AI’s Impact on Patent Examination: A Forward-Looking Perspective

This article was republished by Law360 on October 7, 2025.

Faster office actions, improved prior art searches, and a more consistent application of the law — these will be hallmarks of the future of patent examination at the U.S. Patent and Trademark Office (USPTO) due to increased use of artificial intelligence (AI).[1] Applicants and patent practitioners are faced with an unprecedented challenge that could significantly change prosecution strategy in the future.

While currently aimed at augmenting the tools available to patent examiners, the USPTO is developing more advanced AI systems to help examiners deal with the growth in patent filings and the complexity of modern inventions.[2] In recent years, the USPTO has introduced several tools to aid with classification and prior art searching.[3] Specifically, the USPTO’s Search AI tool includes Similarity Search, which enables examiners to locate documents similar to a patent application by leveraging AI algorithms to sift through millions of domestic and foreign patents and publications.[4] Recently, the USPTO also introduced DesignVision, an image-based search tool that allows patent examiners to search for visually similar designs across more than 80 global databases using image inputs, which improves the identification of relevant prior art for design patents.[5] The USPTO is also developing SCOUT, a generative AI platform designed to assist examiners with analytical and drafting tasks, including reviewing incoming documents, suggesting corrections, and navigating the Manual of Patent Examining Procedure (MPEP).[6]

As a result, the USPTO is setting the stage for significant changes in how patent prosecution unfolds. Applicants and patent practitioners should consider new and evolving strategies to overcome AI-enhanced patent examination, particularly by taking a proactive approach when drafting new applications. This article extrapolates the impacts of AI on patent examination processes across several critical categories.

Prior Art Rejections Under 35 USC Section 102 and Section 103

Traditionally, examiners were constrained by time and the limitations of keyword-based search strategies, potentially missing relevant references or relying on multiple references to support obviousness rejections when other references may be more on point. With developing AI tools, examiners can access a broader and more relevant pool of prior art, leading to potentially more robust novelty (Section 102) and obviousness (Section 103) rejections. The ability to quickly identify highly similar references means examiners will have more time to develop their arguments, and applicants may face more substantial hurdles in overcoming rejections. In the future, AI may be capable of suggesting multiple logical combinations of references, potentially reducing the number of references needed to support a rejection. This could make Section 103 rejections more difficult to rebut, as the combinations may rely on fewer references and be more closely aligned with the claimed invention.

Accordingly, as AI improves the quality of prior art combinations, patent practitioners may increasingly focus their arguments on attacking the rationale for combining references, rather than disputing the features of the references. This shift may result in arguments with more nuanced legal and technical reasoning, such as following a more comprehensive Graham analysis (e.g., an increased reliance on secondary considerations).[7] Arguments may further focus on analyzing the operation of prior art and considering whether proposed modifications are actually possible or change the fundamental principles of how a prior art reference works (e.g., whether the modification of the prior art makes the prior art inoperable for its intended purpose or changes a principle of operation).[8] This also highlights an increasing need for practitioners to perform thorough prior art searches before drafting applications and to proactively tailor draft applications to emphasize any novel points among the prior art landscape.

Patentability Rejections Under 35 USC Section 101

Subject matter eligibility under Section 101 has long been a difficult area to navigate, particularly for financial and software-related inventions. The USPTO’s AI-driven tools have the opportunity to bring greater consistency and predictability to these rejections. Specifically, AI models trained on thousands of Section 101 decisions can help examiners apply eligibility standards more consistently. By providing enhanced guidance to examiners, AI tools may reduce the variability in Section 101 rejections and improve outcomes for applicants.

Accordingly, practitioners may benefit from focusing arguments on similarities to known USPTO examples[9] and relevant cases, as AI analysis may more easily recognize similarities to known approved elements, such as known or similar practical applications (under Step 2A, prong 2).[10] However, conversely, practitioners may have increased difficulty overcoming Section 101 through more original arguments, as AI analysis tools may not understand the nuance of more individualized arguments. The use of AI tools also increases the importance of bearing in mind Section 101 while drafting patent specifications, for example, by preemptively characterizing technologies in the proper light and setting forth technical improvements with sufficient detail.[11]

Claim Objections and 35 USC Section 112 Rejections

The USPTO’s SCOUT tool currently has the capability to identify potential Section 112 issues.[12] As AI models progress, flagging issues related to claim support, enablement, and indefiniteness will occur with greater accuracy, potentially leading to more frequent and detailed Section 112 rejections. Examiners may be able to more routinely identify vague or unsupported claims, prompting applicants to provide clearer and more robust disclosures. Automated review of applications can surface formal objections and inconsistencies that might otherwise be overlooked, streamlining the examination process and improving patent quality. As with Section 101, during the application drafting stage, practitioners should consider providing support for not only the present claims but also for any anticipated future applications, such as continuations.

Other Impacts

The integration of AI is also expected to accelerate the pace of patent examination. By automating routine tasks and enhancing search efficiency, examiners can process applications more quickly, potentially reducing pendency. With future AI-driven tools aiding in drafting office actions and analyzing applications, examiners will be able to issue responses more rapidly. As examination efficiency improves, the USPTO may be able to address its longstanding backlog of applications, benefiting applicants.[13] Furthermore, as patent examination occurs more quickly, it may result in fewer delays due to the USPTO. Accordingly, this may lead to fewer cases receiving patent term adjustment (PTA), and the amount of PTA granted to cases may be reduced.

After-final practice may also become more fruitful with the use of AI, as examiners may need less time to determine whether proposed claim amendments overcome the art. Similarly, interview practice may also become more beneficial as examiners may be able to more quickly assess, through quick AI searching, if proposed claim amendments overcome the art. This may additionally improve the speed of examination.

A Future of Stronger Patents

The increasing use of AI will empower examiners with advanced search and analytical tools, providing stronger and more thorough searches, strengthening potential rejections, and providing more consistent examination. As a result, applicants and practitioners will need to preemptively adapt their strategies to address stronger and more sophisticated rejections and should consider the future potential of the USPTO’s AI tools when drafting applications. Ultimately, the use of AI will likely result in patents that are more thoroughly vetted, better supported, and less vulnerable to post-grant challenges.

[1] See https://www.uspto.gov/sites/default/files/documents/USPTO-HOUR-AI-June-17.pdf

[2] See id.

[3] See id.

[4] See https://www.uspto.gov/sites/default/files/documents/ai-sim-search.pdf

[5] See https://www.uspto.gov/about-us/news-updates/uspto-launches-new-design-patent-examination-ai-tool; https://www.uspto.gov/sites/default/files/documents/og-designvision-2025-07-16.pdf

[6] See https://www.uspto.gov/sites/default/files/documents/USPTO-HOUR-AI-June-17.pdf

[7] See MPEP § 2141(V)

[8] See MPEP § 2143.01(V) and (VI)

[9] See, e.g., USPTO Subject Matter Eligibility Examples, July 2024, https://www.uspto.gov/sites/default/files/documents/2024-AI-SMEUpdateExamples47-49.pdf

[10] See MPEP § 2106.04(d)(1)

[11] See MPEP § 2106.04(d)(1)

[12] See https://www.uspto.gov/sites/default/files/documents/USPTO-HOUR-AI-June-17.pdf

[13] See https://www.uspto.gov/dashboard/patents/production-unexamined-filing.html