Up: AI Office Action response software

Claim amendments with support, claim history, and attorney approval

Amending a claim is the most consequential thing a response does. It narrows what the client will own, it can surrender scope that cannot be recaptured, and it must be supported by the application as filed and presented in the form the rules prescribe. This guide covers the decision to amend rather than traverse, where the support has to come from, why the claim history matters before any narrowing, how the claim listing under 37 CFR 1.121 is built, and where the attorney’s approval sits in PatentReply’s workflow.

Scope: U.S. utility applications before the USPTO. Statutory and regulatory statements are cited to the primary source; the product statements describe PatentReply as it ships. Nothing here is legal advice, and the decision to amend is the practitioner’s.

When to amend versus traverse

A traversal argues that the rejection is wrong on the record as it stands: the cited passage does not teach the element in the claimed sense, the stated reason to combine the references does not hold, the examiner has construed a term more broadly than the specification allows, or a procedural requirement was not met. An amendment concedes nothing about the rejection but changes the claim so the rejection no longer applies. The choice turns on what the analysis found. If the examiner’s passage really does disclose the element, arguing that it does not is a losing position and an invitation to a final rejection; if it does not, narrowing the claim gives away scope for nothing.

In practice many responses do both: amend the independent claim to add an element the art of record does not show, and traverse the grounds against the dependent claims that already recited it. A reply under 37 CFR 1.111(b) must in any case “reply to every ground of objection and rejection” and point out the supposed errors in the examiner’s action; an amendment without Remarks that distinctly and specifically address each ground is not a complete reply.

Finding written-description support before narrowing

Every word added to a claim has to be supported by the application as filed. Two rules enforce this. 35 U.S.C. § 132(a) forbids an amendment that introduces new matter into the disclosure, and § 112(a) requires the specification to describe the claimed invention; an amended claim the original disclosure does not reasonably convey fails the written description requirement even if the words themselves are not new. The practical test is whether the specification, the drawings, or the original claims actually disclose the limitation in the form it is being claimed, not merely a related idea.

The support should be found before the claim is drafted, and cited in the Remarks by paragraph or figure. Because a compiled matter holds the application as filed with every paragraph addressable, your AI can search the specification for candidate support and return the passages with their addresses; the attorney reads them and decides whether they carry the amendment. A passage that merely uses the same words is not support for a limitation the inventor never described.

Why claim history matters: prior amendments, arguments, and estoppel

A claim does not arrive at a response with a blank history. It may already have been narrowed to overcome an earlier rejection; the applicant may already have told the examiner what a term means, or argued that a reference lacks a feature. Those statements bind the next response twice over. Within prosecution, a new argument that contradicts an old one invites the examiner to quote the file back. After grant, prosecution history estoppel can bar the patentee from recapturing, through the doctrine of equivalents, scope that a narrowing amendment surrendered.

This is why PatentReply reconstructs every claim set ever presented when it compiles a matter, and why its analysis reports the history constraints on each issue: what was argued before against the same reference, and which narrowing amendments bear on scope. Each amendment proposal is pinned to the claim version it amends and carries a scope-cost note and an estoppel note, so the decision to narrow is made with the history in view rather than discovered later.

One special case deserves its own mention. A nonstatutory double patenting rejection is often answered not with an amendment but with a terminal disclaimer, which caps the term of the later patent to the earlier one. That is a term decision as much as a claim decision; the guide on how terminal disclaimers affect patent expiration covers what it costs.

The amendment listing under 37 CFR 1.121

Amendments to the claims are made under 37 CFR 1.121(c), which requires a complete listing of all claims ever presented, in ascending order, with a status identifier in parentheses after each claim number: Original, Currently amended, Previously presented, Canceled, Withdrawn, New, Not entered, or Withdrawn – currently amended. A currently amended claim shows added text underlined and deleted text with strikethrough, or double brackets for five or fewer characters; canceled and withdrawn claims are listed by number and status only. The listing replaces every earlier version of the claims, so a claim omitted from it is treated as canceled, and an amendment that does not comply is held non-compliant and the period for reply keeps running. MPEP § 714 walks through the format with examples.

In PatentReply the listing is a server artifact. Once the plan is approved, it is generated deterministically from the proposed claim text, either from the durable amendment proposals on the approved revision or from full text supplied per amended claim plus any cancellations. Status identifiers, markings, and ordering follow the rule, and the markup is computed by comparing the proposed text with the operative claim it amends; the model never assembles the listing from claim text it read.

The attorney approval gate before any amendment is prepared

Proposing an amendment and preparing it as a filing paper are two different acts in PatentReply, separated by the practitioner’s approval. During strategy, the attorney and the AI record amendment proposals on the plan: proposed full text per claim, pinned to the version it amends, with the scope-cost and estoppel notes. When the plan is settled, it is confirmed and the attorney is handed a link to an approval page on patentreply.ai. Approving there, signed in, is the sign-off; a yes in the chat is not, and the model cannot approve on the attorney’s behalf. The approval is recorded with the matter and the plan revision it approved.

Only after that approval does PatentReply generate the 1.121 listing and assemble the response document, with the Remarks your AI drafted from the approved strategy. If the plan changes afterward, a new revision needs a new approval before drafting resumes. The Office Action response workflow guide explains why the gate sits exactly there; the step-by-step instructions are in the documentation under Amendments and the response document.

Up: AI Office Action response software