How should patent applicants use Rule 132 declarations (SMEDs) to overcome Section 101 rejections, especially for AI inventions?

A Rule 132 declaration — increasingly called a Subject Matter Eligibility Declaration, or SMED — is sworn evidence submitted under 37 C.F.R. § 1.132 to rebut an examiner's factual assertion during prosecution. In 2025 and into 2026, the USPTO issued updated guidance clarifying how these declarations can be used to overcome subject matter eligibility rejections under 35 U.S.C. § 101, with particular attention to AI and software claims that examiners routinely characterize as abstract ideas or mere mathematical concepts. The definitive strategy is this: use a SMED not as a substitute for strong claim drafting, but as targeted rebuttal evidence aimed at the specific factual assertions in the examiner's eligibility rejection — typically the assertions that a claimed improvement is well-understood, routine, or conventional.

What a Rule 132 Declaration Actually Is

Also worth reading: How does the AI patent application review process work and what should applicants expect in 2026? · What are the AI patent eligibility 2027 USPTO guidelines and how should applicants prepare under the September 2026 framework? · What are the AI patent disclosure compliance requirements for 2026 and how do they affect patent applicants?

Rule 132 declarations have existed for decades as a general prosecution tool. Under 37 C.F.R. § 1.132, an applicant may submit affidavits or declarations to traverse an examiner's rejection where the rejection relies on factual contentions rather than pure legal conclusions. Historically, practitioners used them most often in obviousness disputes under § 103 — for example, to show that a claimed feature achieved unexpected results. The 2025 shift was the USPTO's explicit endorsement of using the same mechanism against § 101 rejections, which many practitioners had assumed were too legalistic for factual rebuttal.

The logic is straightforward once you see it. When an examiner rejects claims under § 101 after applying the Supreme Court's Alice/Mayo framework, Step 2A Prong Two asks whether additional elements amount to significantly more than a judicial exception — often by assessing whether they reflect an inventive concept. Examiners frequently support that conclusion with factual statements: that the claimed data-processing steps are 'well-understood, routine, conventional activities,' or that a generic computer performing them adds nothing. Those are factual assertions, and Rule 132 exists precisely to let applicants contest facts with competent, sworn evidence. The USPTO's guidance confirmed that examiners must consider such evidence and cannot simply dismiss it because it appears in a § 101 response rather than a § 103 response.

Why the USPTO Embraced SMEDs for AI Patents

AI applications face some of the highest § 101 rejection rates at the USPTO. Claims reciting neural networks, machine learning training pipelines, or AI-driven decision systems are frequently mapped to 'mental processes,' 'mathematical concepts,' or 'certain methods of organizing human activity.' The classic example cited across commentary on the guidance is AlphaGo-style reinforcement learning: a program taught only the game's rules that developed its own strategy through self-play. An examiner might view 'learning a strategy' as an abstract mental process; a SMED can establish that the specific technical implementation — the evaluation function architecture, the Monte Carlo tree search integration, the hardware-level parallelization — produces concrete technological improvements that no human mind performs mentally.

The USPTO's motivation is partly docket management. Eligibility rejections generate years of back-and-forth, RCE filings, and appeals. By channeling disputes into a structured evidentiary framework, the Office gives examiners a principled basis to withdraw rejections early when the applicant's evidence is credible, and gives applicants a path short of appeal. Early practice reports from mid-2025 — covered by Patently-O and several firm client alerts — suggested examiners were engaging with SMEDs more than skeptics predicted, though withdrawal rates varied sharply by technology center and examiner. The honest read three months in was 'tie goes to the runner': a well-drafted SMED helped, but it did not guarantee allowance.

The Core Strategy: Attack Factual Assertions, Not Legal Conclusions

The single most important strategic principle is scope discipline. A SMED succeeds when it rebuts specific factual predicates; it fails when it argues law. If the examiner wrote that 'sorting data arrays is a well-understood, routine activity performed by generic computers,' your declarant should address exactly that: what sorting algorithm, in what context, with what performance characteristics, and why the claimed approach differs from the routine baseline. If the examiner instead concluded that the claim 'as a whole' lacks an inventive concept, a declaration disputing that legal conclusion will be given little weight — and may irritate the examiner.

