A systematic approach to error-free filings in a year of new fees, new guidance and new scrutiny

1. Why Proofreading Is a Strategic Function

A patent application is a legal instrument written in technical language and its text is the only thing that defines what is disclosed and what is claimed. A mistaken reference numeral, a claim that depends on a non-existent element, a transposed digit in a priority claim or a misplaced decimal in a numerical range can have consequences that last for the life of the patent. Some errors can be corrected cheaply; others cannot be corrected at all because correction would add new matter, or because a priority or filing-date deadline has already passed.

The environment in which proofreading takes place has also shifted. During 2025 the USPTO implemented a new fee schedule that puts a price on length, continuation practice and information-disclosure volume; it reissued guidance on AI-assisted inventions and on practitioners’ use of AI tools; it elevated Ex parte Desjardins to precedential status, signalling that examiners will lean harder on written-description and enablement under 35 U.S.C. § 112 in software and AI cases; and it changed how inter partes review petitions are handled, making the quality and defensibility of issued claims more important than ever. Proofreading therefore is no longer only about typographical neatness – it is a front-line quality-control and cost-control activity.

This article lays out the 2025 changes that matter most to proofreaders, a layered proofreading methodology for each document type and practical checklists. It reflects developments through late 2025; because rules and guidance continue to move, confirm current requirements before relying on any detail.

2. What Changed at the USPTO in 2025 (and Why It Affects Proofreading)

2.1 Fee changes effective 19 January 2025

The USPTO’s final fee rule took effect on 19 January 2025. Filing, search and examination fees for utility and design applications rose by roughly 6–10%, requests for continued examination became more expensive (the first RCE by about 10% and second and subsequent RCEs by about 43%) and several new fee categories were introduced. Three of them directly affect proofreading behaviour:

• Continuing-application fees. A continuing application filed more than six years after the earliest benefit date to which it claims priority attracts an additional fee, which rises again if the benefit claim reaches back more than nine years. A mistaken or unnecessary benefit claim – or a misdated one – can therefore cost thousands of dollars. Verify every priority and benefit claim against the filing receipt and the application data sheet (ADS).
• IDS size fees. A tiered fee now applies to information disclosure statements that cite large numbers of references, with additional charges for lengthy non-patent literature. Proofread IDSs for duplicate citations, cumulative references and mis-keyed numbers and plan whether to file a single large IDS or to rationalise the list to what is truly material.
• Claim and length-related fees. Excess independent and total claims and excess pages for large specifications, remain priced; a careful claim-set proofreading pass should also confirm that claim counts and dependencies match the fee calculation.

Entities qualifying for small or micro status benefit from reduced fees, but a wrong entity-status assertion can lead to serious problems. Verify entity status at each fee payment.

2.2 DOCX filing

The USPTO has promoted filing the specification, claims and abstract in DOCX format and applies a surcharge to non-DOCX filings (introduced in 2023 and adjusted in subsequent fee rules). Authors should use the DOCX that Patent Center can parse reliably: avoid text boxes for claim text, unusual embedded objects, complex tracked changes and non-standard fonts. A proofreading pass on the final DOCX – rather than on a PDF export – catches formatting artefacts such as hidden text, residual comments, broken numbering and auto-numbered paragraph references that renumber unexpectedly. Always strip metadata and comments before filing.

2.3 AI-related guidance

In February 2024 the USPTO issued guidance reminding practitioners that existing duties – candour, the certification made by every signature under 37 CFR 11.18, confidentiality and competence – apply fully when AI tools are used to draft or review submissions. In November 2025 the Office replaced its earlier inventorship guidance for AI-assisted inventions with a revised framework that applies the same conception-based inventorship test used for all inventions and reaffirms that only natural persons can be inventors. For proofreaders the lessons are practical:

• Treat any AI-generated text as an unverified draft. A practitioner remains responsible for every statement, citation and technical assertion filed.
• Verify that every case, rule and reference an AI tool suggests actually exists and says what the draft claims.
• Do not upload confidential invention disclosures to tools that do not meet the client’s confidentiality requirements.
• Check that inventorship statements and declarations reflect human conception and document the human contribution where AI was used in the inventive process.

