Patent Subject Matter Eligibility: 2-Step Alice Test

You spent two years building something genuinely new, filed a patent, and then the examiner says your invention is an “abstract idea” that nobody can own. Welcome to the most frustrating doctrine in patent law. Patent subject matter eligibility under 35 U.S.C. § 101 decides whether your invention is even the kind of thing a patent can protect — before novelty or non-obviousness ever come up. Most guides explain the rule and stop. This one shows you how the two-step Alice test actually works, where software, business methods, and diagnostics fall, and the concrete drafting moves that get a claim across the line.

What Patent Subject Matter Eligibility Means Under § 101

Section 101 of the Patent Act says you can patent “any new and useful process, machine, manufacture, or composition of matter.” Those four buckets are the statutory categories. If your claim doesn’t fit one of them, it’s out before the analysis even starts — though in practice that’s rarely the problem.

The real fight is over a rule the Supreme Court read into the statute. Over more than a century, the Court carved out three things you cannot patent no matter how you claim them. These are the judicial exceptions, and patent subject matter eligibility turns almost entirely on whether your claim falls into one of them:

• Abstract ideas — mathematical concepts and formulas, certain methods of organizing human activity (fundamental economic practices, commercial interactions), and mental processes a person could perform in their head or with pen and paper.
• Laws of nature — naturally occurring principles and relationships, such as the correlation between a drug dose and a metabolite level in the blood.
• Natural phenomena — products of nature that haven’t been meaningfully transformed, like an isolated gene sequence or a naturally occurring bacterium.

The logic is that no one should monopolize the basic tools of science and commerce. The hard part is that almost every invention uses an abstract idea or a natural law somewhere. The question is never “does the claim touch one,” but “does the claim do enough beyond it.”

The Alice/Mayo Two-Step Test in Plain English

The framework comes from two Supreme Court cases. Mayo Collaborative Services v. Prometheus Laboratories (2012) supplied the structure for laws of nature; Alice Corp. v. CLS Bank International (2014) extended it to abstract ideas and gave us the version examiners and courts use today. It’s a two-step inquiry.

Step 1: Is the claim directed to a judicial exception? The examiner reads your claim and asks whether, at its core, it’s aimed at an abstract idea, a law of nature, or a natural phenomenon. A claim to “a method of hedging financial risk” is directed to a fundamental economic practice — an abstract idea. If the claim isn’t directed to an exception at all, it’s eligible and the analysis stops. If it is, you move to step two.

Step 2: Is there an inventive concept — something significantly more? Here the examiner looks at the claim’s elements, individually and as an ordered combination, to see whether they add enough to transform the exception into a patent-eligible application. Generic computer hardware doing what computers normally do is not enough. Reciting the exception and then saying “apply it” is not enough. You need a concrete, non-conventional implementation.

The clearest contrast in the case law is Diamond v. Diehr (1981), still cited as the eligible side of the line. The claim used a well-known mathematical equation but tied it to a specific process for curing rubber — constantly measuring mold temperature and using the calculation to control when the press opened. The claim wasn’t trying to own the equation; it improved a physical manufacturing process. That’s an inventive concept.

How the USPTO Examines Eligibility: The 2019 Guidance Framework

Examiners don’t apply Alice freehand. They follow the 2019 Revised Patent Subject Matter Eligibility Guidance, now baked into MPEP § 2106. The guidance breaks the two-step test into a more predictable flow, which is good news because it gives you specific places to win.

1. Step 1 — Does the claim fall within a statutory category (process, machine, manufacture, composition)? Almost always yes.
2. Step 2A, Prong One — Does the claim recite a judicial exception? The examiner checks whether it recites a mathematical concept, a method of organizing human activity, or a mental process. If no, the claim is eligible.
3. Step 2A, Prong Two — If it does recite an exception, do the additional elements integrate it into a practical application? A claimed improvement to the functioning of a computer or another technology gets you out here, before you ever reach the inventive-concept search.
4. Step 2B — If the claim is still directed to an exception, do the additional elements amount to significantly more — an inventive concept that’s more than well-understood, routine, conventional activity?

The practical takeaway: Prong Two is your best friend. If you can frame your invention as a specific technical improvement, you can secure eligibility at Step 2A without ever arguing about “inventive concept.” That’s why how you draft the improvement matters so much — more on that below.

In July 2024, the USPTO issued a guidance update addressing AI-assisted inventions, with three new worked examples (Examples 47–49). The headline point: the fact that an invention was developed with help from AI does not affect its eligibility. What matters is the claimed invention itself — whether it recites an exception and whether it improves a technology. If you’re working in this space, our guide to patenting AI inventions goes deeper on how machine-learning claims fare under § 101.

