that fix it, and the effect: less memory, lower latency, fewer network calls, better accuracy.
Under current USPTO guidance the improvement has to be described in enough detail that a skilled engineer would recognize it, not just asserted, and the claim has to include the steps that deliver it.
If an examiner still rejects under Section 101, the USPTO has since December 2025 issued guidance on declarations that offer objective evidence of the improvement.
What to bring your attorney
The architecture diagram, the part of the code that does the new thing, what you tried first and why it did not work, benchmarks, and every variation you have considered.
Tao AI asks for exactly this when it writes your invention disclosure, and answers what it can from your code and documents, so you mostly confirm.
Section 101 is only the first gate
The invention also has to be new under Section 102 and not obvious over what exists under Section 103, and the application has to describe it fully under Section 112. A prior art search answers the first two questions before you pay for drafting.
Ours is a $300 report covering US and worldwide patents, published applications and technical literature, with an overlap map, within 2 business days.