2.4 Examination emphasis: Section 101 and Section 112

The Appeals Review Panel decision in Ex parte Desjardins, designated precedential in November 2025, held claims directed to improving how a machine-learning model operates to be patent-eligible and was later reflected in updated examination guidance. Commentators have observed that examiners are turning more attention to written-description and enablement rejections under Section 112 in AI-related cases. For proofreaders, this means technical clarity matters: the specification should explain how the claimed improvement is achieved, define terms consistently and ensure each claim element has an identifiable source of support in the description.

2.5 Post-grant dynamics

The USPTO also significantly changed how it institutes IPR proceedings during 2025, including rescission of prior discretionary-denial guidance and centralisation of institution decisions with the Director. Whatever one thinks of the policy, issued claims that are clear, internally consistent and well supported are easier to defend in any forum. Proofreading at the prosecution stage is cheaper than defending ambiguity later.

Always verify against current sources Fee amounts, thresholds and guidance cited in this article reflect the situation reported in 2025. Before filing, check the USPTO’s current fee schedule, Patent Center notices, the MPEP and any Federal Register notices issued since.

3. A Layered Proofreading Methodology

Single-pass reading by the person who drafted a document is the least reliable form of proofreading: the author reads what they meant to write. A robust process uses several independent layers, each with a distinct purpose.

Layer	Purpose	Typical tools / method
1. Author self-check	Catch obvious errors and fix structure before others see it	Spell-check, claim-element checklist, read aloud
2. Automated consistency check	Find antecedent-basis problems, undefined terms, numeral mismatches, claim-dependency errors	Claim-analysis software, macros, Word’s Compare and Navigation Pane
3. Independent human review	Fresh eyes on language, logic, support and style	A second professional who did not draft the text
4. Cross-document reconciliation	Align specification, claims, drawings, abstract, ADS, IDS and oath/declaration	Cross-check tables; reference-numeral table
5. Filing-format review	Ensure the file complies with USPTO format and parses cleanly	Patent Center validation, DOCX inspection, PDF preview
6. Post-filing verification	Confirm the filing receipt, filing date, priority claims and fees	Compare filing receipt to ADS and instructions

4. Document-by-Document Checklists

4.1 Claims

Claims deserve the most rigorous review because they define the boundaries of the patent and because the doctrine of claim construction depends on their exact words. Read every claim at least twice – once for language and once for logic – and then read the entire set backwards, from the last dependent claim to claim 1, to detect broken dependencies.

• Antecedent basis. Every “the” or “said” element must have a prior introduction using “a” or “an,” in the same claim or in a claim from which it depends.
• Dependency. Each dependent claim refers to an earlier claim (never to a later one) and multiple dependent claims comply with the rules and fee consequences.
• Numbering and format. Claims are numbered consecutively, begin with a capital letter, end with a period and each claim is a single sentence.
• Terms of degree and indefiniteness. Phrases such as “substantially,” “about,” “relatively” and “optimal” should be defensible in light of the specification; the indefiniteness standard of Nautilus v. Biosig requires reasonable certainty for a skilled person.
• Consistency of terms. Use one term for one concept. Do not switch between “controller,” “processor” and “control unit” unless they are meant to be different.
• Transitional phrases. “Comprising,” “consisting of” and “consisting essentially of” are used intentionally.
• Means-plus-function language. Where claim elements use “means,” “module” or “unit” in a way that may invoke Section 112(f), confirm that corresponding structure, material or algorithm is disclosed.
• Numerical ranges and units. Check endpoints, inclusive/exclusive language and units against the examples and tables.
• Claim count. Confirm the number of independent and total claims against budget and fee calculations.

4.2 Specification

• Support for each claim element. Create a claim-support chart linking each limitation to paragraph numbers.
• Terminology. Define special terms and make sure the description uses them consistently.
• Reference numerals. Confirm every numeral in the text appears in the drawings and vice versa; confirm each numeral refers to the same element throughout. Check for duplicate numerals assigned to different elements.
• Figure descriptions. Confirm the brief description of each figure matches the figure content and numbering (FIG. 3A, 3B and so on).
• Cross-references. Check references to related applications, paragraphs, tables and incorporated documents.
• Numbers, units and chemical data. Verify mathematical values, formulas, molecular weights, temperatures, concentrations and units against source data. Proofread chemical names and structure labels character by character.
• Sequence references. Ensure every sequence in the text has the correct SEQ ID NO: and that the XML listing conforms to the current standard.
• Boilerplate. Remove leftover template text, client names, internal notes and bracketed instructions.
• Section 112 sufficiency. Look for statements that overpromise (“always,” “never,” “the invention is”) and rewrite them as embodiments, unless the intent is to limit the invention.

