Start with the Examiner’s Actual §101 Analysis
The first mistake in a subject-matter-eligibility response is to answer a rejection that the examiner did not actually make. Under the USPTO’s current framework, the analysis proceeds through the statutory-category inquiry and then the Alice/Mayo-based steps reflected in MPEP §§2103–2106.07. For many software and AI claims, the practical dispute is Step 2A: whether the claim recites a judicial exception and, if so, whether the claim as a whole integrates that exception into a practical application.
A useful response therefore begins by reconstructing the rejection limitation by limitation. What language did the examiner identify as the alleged abstract idea? Which remaining elements were treated as extra-solution activity, generic computer implementation, or insignificant data gathering? Did the Office Action address the claim as an ordered combination, or only list individual elements? Those questions matter because an eligibility response should attack the analytical link that is weak, not simply recite favorable cases.
The USPTO’s December 2025 eligibility update expressly separates eligibility from traditional patentability. An invention can be eligible and still fail novelty, obviousness, written description, enablement, or definiteness. A good response keeps those statutory questions distinct.
The Specification Must Supply a Credible Technological Improvement
After Enfish and the USPTO’s precedential Ex parte Desjardins decision, an asserted improvement to computer functionality or another technology remains one of the most important pathways through Step 2A, Prong Two. But the word “improvement” is not self-proving. The specification should explain what technical limitation existed in the prior approach, what mechanism the claimed invention changes, and what technical consequence follows from that change.
Compare two stories. “The invention processes information more efficiently” is a conclusion. “The system reorganizes data in a defined structure that eliminates repeated database traversal and reduces the number of memory accesses required for each query” describes a mechanism and an effect. The second story gives the claim drafter and prosecution team something concrete to connect to claim language.
The same discipline applies to AI-related inventions. Saying that a model produces a better result may still sound like the automation of analysis. Explaining how a particular data representation, model architecture, training control, hardware interaction, or inference pipeline changes computer operation gives the eligibility argument a technical center of gravity.
The Improvement Has to Be in the Claim—not Only in the Background Story
A specification can describe an impressive technical advance while the pending claim remains drafted at the level of “receive data, analyze data, output a result.” That disconnect is a common reason a persuasive invention story fails to move a §101 rejection. Step 2A asks about the claim as a whole. If the limitations that create the technological improvement are absent, the response may need an amendment rather than another page of argument.
The amendment should not simply add technical nouns. The better question is which supported limitations cause the asserted improvement. If the technical benefit depends on a particular ordering of operations, data structure, control relationship, sensor interaction, or resource allocation, the claim should capture enough of that mechanism to make the improvement legally visible.
That creates a second constraint: new matter. An applicant cannot repair a thin disclosure by inventing a mechanism after filing. For that reason, §101 strategy often exposes a drafting problem that began before prosecution. The more the original specification explains the “how” of the improvement, the more room the applicant has to amend without sacrificing the invention to unsupported detail.
Argue the Claim as an Ordered Combination
Eligibility analysis should not end with a checklist saying that each limitation is individually known or generic. The USPTO’s current guidance emphasizes considering the claimed invention as a whole. A combination of conventional components can still implement a nonconventional technological arrangement; conversely, a long claim can still be abstract if its added elements merely surround the exception with routine activity.
A practical response can therefore identify the interaction among limitations: limitation A generates a particular intermediate representation; limitation B uses that representation to control C; and C changes the operation of a device, network, database, or computing process. This is more useful than arguing that each element is “technical” in isolation.
When the examiner has characterized the claim too broadly, the response should also show why that characterization abstracts away the very limitations that define the improvement. The point is not to win a semantic fight over the label assigned to the abstract idea. The point is to demonstrate that the claim contains a specific application that the characterization fails to account for.
Evidence Can Help When the Technical Proposition Is Disputed
Eligibility is a legal question, but some disputes turn on technical facts: whether a claimed architecture improves computer functionality, whether a limitation changes resource use, or whether a result reflects an engineering improvement rather than an abstract business objective. The USPTO now provides best-practice guidance for voluntary Rule 132 Subject Matter Eligibility Declarations (SMEDs). The April 30, 2026 memorandum superseded the initial December 2025 best-practices memo.
A declaration is not a routine add-on. It is most useful where a competent declarant can explain a specific technical proposition grounded in the application and evidence. It should not merely restate attorney argument. Before using one, the team should identify the factual issue the declaration is intended to prove, make sure the proposition is consistent with the original disclosure, and consider how the statement may affect later claim interpretation or litigation.
A Practical Response Workflow
- Parse the examiner’s Step 2A and Step 2B reasoning claim by claim.
- Identify the specification passages that describe the technical problem, mechanism, and effect.
- Map those passages to the actual claim limitations; do not rely on unclaimed advantages.
- Decide whether the best path is argument, supported amendment, evidence, or a combination.
- Address the ordered combination rather than only arguing each limitation separately.
- Keep §101 arguments distinct from novelty and obviousness; do not imply that unconventionality alone proves eligibility.
- Check every proposed amendment for written-description support and new-matter risk.
- Where a technical factual dispute is material, evaluate whether a Rule 132 SMED would add real evidentiary value.
Takeaway
The most durable §101 response is not the one with the most case citations. It is the one in which the claim, specification, and legal framework converge on the same technical proposition. The applicant should be able to say what the technology improves, how it achieves that improvement, and where those features appear in the claim.
That is also why eligibility work often starts before the Office Action arrives. A specification that records technical mechanisms and effects gives prosecution counsel options. A specification that describes only desired results leaves fewer options, no matter how sophisticated the later legal brief becomes.