Introduction.

You’ve built something new. Maybe it’s a piece of software, a medical device, a new process for manufacturing a product, or a diagnostic test that detects disease faster than anything on the market. Now you’re wondering: can I patent this? Should I?

These are some of the most common questions we hear from startups, and the answers are often more nuanced than a simple yes or no. This guide will walk you through the basics of what the U.S. patent system protects, address the persistent confusion around software patents, and help you understand one of the most challenging legal hurdles in patent law today: the Section 101 eligibility rejection.

The foundation: what is a patent, and why does it matter?

A patent is a government-granted right that allows you to exclude others from making, using, selling, or importing your invention for a limited period — generally 20 years from the date you file your application. In exchange, you disclose how your invention works to the public.

Patents are powerful. They can deter competitors, attract investors, generate licensing revenue, and form the backbone of a company’s valuation, especially in R&D-heavy spaces. But they are not free, not fast, and not guaranteed. Understanding what is and isn’t patentable before you invest in the process is critical.

The four categories of patentable subject matter.

Under U.S. patent law (specifically 35 U.S.C. § 101), patent protection is available for anything new and useful in the following categories:

• Process: A series of steps or acts, such as a method of manufacturing a chemical compound, a surgical technique, or a business method implemented by a computer.

• Machine: A concrete device or apparatus with moving or operative parts, such as an engine, a circuit board, or a medical scanner.

• Manufacture: An article of manufacture, meaning a tangible item made from raw materials, such as a new type of fabric, a tool, or a smartphone component.

• Composition of Matter: A combination of ingredients or substances, such as a new drug formulation, a polymer blend, or a genetically modified organism.

If your invention falls into one of these categories and it is new (novel), non-obvious to someone skilled in your field, and useful, it is a strong candidate for a patent.

What cannot be patented.

Just as important as knowing what can be patented is knowing what cannot. The law carves out three major exceptions:

• Abstract ideas: Mathematical concepts, certain mental processes, and methods of organizing human activity are not patentable on their own. This is where software often runs into trouble (more on this below).

• Laws of nature: You cannot patent a natural phenomenon, such as the relationship between a gene mutation and a disease, or the fact that gravity exists.

• Natural phenomena: Naturally occurring substances — a mineral found in the earth, a naturally isolated protein — are generally not patentable without meaningful human modification or application.

These exclusions exist because allowing anyone to own foundational principles of nature or mathematics would stifle, rather than encourage, innovation.

The software question: Can you patent software?

This is easily the most misunderstood area of patent law, and the answer is: it depends — but often yes, with the right approach.

Software itself, as a set of abstract instructions, is not patentable on its own. You cannot patent an algorithm in the abstract, a mathematical formula, or a purely mental process. However, software that produces a concrete, real-world result — especially when it is tied to specific hardware, improves the functioning of a computer itself, or solves a specific technical problem in a novel way — can absolutely be patented.

Think of it this way: you cannot patent “sorting data.” But you may well be able to patent a specific, novel method of compressing video data that reduces bandwidth consumption by 40% while maintaining resolution, implemented through a defined series of computational steps. The key is that the invention must do something specific and technical, not merely automate an abstract idea.

What kinds of software are most likely patentable?

Software that tends to fare well in the patent system includes:

• Algorithms that improve computer performance, speed, or efficiency in a technically defined way.

• Novel data compression or encryption methods with specific technical implementations.

• Machine learning architectures applied to solve a defined technical problem (not simply “use AI to predict X”).

• Software that controls physical systems, such as autonomous vehicle navigation, industrial robotics, or medical device operation.

• User interface innovations that solve a specific usability or technical problem in a non-obvious way.

• Network security methods that prevent specific categories of cyberattacks through defined technical steps.

Software that tends to struggle includes: generic “do it on a computer” claims, basic data collection and display without meaningful transformation, and purely mathematical operations with no clear technical application beyond the math itself.

The toughest hurdle: Section 101 rejections.

Section 101 of the U.S. patent statute defines patent-eligible subject matter, and it is the source of some of the most contentious and unpredictable rejections patent applicants face today. Understanding why requires a brief look at two landmark Supreme Court cases.

Alice and Mayo: The cases that changed everything.

In Alice Corp. v. CLS Bank International (2014), the Supreme Court held that a computer-implemented method for mitigating settlement risk in financial transactions was not patentable because it was merely an abstract idea — a longstanding commercial practice — implemented on a generic computer. The Court established a two-step test (now called the Alice/Mayo framework) for evaluating patent eligibility:

• Step 1: Is the claim directed to a patent-ineligible concept — an abstract idea, law of nature, or natural phenomenon?

• Step 2: If so, does the claim contain an “inventive concept” — something significantly more than the ineligible concept itself — that transforms the nature of the claim into a patent-eligible application?

In Mayo Collaborative Services v. Prometheus Laboratories (2012), the Court struck down patents covering methods of determining optimal drug dosage based on metabolite levels in the blood. The Court found that the patents simply described a natural relationship between drug metabolites and health outcomes, with conventional steps added that did not meaningfully limit the claims.

What a 101 Rejection looks like in practice: Software.

