Note · Aug 2026
Can a patent examiner ask how an invention was developed? Understanding 37 CFR § 1.105 in the age of AI
A requirement for information is a broad examination tool, but it is not an AI-detection rule. The meaningful question is who conceived the claimed subject matter.
Artificial intelligence is increasingly being used in patent practice. Inventors may use AI to organize technical notes, explore alternatives, summarize research, generate diagrams, or help turn an invention disclosure into the formal language of a patent application.
The answer is potentially yes, through a relatively broad examination tool found in 37 CFR § 1.105. But that does not mean applicants have a general obligation to provide the USPTO with a history of every tool used to prepare their application. The distinction is important.
What is a § 1.105 requirement for information?
37 CFR § 1.105 permits an examiner or other USPTO employee, during examination, to require information that is reasonably necessary to properly examine or treat the matter. The rule provides several examples of information that may be requested, including whether a prior-art search was performed and what was searched; information relating to the claimed invention; literature or patents used to draft the application; literature or patents used during the invention process; identification of what is being improved when an improvement is claimed; and technical information known to the applicant concerning the claimed subject matter or related art.
Two provisions are particularly interesting in the context of AI-assisted inventions. § 1.105(a)(1)(iv) concerns certain literature, published applications, and patents used to draft the application. § 1.105(a)(1)(v) concerns such materials used in the invention process, including materials used to design around an existing technology or develop a solution. These provisions existed long before modern generative AI. Nevertheless, § 1.105 provides a mechanism through which an examiner can seek additional factual information when there is a legitimate examination issue.
The examiner cannot simply demand everything
Section 1.105 is broad, but it is not unlimited. The MPEP explains that there must be a reasonable basis for believing the requested information would aid examination. Because responding can impose a substantial burden on the applicant, the examiner should clearly identify why the information is needed and narrowly define the expected response. In other words, § 1.105 is not supposed to function as an unrestricted discovery process.
An examiner generally should not ask:
Tell me everything the inventor did while developing this invention.
Instead, the request should be directed toward information reasonably necessary to resolve an examination issue. For example, an examiner might have evidence suggesting that a particular technical solution originated somewhere other than with the named inventor. That could potentially raise an inventorship question and provide a reason for requesting additional factual information.
Figure 1 · How a § 1.105 requirement is supposed to narrow
01 · Trigger
An examination issue arises
prior art, inventorship, or an unexplained origin of a claimed feature
02 · Basis
Reasonable basis to believe information would help
not curiosity, not a fishing expedition
03 · Scope
Narrowly defined request
identify why it is needed and what response is expected
04 · Response
Answer the requirement actually made
factual, accurate, consistent with other disclosure duties
Broad authority, focused use
Where AI makes this more interesting
Suppose an inventor independently conceives a new device. The inventor develops the architecture, determines how the components interact, performs experiments, and prepares extensive technical notes. Later, the inventor or patent practitioner uses a generative AI system to help convert those materials into patent-style prose. Those are two very different activities.
The USPTO's revised November 2025 inventorship guidance for AI-assisted inventions emphasizes that the existing human inventorship standard, rooted in conception, remains the controlling framework. The guidance does not create a new, routine obligation for examiners to investigate whether AI was used, nor does it impose a general additional duty on applicants to disclose the extent to which AI was used to develop an invention. Therefore, the mere use of AI as a drafting tool does not automatically mean an applicant must submit a statement saying that generative AI was used to draft portions of the specification. The relevant question is what information the USPTO actually requires, and why.
How might an applicant respond to a § 1.105 request?
Consider an examiner who has a legitimate question concerning how a particular claimed feature was developed. The applicant's response could focus directly on the requested factual issue. Under MPEP §§ 704.12 and 704.12(b), a complete response addresses each enumerated requirement by either providing the requested information or stating that it is unknown and/or not readily available. The applicant must provide information that is reasonably and readily available, but is generally not required to search for unknown information.
The claimed architecture was conceived by the inventor while investigating the identified technical problem. The inventor began with the conventional architecture described in Reference A but identified limitations associated with X. The inventor subsequently developed the alternative configuration shown in the application to address those limitations. Attached are the inventor's technical notes and References B and C, which were considered during development.
That response establishes a factual development history. Depending upon what the examiner requested, the applicant might trace a path from problem to human reasoning to technical development to the claimed solution, supported by laboratory notebooks, engineering drawings, research papers, design documents, simulation results, dated technical notes, emails, prototypes, or other relevant evidence. The purpose is not necessarily to provide a biography of every software tool involved in preparing the application. It is to answer the examiner's actual requirement.
What if AI was used only to draft the patent?
This is where careful terminology matters. Imagine the inventor tells an AI system:
The device uses components A, B, and C. I developed configuration D because conventional configuration E produces excessive thermal loading. Write a detailed patent specification describing the architecture and possible embodiments.
The AI may produce twenty pages of polished patent prose. But the underlying inventive concept may still have come from the human inventor. If an examiner asks for information concerning how the claimed invention was conceived, the relevant response may concern the inventor's technical reasoning, development history, documents, experiments, and other evidence responsive to that request. There is not automatically a requirement to volunteer every drafting tool that subsequently helped express that invention in patent language.
That changes, however, if the examiner's lawful § 1.105 requirement specifically reaches information about the drafting process or AI interaction. An applicant cannot evade a properly framed request by answering a different question. Responses to the USPTO must also remain accurate and consistent with the applicant's other disclosure obligations.