Effective SMEDs typically include testimony from a person of ordinary skill in the art (or a qualified expert) addressing: (1) the state of the art at the priority date; (2) whether the specific claimed improvement was known, recognized, or routine; (3) quantifiable technical results — speed improvements, accuracy gains, memory reduction, energy efficiency — tied to the claimed features; and (4) recognition in the field that the problem being solved was difficult or unsolved. Declarations anchored in measured results and contemporaneous literature carry far more weight than conclusory assertions that 'this was not conventional.'

Practical Steps to Build a Winning SMED

First, dissect the office action and build an assertion map. List every factual statement the examiner makes in the § 101 rejection, sentence by sentence. Each becomes a target for declarant testimony. Second, choose the right declarant. The inventor works when the invention history matters; an independent expert works better for state-of-the-art testimony because there is no self-interest taint. Third, gather objective corroboration before drafting: benchmark data, published papers predating the filing, conference talks, product documentation showing the problem persisted until the invention. Fourth, draft the declaration to mirror the examiner's language — if she said 'routine,' the declaration says why it was not routine, using her vocabulary. Fifth, submit it with an amended response that also addresses any legal arguments separately, so the declaration does the factual work while the brief does the legal work.

Timing matters. A SMED filed in a first office action response can short-circuit prosecution entirely. Filed only after two rejections and an RCE, it signals desperation and invites skepticism about why the evidence surfaced so late. Practitioners reporting on early SMED practice noted that examiners responded best when declarations arrived with substantive claim amendments or at least detailed argumentation connecting declared facts to claim limitations.

Comparing SMEDs to Alternative § 101 Strategies

The comparison reveals the real doctrine: these tools are sequential, not competing. Amend first if the claims genuinely drift toward abstraction. Deploy a SMED when the claims are sound but the examiner's factual premises are wrong. Appeal only when the examiner disregards competent evidence — which itself builds a stronger appellate record, because the board reviews whether the examiner properly considered rebuttal evidence.

Common Mistakes That Sink SMEDs

The most frequent error is the legal-argument declaration: an affidavit that spends four pages explaining why the claim satisfies the Alice test. Examiners and the board give that essentially zero evidentiary weight because it invades the examiner's legal territory. The second mistake is unsupported conclusory testimony — a declarant stating the invention 'was not known in the art' without citing prior art, literature, or data. Third, inventors sometimes overreach, testifying outside their competence (an ML researcher opining on semiconductor fabrication conventions). Fourth, applicants file SMEDs without amending obviously problematic claim language, asking the declaration to cure drafting defects evidence cannot fix. Fifth, timing failures: sitting on available evidence through multiple office actions undermines credibility. Finally, some applicants treat the SMED as a substitute for traversing the rejection at all; the declaration supplements, never replaces, a proper response with legal argument and proposed claim positioning.

Cost, Economics, and When to Act

Budget realistically. A qualified independent expert typically bills $300–$600 per hour, and a defensible declaration requires 20–60 hours of expert time plus attorney drafting, putting total cost between roughly $10,000 and $30,000 for a serious effort. Against that, compare the alternative: each RCE costs $1,360–$2,000 in government fees plus response costs, appeals run five figures minimum, and continued prosecution delay pushes revenue-generating issuance back quarters or years. For a high-value AI portfolio asset, a well-targeted SMED is usually cheaper than two additional prosecution cycles. For a marginal case with weak technical differentiation, the money is better spent narrowing claims or abandoning the application.

Act early. The optimal window is the first office action containing a § 101 rejection that rests on 'well-understood, routine, conventional' characterizations. Build the assertion map immediately, commission the state-of-the-art survey, and decide within weeks whether evidence exists to support a declaration. Waiting converts a proactive evidentiary strike into a reactive salvage operation.

The Honest Caveats

SMEDs are not a silver bullet, and the early enthusiasm in firm alerts deserves tempering. Reported examiner engagement varied considerably across technology centers, and some examiners treated declarations as argument to be weighed rather than evidence requiring rebuttal. The Federal Circuit has not yet produced a body of precedent squarely blessing SMED-driven § 101 allowances, so the practice rests on USPTO guidance and internal examination procedure rather than binding appellate authority. Applicants should also recognize that anything declared becomes part of the public record and can be used in later litigation or IPR proceedings — declarants must be accurate, consistent, and prepared to defend their testimony years later. Used with discipline, targeting real factual disputes with corroborated testimony, a Rule 132 SMED is currently the most direct evidentiary lever AI applicants have against boilerplate eligibility rejections. Used as a rhetorical dump, it wastes money and hardens examiner resistance.