Software is one of the most valuable assets a technology company can own, but protecting it through patents in the United States is more complex than most people expect. Unlike a physical invention, software exists in a legal grey zone that has kept courts and patent offices busy for decades.
The good news is that software is absolutely patentable in the US. The rules around how to get there, though, require some careful navigation, particularly with the rise of AI.
Yes, but with conditions. Under Section 101 of the US Patents Act, the types of things that can be patented fall into four broad categories:
Software can fit within these categories, but courts have carved out several exceptions. The core challenge is that courts treat many software inventions as “abstract ideas,” which are not patentable on their own.
The way US patent examiners assess software eligibility today traces back to a 2014 US Supreme Court decision: Alice Corp. v. CLS Bank International. In that case, Alice Corporation held a patent on a computerised method for managing financial transactions, essentially using software to act as a trusted middleman in a trade. The Supreme Court struck it down, finding that it was just an abstract idea dressed up in computer language.
The decision gave rise to what’s now known as the Alice two-step test, which patent examiners apply to every software-related patent application.
The examiner first asks whether your invention, at its core, is really just an abstract idea, a law of nature, or a natural phenomenon. For software, the most common issue is abstract ideas like mathematical concepts, methods of organising human behaviour, or processes that could theoretically be done in someone’s head.
Take a piece of software that monitors stock prices and automatically buys or sells based on user-defined rules. An examiner might argue that the underlying process of checking a price and deciding to act is something a human could do mentally, with a computer just doing it faster. That kind of reasoning is what sinks many software patent applications.
If the claim does involve an abstract idea, it isn’t automatically dead. The examiner then asks whether the claim adds something genuinely significant that transforms it into a real-world, practical application.
This is where most software patents either survive or fail. The additional elements need to be substantial and meaningful, not just window dressing around an abstract core.
US courts have given us some useful reference points here.
In Apple Inc. v. Ameranth Inc. (2016), the court found that simply printing a menu generated by an abstract software process wasn’t enough of an addition to make the claim patent-eligible. Similarly, in Electric Power Group v. Alstom (2016), presenting collected data in a generic display wasn’t considered sufficient.
On the other hand, the Federal Circuit in BASCOM Global Internet v. AT&T Mobility (2016) found a software claim was patent-eligible. The invention filtered internet content, but the key was that it did so through a non-conventional arrangement of computer components, not just by taking an existing process and automating it.
If your software improves how computers work, or solves a technical problem in a genuinely new way, you’re in much stronger territory. If your software just uses a computer to do something that was previously done by hand or in someone’s head, you’ll face a tougher road.
The United States Patent and Trade Mark Office (USPTO) has included a helpful illustration in its guidance. Cattle monitoring software that assigns identifiers to each animal might be considered abstract on its own. But if that same software controls a gate mechanism to physically direct cattle to different locations based on the data, it now has a concrete, real-world application that makes it far more likely to be patent-eligible.
US courts have given us some useful reference points here.
In Apple Inc. v. Ameranth Inc. (2016), the court found that simply printing a menu generated by an abstract software process wasn’t enough of an addition to make the claim patent-eligible. Similarly, in Electric Power Group v. Alstom (2016), presenting collected data in a generic display wasn’t considered sufficient.
On the other hand, the Federal Circuit in BASCOM Global Internet v. AT&T Mobility (2016) found a software claim was patent-eligible. The invention filtered internet content, but the key was that it did so through a non-conventional arrangement of computer components, not just by taking an existing process and automating it.
If your software improves how computers work, or solves a technical problem in a genuinely new way, you’re in much stronger territory. If your software just uses a computer to do something that was previously done by hand or in someone’s head, you’ll face a tougher road.
The United States Patent and Trade Mark Office (USPTO) has included a helpful illustration in its guidance. Cattle monitoring software that assigns identifiers to each animal might be considered abstract on its own. But if that same software controls a gate mechanism to physically direct cattle to different locations based on the data, it now has a concrete, real-world application that makes it far more likely to be patent-eligible.
One of the more practically useful distinctions to come out of the USPTO’s 2024 guidance update is the difference between a claim that recites an abstract idea versus one that merely involves one.
A claim recites an abstract idea when it explicitly names or describes it. For example, if the claim language directly references a specific algorithm or mathematical formula. A claim that merely involves an abstract idea uses one as a foundation without making it the subject of the claim itself.
Why does this matter? Because if your claim only involves an abstract idea rather than reciting it, a full Alice analysis may not even be necessary. This distinction has opened up a useful drafting strategy for patent attorneys and is one of the reasons that how your claims are written matters just as much as what your software actually does.
If you’re working on AI or machine learning software, the landscape has become both more promising and more complicated in recent years.
The USPTO released significant new guidance in July 2024, specifically addressing how the Alice test applies to AI-related inventions. Simply adding “AI” or “machine learning” to a patent claim does not automatically make it patent-eligible. The underlying question is still whether the invention makes a concrete technical contribution.
This was reinforced by a significant Federal Circuit decision in April 2025 which many practitioners are treating as a cautionary tale. The patent in question involved using predictive analytics to generate TV advertising proposals. The court struck it down, finding that the invention was essentially applying a known analytical process to a new context using AI, making it a modern version of the “do it on a computer” problem that’s been fatal to software patents since Alice.
Claiming that your software uses AI isn’t a shortcut to patent eligibility. The AI implementation itself needs to represent a genuine technical advancement.
If you’re an Australian business looking to protect AI or software inventions in the US market, navigating these requirements alongside your Australian IP strategy requires careful coordination. Our international intellectual property team works with clients across both jurisdictions to ensure protection strategies are aligned from the start.
Focus on what your software actually improves
The strongest software patent claims are anchored in a specific technical problem that the software solves in a new way. If your software makes computers work better, faster, or more efficiently in a technically novel way, lead with that.
Avoid generic computer language
Claims that describe an abstract process and then say it’s implemented “on a processor” or “via a network” without more detail are easy targets for Section 101 rejections. Be specific about the technical mechanism.
Think about practical application from the start
The question “does this have a real-world, concrete application?” should be built into how you draft your claims, not treated as a patch to apply after a rejection.
How your claims are written matters enormously
The difference between a claim that recites an abstract idea and one that merely involves one can determine whether your application survives its first examination. This is an area where experienced patent attorney guidance makes a meaningful difference.
Software patent eligibility in the US is a fast-moving area of law, and getting your claims right from the start is far easier than fighting rejections down the track. At Kings, our patent attorneys stay across the latest USPTO guidance and case law so your application is positioned as strongly as possible from day one.
Get in touch with our team or book a meeting to talk through your software invention.
Yes. Software is patentable in the US, but claims must satisfy Section 101 of the Patents Act and pass the Alice two-step test. The invention must go beyond an abstract idea and demonstrate a genuine, practical technical contribution.
The Alice test is a two-step framework used by the USPTO to assess whether software-related patent claims are eligible for protection. Step one asks whether the claim is directed to an abstract idea. Step two asks whether there’s a real technical contribution that transforms the abstract idea into a patent-eligible invention.
Claims that describe a specific technical improvement to how a computer or system functions tend to hold up best. Vague claims that describe an outcome without explaining the technical mechanism are the most vulnerable to rejection.