Eligible vs. Ineligible: Concrete Examples That Set the Line

Abstract rules are hard to apply. Real fact patterns are easier. Here’s how the line tends to fall across the three areas where § 101 bites hardest.

Software. A claim to “collecting data, analyzing it, and displaying the result” on a generic computer is the classic ineligible software claim — it automates a mental process without improving the computer. Compare a claim that changes how a database is structured to run searches faster, or improves how a network handles traffic. Courts have upheld claims directed to a specific improvement in computer functionality. The dividing question is whether the computer is the tool or the thing being improved — see our deeper treatment of what can and cannot be patented in software.

Business methods. This is the hardest category. Bilski v. Kappos (2010) struck down a claim to hedging risk in commodities trading as an abstract idea, and Alice did the same to computerized escrow settlement. A business method claimed at the level of “the idea, plus a computer” almost always fails. The rare survivors solve a problem that’s specifically rooted in technology — for example, a way of doing something on the internet that overrides how the internet conventionally behaves — rather than just moving an offline business practice online.

Diagnostics. Mayo set the trap here. A claim that observes a natural correlation — “measure the metabolite, and if it’s above X, adjust the dose” — is directed to a law of nature, and routine “determining” and “administering” steps don’t save it. Eligible diagnostic claims generally need a new, non-conventional technique to detect or apply the correlation, or a specific treatment step that does more than tell a doctor to think about a natural relationship.

Drafting Tactics That Survive a § 101 Rejection

This is where most guides go quiet. Eligibility isn’t only decided by what you invented — it’s heavily shaped by how the claim and specification are written. These are the moves that consistently improve your odds before the examiner ever picks up the file.

• Frame a specific technical improvement. Don’t claim a result (“reduce fraud,” “improve efficiency”). Claim how the system achieves it in a way that improves the functioning of a computer or another technology. The specification should name the technical problem and explain, in detail, the technical solution — that disclosure is what Step 2A, Prong Two rewards.
• Claim concrete steps and structure, not pure outcomes. Result-oriented claiming (“a system that determines the optimal price”) invites an abstract-idea rejection. Recite the specific operations, data transformations, and components that produce the result.
• Avoid generic, conventional language for the technical parts. “A processor configured to” plus an abstract idea reads as “apply it on a computer.” Tie the steps to non-conventional structure or an unconventional ordered combination of elements.
• Anchor the abstract part to something concrete. A physical transformation, a measurement, a sensor input, a specific hardware interaction, or a change in how a machine operates can pull a claim out of the abstract bucket.
• Build a layered claim set. Use independent claims for breadth and dependent claims to add the technical detail you can fall back on if the independent claim draws a § 101 rejection.
• Write the specification to support eligibility. Your abstract and detailed description should make the technical improvement obvious to a reader. You can usually only argue from what the spec already discloses, so put the improvement story in there from the start.

None of this is clever wording for its own sake. It’s about making sure the genuine technical contribution you actually made shows up clearly in the claim and the disclosure — because an examiner can only credit what’s on the page.

What to Do When You Receive a § 101 Rejection

A § 101 rejection is common, not fatal. The examiner has to make a specific case, and that gives you specific openings to push back.

1. Read the rejection precisely. Identify which exception the examiner says you recite (which abstract-idea grouping?) and exactly where they say the analysis fails — Prong Two or Step 2B. Your response is only as good as the part of the framework it targets.
2. Argue Prong Two first. If you can show the additional elements integrate the exception into a practical application — a real improvement to technology — you win at Step 2A and skip the inventive-concept fight entirely.
3. Attack the “conventional” finding. Under Berkheimer, whether an element is well-understood, routine, and conventional is a factual question. If the examiner just asserts it, you can demand evidence and point to your specification showing the combination isn’t conventional.
4. Amend toward the technical detail. Pull specific, eligibility-supporting limitations from your dependent claims or specification into the independent claim. This is where a layered claim set pays off.
5. Use the USPTO’s own examples. The guidance examples (including the 2024 AI set) are persuasive with examiners. Map your claim to an example the office already treats as eligible.
6. Request an interview. A short call with the examiner often resolves a § 101 rejection faster than another written round.

If amendments and arguments stall, you still have appeal options. But the cheapest fix is almost always the one you build in at drafting — which is why getting eligibility right before you file beats arguing about it for two years.

How PerspireIP Can Help With Patent Eligibility

Eligibility is won or lost in the claims and the specification, long before the office action arrives. PerspireIP’s patent professionals draft software, AI, and life-sciences applications with the Alice framework in mind — framing the technical improvement, building layered claim sets, and crafting disclosures that hold up under § 101. Already facing a rejection? We respond strategically, targeting the exact prong the examiner relied on. Contact us to protect your invention the right way.

This article is general information, not legal advice; consult a qualified attorney for your situation.

Frequently Asked Questions