Section 101 Declarations: The USPTO's 2026 SMED Strategy for AI and Software Patents

On April 30, 2026, USPTO Director John A. Squires issued an updated memorandum on Subject Matter Eligibility Declarations (SMEDs). The memo supersedes the December 4, 2025 best-practices guidance and refines how applicants should present §101 evidence to examiners.

SMEDs are voluntary. They are not a new procedural requirement, and they are not specific to AI or software patents. They are an existing tool under 37 C.F.R. § 1.132 ("Rule 132") that the USPTO has clarified, formatted, and is now actively encouraging for inventions where eligibility is the bottleneck. The April 2026 memo says the quiet part out loud: the technologies the office expects SMEDs to support are software, AI, data analytics, and medical diagnostics.

This page covers what SMEDs are, what changed in the April 30 memo, how to structure a strong SMED for an AI-assisted or software invention, and where the evidence comes from. If you are an attorney preparing a §101 response, this is the workflow the USPTO is signaling it wants to see.

This article presents published USPTO guidance and law-firm analysis for educational purposes. It is not legal advice. Consult a registered patent attorney for filing decisions.

What a SMED actually is

A Subject Matter Eligibility Declaration is a sworn statement filed under 37 C.F.R. § 1.132 (Rule 132) supporting an applicant's §101 eligibility argument with factual evidence. It is the same procedural vehicle attorneys have used for years to support §§102/103/112 arguments with declarations from inventors, experts, or third parties. The SMED is the §101-specific application of that vehicle.

The declaration is sworn, signed by a competent declarant, and made of record. Examiners are obligated to consider it under In re Oetiker, which places the initial burden on the office to establish a §101 case. A well-supported SMED shifts the evidentiary balance: the examiner now has to engage with specific factual claims about how the invention works and why it represents a technical improvement.

The mechanism is not new. The April 30, 2026 memo's contribution is to standardize the format and substance, so examiners encounter consistent declarations across applications and applicants encounter consistent treatment across art units.

What the April 30, 2026 memo changes

1. Standalone declarations, separated from other §§ declarations

The single most actionable change. The USPTO now encourages applicants to file a SMED focused solely on §101, distinct from any Rule 132 declarations supporting novelty (§102), obviousness (§103), or written description (§112). Combining all of an application's evidentiary arguments into one declaration was permissible under the December 2025 memo. It is still permissible. The April 2026 memo discourages it because it conflates issues and gives the examiner cover to engage with only the easiest argument.

A standalone SMED puts §101 in front of the examiner as a discrete decision. It also creates a cleaner record for any future district-court or PTAB review.

2. Substantive standard: "better, faster, cheaper, or more efficient"

The memo articulates a substantive standard for what a SMED should establish: the claimed invention is better, faster, cheaper, or more efficient than existing solutions, with a clear evidentiary nexus to the claims. This is the language the USPTO wants examiners and applicants to converge on.

It is not the legal test for eligibility (Alice/Mayo two-step is). It is a presentation standard for the evidence package an applicant submits in support of step-two ("inventive concept") arguments. The implication is concrete: a SMED that asserts technical improvement without quantifying it against existing solutions is weak; a SMED that benchmarks the invention against named prior approaches with measurable metrics is strong.

3. The "living document" framing

The April 2026 memo explicitly positions itself as a living document. The USPTO will issue periodic updates as practitioner experience develops. MPEP § 716.01(c)(III) (general guidance on declarations) is under review for potential revision, which would codify the SMED approach more formally.

Practical takeaway: the SMED format is in active development. Practitioners filing SMEDs in mid-2026 should expect format guidance to evolve, and should write to the underlying substance (technical improvement, evidentiary nexus, named comparators) rather than to today's specific formatting preferences.

4. The USPTO is reporting early success without numbers

The memo's preamble notes that "early feedback indicates that voluntary submitters are having success" with the SMED approach. No specific success-rate data is published. Read this as encouragement, not as a measured outcome study. The office is signaling: practitioners who are filing SMEDs are seeing engagement; the office wants to see more of them.

What a strong SMED looks like for AI and software inventions

The April 30 memo identifies software, AI, data analytics/computation, and medical diagnostics as the technologies "particularly applicable" for SMED use. That is not because these technologies are subject to different §101 law; the Alice/Mayo two-step applies to all. It is because these technologies disproportionately encounter §101 rejections at the office and need supplemental factual support to clear step two.