4.3 Abstract and title

The abstract must be a concise summary of the disclosure, not exceeding 150 words under 37 CFR 1.72(b) and the title should be short and specific (37 CFR 1.72(a) limits it to 500 characters and shorter is better). Check that neither contains implied limitations that conflict with the claims, that abbreviations are expanded and that no numeral or legal phrase was left in by mistake.

4.4 Drawings

Check drawings for compliance with 37 CFR 1.84 (margins, black lines, character height, legends, sheet numbers) and for consistency with the specification. Compare numerals in each figure with the reference-numeral table, check that all claimed features are shown (37 CFR 1.83) and view the final PDF at reduced size to confirm legibility.

4.5 Application Data Sheet and administrative documents

The ADS is a frequent source of costly errors because the Office treats it as controlling for bibliographic information. Carefully verify:

• Inventor names, residences and mailing addresses, spelled as they appear in legal documents.
• Applicant and assignee names and entity types.
• Priority and benefit claims, including application numbers, filing dates, relationship types (continuation, divisional, continuation-in-part, provisional) and the timing of claims.
• Foreign filing licence information and any prior publication.
• Entity status assertions (small or micro) and supporting certifications.
• Correspondence address, customer number and attorney registration numbers.
• Patent Law Treaty-related requirements for filing-date components and for restoration of priority.

4.6 Information disclosure statements

• Verify each U.S. patent and publication number, kind code, date, first-named inventor and country.
• Check non-patent literature citations for author, title, journal, date and pages.
• Remove duplicates and, where appropriate, cumulative references; consider the new size-based fees.
• Ensure every document cited is available and legible and that foreign-language documents are accompanied by the required concise explanation of relevance or a translation where available.
• Cross-check the IDS against references cited in related applications and against the search report in any foreign counterpart.

4.7 Responses and amendments

Amendments are governed by the formatting rules of 37 CFR 1.121 (claim listing with status identifiers, underlining for additions, strikethrough or double brackets for deletions). Compare the amended claims with the previous version using Word’s Compare, ensure no unintended changes slipped in and verify that every cited paragraph or figure for support actually supports the amendment. Check deadlines and extension fees and review remarks for statements that could create prosecution history estoppel.

5. Technique: How to Proofread Effectively

5.1 Change the medium and the pace

• Print the document or change the font and size on screen; unfamiliar presentation helps the eye see what is actually there.
• Read aloud or use a text-to-speech tool; the ear catches missing words and clumsy sentences.
• Read slowly, with a ruler or a cursor guide, one line at a time.
• Proofread numbers separately from words; verify each against its source, not against memory.
• Take breaks. Fatigue is a primary cause of missed errors; proofread in sessions of manageable length.

5.2 Use independent verification

Wherever possible, have someone other than the author review the text. The “two-person rule” is especially valuable for the claims, the ADS and the priority claims. For critical documents consider a blind comparison: have two reviewers separately build the reference-numeral table and compare their results.

5.3 Build and use checklists

Checklists are not a sign of weakness; they are the standard tool of high-reliability professions. Maintain a living checklist for each document type and update it whenever an error escapes to the Office. Record the types of errors found in each review so that recurring weaknesses can be addressed by training or template changes.

5.4 Use technology wisely

• Word processors. Use styles for headings, paragraph numbering and claims; use the Navigation Pane and Compare feature; turn on non-printing characters to find stray spaces and manual line breaks.
• Claim and specification analysers. Several commercial tools check antecedent basis, claim terms, reference numerals and figure references. They are useful but not infallible; treat their output as a prompt for human review.
• Macros and scripts. Simple macros can find double spaces, inconsistent hyphenation, missing units, straight versus curly quotation marks and mismatched parentheses.
• AI assistants. If you use AI to propose edits or to flag inconsistencies, treat the output as a suggestion; verify each change; ensure compliance with confidentiality obligations; and keep a record of what was changed and why.
• Document comparison. Always compare the filed version to the approved version and the amended claims to the pending claims.