Imagine a startup that develops a new app for detecting fraudulent credit card transactions. They apply for a patent on their method. The examiner issues a 101 rejection arguing that the claims are directed to the abstract idea of “detecting fraud” — a mental process humans could perform — implemented on a generic computer, without any inventive concept beyond the abstract idea itself.

To overcome this rejection, the applicant would need to show that the claims are directed to a specific technical implementation: for example, a method that uses a specific type of neural network architecture trained on a defined class of behavioral signals, operating in real time on transaction data streams in a way that reduces false positives below a defined threshold. The more the claims are tied to a specific technical solution — rather than a result — the better the chances of survival under Alice.

Another common software example involves personalization algorithms. A claim directed to “personalizing content for a user based on their preferences” is almost certain to receive a 101 rejection. That is an abstract idea. But a claim directed to a method that uses a defined content-ranking algorithm that weights recency, interaction frequency, and co-occurrence patterns within a specific data structure, delivered through a specific API architecture, is far more likely to pass muster — because it solves a technical problem with a specific technical solution.

What a 101 Rejection looks like in practice: Diagnostic patents.

The diagnostic context is where Section 101 rejections are perhaps most painful — and most consequential. After Mayo, patents claiming methods of diagnosing a disease or condition by observing a natural correlation (e.g., a correlation between a biomarker level and a disease state) face an extremely high bar.

Consider a company that discovers that elevated levels of a specific protein in the blood are strongly predictive of early-stage pancreatic cancer. They want to patent a diagnostic method. The examiner will almost certainly issue a 101 rejection arguing that the correlation between the protein and the disease is a law of nature, and that the steps of measuring the protein and comparing it to a threshold are conventional laboratory techniques that add nothing “significantly more.”

This is a genuine and difficult problem. In the years since Mayo, courts have struck down diagnostic patents covering genuinely innovative discoveries, on the grounds that the natural relationship between a biomarker and a disease cannot be owned. Critics — including many in the medical and scientific communities — argue that this has had a chilling effect on investment in diagnostic innovation.

To have the best chance of surviving a 101 rejection in the diagnostic context, applicants should consider:

• Tying the method to a specific, unconventional detection technique (not just a standard lab assay).

• Claiming a specific treatment step that follows from the diagnostic result, creating a more integrated method.

• Emphasizing any unconventional steps in sample preparation, data analysis, or result interpretation.

• Framing claims around a specific technical improvement to the diagnostic process, not just the underlying natural correlation.

The honest reality is that purely correlational diagnostic methods face very difficult odds after Mayo, even when the underlying discovery is genuinely breakthrough science. This is one area where working closely with a patent attorney/agent from the very beginning of a research program can shape how discoveries are documented and framed in ways that give you a meaningful chance at protection.

Other things you may rot realize can be patented.

Patent law is broader than many people assume. Beyond traditional inventions, the following are often patentable:

• Business methods: Methods of conducting business, when implemented through a specific technical process, can be patented. Think of Amazon’s one-click purchasing patent, which was a specific technical implementation of an ordering process.

• Biological inventions: Genetically modified organisms, synthetic DNA sequences, engineered proteins, and novel microorganisms created through human ingenuity (not isolated from nature) can be patented. The Supreme Court’s Myriad Genetics decision (2013) drew a clear line: naturally occurring DNA sequences cannot be patented, but synthetic cDNA sequences can.

• Plants: Novel plant varieties created through human breeding or genetic engineering can be protected under plant patents or plant variety protection.

Should you patent your invention? A practical framework.

Here are the key questions to ask when deciding whether to pursue patent protection:

• Is it truly novel? If something very similar already exists in the prior art, you may be wasting resources.

• Can you describe it technically enough? The vaguer your innovation, the harder it is to claim. For instance, if you can only describe what your software does (the result), not specifically how it does it, you may have a 101 problem before you start.

• Is the competitive landscape worth it? If your competitors are unlikely to copy your specific implementation, trade secret protection may serve you better.

• What is your business model? If you license technology, patents are often essential. If you are building a product and rely on speed-to-market and continuous innovation, a portfolio strategy might be more important than any single patent. Additionally, if you are a startup, patent protection or simply a pending application may be essential to securing investors. Many investors do not want to invest significant money into a company if there is not adequate patent protection in place as competitors may copy the product.

• How fast is your space moving? In some technology areas, the time it takes to get a patent means the technology is obsolete by the time the patent issues. In others, a patent granted years from now may still be highly valuable.

The bottom line.

Patent law is designed to reward genuine innovation, but navigating it requires understanding its boundaries and quirks. Software can be patented, but only when it solves a specific technical problem with a specific technical solution. Diagnostic methods can be patented, but the bar is high and has gotten higher since Mayo. Natural phenomena and abstract ideas, in their raw form, belong to everyone.

The best time to think about patent strategy is before you finalize your product architecture, before you publish research results, and before you talk to investors without a non-disclosure agreement in place. Early strategic thinking can make the difference between a portfolio of enforceable patents and a set of expensive rejections.

This article is intended for general informational purposes and does not constitute legal advice. Patent eligibility analysis is highly fact-specific, and you should consult a registered patent attorney/agent regarding your particular invention and circumstances.