Five attributes of a strong SMED for an AI or software invention:

1. Named comparators, not generic prior approaches. "The invention performs better than prior systems" is weak. "The invention performs better than the approach disclosed in U.S. Patent No. 9,XXX,XXX" or "the conventional approach described in Smith et al., 2023" is strong. Specificity creates evidentiary nexus the examiner has to engage with.
2. Measurable metrics. Throughput, latency, accuracy, energy consumption, memory footprint, training cost, inference cost. The SMED that quantifies improvement is the SMED that survives. The qualitative claim is the SMED that gets rejected.
3. Evidence chain to the claims. Each technical improvement asserted in the declaration should map to specific claim limitations. The examiner is reading the SMED with the claim language open in another tab; the SMED writer should anticipate that.
4. Declarant qualifications matched to the technical claim. An inventor declarant who built the system can attest to its measured behavior. A subject-matter expert declarant who did not build the system can attest to its novelty against the broader prior art. The strongest SMEDs use both, in separate sections or separate declarations.
5. Source citations. Benchmarks, prior-art references, public benchmarks, and (where applicable) the source code that was tested. A SMED that points to a reproducible measurement is harder to dismiss than a SMED that does not.

A SMED template: what to populate, where the evidence lives

The April 2026 memo's substantive standard ("better, faster, cheaper, or more efficient than existing solutions" with a clear nexus to claims) maps to a small number of structured sections. Practitioners drafting SMEDs for AI and software inventions can use the following as a working template:

Template structure derived from the April 30, 2026 USPTO memorandum's emphasis on claim-linked evidence with named comparators and objective metrics. The memo does not prescribe this exact table format; it prescribes the substance. Format is at the practitioner's discretion.

Where the evidence comes from

This is where the SMED workflow intersects modern disclosure tooling. The April 2026 memo asks for evidence anchored to specific claim limitations, with measurable metrics and named comparators. That is exactly the output structure of a well-designed code-scanning disclosure pipeline:

• Distinctive algorithms, architectures, and methods identified by automated scanning of an actual codebase, with evidence citations to specific files and commits.
• Strategic concepts with scoring against named prior approaches and measurable distinctions.
• An evidence trail that an examiner (or, downstream, a PTAB panel or district court) can follow back to the technical artifact.

ObviouslyNot's scanner produces output structured for exactly this kind of evidentiary chain. The scanner identifies; the patent attorney determines whether what was found supports a §101 declaration. The attorney's role is unchanged. What changes is that the attorney now has a structured technical baseline to draft the SMED against, rather than building the evidence package from scratch each time. See the attorney landing page for how this fits prosecution practice, or scan your codebase directly.

What this is not

The April 2026 memo is widely being discussed as "the SMED guidance." Some of the discussion has overstated what the memo actually does. To be clear about what it does not do:

• SMEDs are not mandatory. They remain voluntary. An applicant who does not file one is not procedurally disadvantaged. The applicant who does file one well is evidentially advantaged.
• The memo does not change Alice/Mayo. The legal test for §101 eligibility is unchanged. The memo is about how to present evidence in support of step-two arguments under that test.
• SMEDs do not guarantee allowance. A strong SMED creates a defensible record and forces the examiner to engage with specific factual claims. It does not replace the underlying eligibility analysis.
• The memo does not exempt AI or software from §101 scrutiny. Those technologies remain disproportionately rejected on §101 grounds. The SMED is a tool for navigating that, not a workaround.

How this fits the broader USPTO 2026 picture

The SMED memo is part of a coordinated USPTO posture under Director Squires. Three other 2026 developments worth tracking alongside:

• Discretionary denial framework for PTAB IPR institution (Magnolia Medical Technologies v. Kurin, May 14, 2026, precedential).
• Ex parte reexamination pre-order procedure (effective for requests after April 5, 2026), allowing patent owners to submit up to 30 pages of supplemental information before SNQ determinations.
• Updated AI inventorship guidance (November 2025) treating AI strictly as a tool under human conception standards, consistent with Thaler v. Vidal.

Read together, the office is signaling that quality patents with documented evidentiary support are favored, while procedurally weak filings face increasing scrutiny. SMEDs are the §101-specific manifestation of that posture.

See also our Software Patents After Alice guide for the broader §101 landscape and AI Patent Legal Precedents for the case-law backdrop.

Sources

Related from the resources