6. Technical Fields: Special Proofreading Points

Field	Points to verify
Software / AI	Consistent naming of modules and data structures; explained algorithms and training data; flow-chart steps matching method claims; support for computer-readable-medium and system claims; clear description of the technical improvement
Biotech / pharma	Sequence identifiers and ST.26 compliance; dosage units and ranges; tables of data matching text; Markush groups and definitions; deposit information; terms such as “comprising” versus “consisting of”
Chemistry	Chemical names and structure drawings; formulas; reaction conditions; percentages that sum to 100; salts and solvates; consistent isomer notation
Mechanical / electrical	Part numbers and directions (upper/lower, proximal/distal); correct figure references; circuit element labels; dimensions and tolerances
Design	Figure descriptions; broken-line statements; consistent view labelling; no disclaimers inconsistent with drawings

7. Managing Deadlines and Post-Filing Verification

Proofreading continues after filing. Within days of filing, compare the filing receipt and the electronic acknowledgement with the instructions: Is the filing date correct? Are all documents received? Are priority claims properly recorded? Are the inventors, applicants and entity status right? Errors found early can often be corrected by a corrected filing receipt request, a corrected ADS or a petition; errors found late may require more costly remedies.

• Priority and benefit deadlines. Check the 12-month deadline for non-provisional and PCT applications claiming a provisional and the timing windows for adding or correcting benefit claims.
• Publication. Review the pre-grant publication for errors in the text as published and request corrections where appropriate.
• Allowance and issue. Proofread the Notice of Allowance, the examiner’s amendment and the drawings and text as they will issue. After grant, check the printed patent. Errors made by the Office may be corrected by a certificate of correction at no fee; errors originating with the applicant may require a fee and have narrower availability, so identify them quickly.
• Foreign filings. Translation errors are a particular risk; compare translations to originals and ensure that terms are translated consistently in each language.

8. Building a Proofreading Culture

• Assign responsibility. Define who is responsible for each layer and who signs off before filing.
• Allow adequate time. Rushed proofreading is the leading cause of missed errors. Build proofreading time into every deadline and push back on unrealistic timelines.
• Train and calibrate. Run periodic exercises with documents containing seeded errors; share lessons from corrections and rejections.
• Measure. Track error rates, types and the cost of corrections to show the value of improved processes.
• Keep records. Maintain final approved versions, comparison reports and review logs. These are invaluable when questions arise later.

9. A One-Page Pre-Filing Checklist

1. Claims: antecedent basis, dependency, numbering, terms, ranges and claim count verified.
2. Specification: claim-support chart complete; terms defined; no boilerplate remnants.
3. Reference numerals: table matches text and drawings; no duplicates or orphans.
4. Drawings: compliance with 37 CFR 1.84 confirmed; legible when reduced.
5. Abstract and title: length and content verified.
6. Sequences: SEQ ID NOs consistent; XML listing validated under the current ST.26 version.
7. ADS: inventors, applicants, priority and benefit claims, entity status and correspondence data verified.
8. IDS: citations accurate, duplicates removed, size fees considered.
9. File format: DOCX inspected for hidden text, comments, metadata and fonts; PDF drawings checked.
10. Fees: correct entity status, claims, continuation and surcharge fees calculated.
11. AI use: any AI-generated content verified; confidentiality obligations met.
12. Independent reviewer has signed off; final comparison against the approved version completed.
13. Post-filing: filing receipt reconciled with instructions within days of filing.

10. Conclusion

The USPTO’s 2025 changes raised the cost of mistakes and shifted attention to the clarity and support of the application. In that environment, proofreading deserves to be treated as a disciplined process rather than a final glance. A layered approach – author self-check, automated consistency checks, independent human review, cross-document reconciliation, format validation and post-filing verification – combined with technical-field checklists, careful use of AI and clear accountability, will reduce errors, protect disclosure and control costs. Practitioners who build these habits now will be well placed to adapt to whatever changes the USPTO introduces next.