A different situation: AI participates in developing the claimed concept
Now change the facts. Suppose an inventor begins only with a problem statement: a better way to solve problem X is needed. The inventor asks an AI system to generate possible technical architectures. The AI proposes configurations A, B, and C. The inventor chooses configuration C, and configuration C becomes the central limitation of the patent claim. That presents a much more complicated inventorship question than using AI merely to improve wording.
The USPTO's November 2025 revised guidance reaffirms that AI systems are tools rather than inventors, and that the ordinary human inventorship and conception standard remains controlling. If facts create a reasonable basis for questioning whether the named human inventor actually conceived the claimed invention, § 1.105 could provide the examiner with a mechanism to seek information relevant to resolving that question. The focus therefore shifts away from whether AI was used at all.
Three different AI scenarios
It can be useful to think about AI involvement as a spectrum rather than a binary fact.
Figure 2 · A spectrum of AI involvement
Scenario 1
Inventorship exposure · low
AI as a writing tool
Principally a drafting question rather than an inventorship question.
Scenario 2
Inventorship exposure · medium
AI as an analytical tool
More factual analysis. Did the human conceive what is claimed?
Scenario 3
Inventorship exposure · high
AI generates the claimed solution
The most significant inventorship concern.
| Scenario | Where AI enters | Primary question | Nature of the issue |
|---|---|---|---|
| 1 · Writing tool | After conception, to express the invention | Is the specification accurate and enabling? | Drafting |
| 2 · Analytical tool | During analysis, with human evaluation and development | Did the human conceive what is claimed? | Factual, requires closer analysis |
| 3 · Solution generator | Before conception, producing the claimed configuration | Whose conception is reflected in the claim? | Inventorship |
The practical lesson for inventors
As AI becomes integrated into research and patent drafting, good documentation becomes increasingly valuable. Inventors should be able to reconstruct the intellectual and technical development of important inventions.
- What problem were we trying to solve?
- What did we know at the beginning?
- What alternatives did we consider?
- What technical decisions did we make?
- Why did we choose the claimed solution?
- What documents, experiments, drawings, or analyses demonstrate that development?
If an examiner later has a legitimate reason to inquire under § 1.105, this record can make answering the examiner much easier. It also allows the applicant to distinguish between tools used to express an invention and the human activity that produced the invention itself.
The strongest citation chain
The central argument of this note rests on three authorities read together. 37 CFR § 1.105 gives the examiner authority to request information reasonably necessary to examination. MPEP §§ 704.11 and 704.14 describe the limits on that authority: the examiner must have a reasonable basis for the request and must narrowly define the expected response. Finally, the USPTO's revised November 2025 inventorship guidance for AI-assisted inventions confirms that the ordinary human inventorship and conception standard remains controlling. AI is treated as a tool, not as an inventor, and there is no general obligation to disclose the mere fact that AI assisted with drafting or analysis.
Read in sequence, these sources support a more defensible proposition: § 1.105 is not itself a rule requiring applicants to volunteer that "AI drafted this application." It is a focused examination tool. If the examiner asks a specific, properly framed question, however, the applicant must respond accurately to the actual requirement rather than simply omit responsive information.
Bottom line
37 CFR § 1.105 gives patent examiners a potentially powerful mechanism for obtaining information reasonably necessary to examination. That authority can become relevant to AI-assisted inventions, particularly where facts raise legitimate questions concerning inventorship or the origin of claimed subject matter. But § 1.105 should not be understood as a general AI-detection or AI-disclosure rule. An examiner's request must have a reasonable basis and should be appropriately focused. Likewise, an applicant's response should address the information actually requested.
When the question concerns conception, an inventor may be able to explain the technical path that led to the claimed invention and provide supporting documents demonstrating that path. When AI merely helped turn an already-conceived invention into polished patent language, that is analytically different from AI generating the claimed inventive concept itself.
As AI becomes commonplace in both research and patent drafting, that distinction, AI as a tool for describing an invention versus AI's role in generating the claimed subject matter, is likely to become increasingly important in patent prosecution.
This is not legal advice. For guidance on a specific application or examiner requirement, consult qualified patent counsel.
References
- 37 CFR § 1.105 — Requirements for Information. Legal Information Institute, Cornell Law School. https://www.law.cornell.edu/cfr/text/37/1.105
- MPEP § 704 — Search and Requirements for Information. U.S. Patent and Trademark Office. https://www.uspto.gov/web/offices/pac/mpep/s704.html
- MPEP Chapter 700 — Examination of Applications. U.S. Patent and Trademark Office. https://www.uspto.gov/web/offices/pac/mpep/mpep-0700.html
- Revised Inventorship Guidance for AI-Assisted Inventions — November 2025. U.S. Patent and Trademark Office. https://www.uspto.gov/subscription-center/2025/revised-inventorship-guidance-ai-assisted-inventions
- USPTO AI-Related Resources. U.S. Patent and Trademark Office. https://www.uspto.gov/initiatives/artificial-intelligence/artificial-intelligence-resources
- FAQs on Inventorship Guidance for AI-Assisted Inventions. U.S. Patent and Trademark Office. https://www.uspto.gov/initiatives/artificial-intelligence/faqs
- USPTO Form Paragraph 7.105 — Requirement for Information. U.S. Patent and Trademark Office. https://www.uspto.gov/web/offices/pac/mpep/mpep-9095-Form-Paragraph-Chapter.html