WaPo Reports USPTO Accelerating AI Integration into Patent Examination
Starting July 21, it will be “mandatory” for examiners to run an AI similarity check
IP stakeholders (who read this blog) are fully aware that the U.S. Patent and Trademark Office (USPTO) is moving assertively to integrate artificial intelligence into the core of its patent examination process. Recent reports, however, confirm that AI tools are not only being tested but are becoming mandatory for examiners.
While the push for modernization promises efficiency gains, it also brings to the forefront critical questions about confidentiality, transparency, and the future role of human examiners, creating a new operational reality for patent professionals.
New Mandates and Pilot Programs
According to a July 15, 2025, report from The Washington Post, the USPTO is expanding its use of AI in significant ways. The Post, citing internal records and an agency employee, revealed that the agency is launching a pilot program where patent applicants can opt-in to have their applications analyzed by an AI tool.
This tool will reportedly "trawl the agency’s databases for existing patents with similar information" and then email the applicant a list of the ten most relevant documents. The stated goal is to encourage applicants to "revise, alter or withdraw their application" more efficiently.
More consequentially for daily practice, the report states that starting July 21, it will become "mandatory" for examiners to utilize an AI-based search tool for a similarity check on all patent applications.
Stakeholders will also read confirmation of AI's expanding role beyond search. The anonymous agency employee told The Post that the large language models behind systems like ChatGPT "are very good at writing reports, and their ability to analyze keeps getting better," pointing to an all-but-certain future where AI will assist in drafting the substantive content of office actions (e.g., “SCOUT,” discussed below).
This rapid deployment has not been without internal friction. The Post noted that another AI search tool scheduled for an earlier rollout was delayed after "concerns arose that USPTO workers — and some top leaders — did not understand what was about to happen."
In response to inquiries from The Post, the Commerce Department, on behalf of the USPTO, issued a statement affirming the evaluation of AI, saying, "At the USPTO, we are evaluating how AI and technology can better support the great work of our patent examiners."
An Ongoing Initiative
Again, for those who read this blog, it is no surprise. This move is the latest step in a longer-term strategy. As discussed previously on this blog, the USPTO has been rolling out "augmented intelligence" tools for some time.
Acting Deputy Commissioner for Patents, Greg Vidovich, has highlighted tools already integrated into the Patents End to End (PE2E) search platform.
Two key examples include:
More Like This Document (MLTD): This feature allows an examiner to select a document and instantly find similar prior art from a database of over 120 million documents. Its adoption is widespread, with examiners using it nearly 850,000 times between March 2024 and February 2025.
Similarity Search: A more powerful tool that analyzes the full text of an application to find relevant art. It has been used in over 125,000 applications and has proven effective; in over 30% of cases, at least one document found by the tool is cited by the examiner.
The new AI tools will likely be SCOUT. In that June 17th “USPTO Hour,” Deputy Chief Information Officer Deborah Stephens discussed the agency's work on a generative AI web application named "SCOUT" (Searching, Consolidating, Outlining, and Understanding Tool).
Developed within a secure, controlled AI lab, SCOUT provides staff with access to advanced generative AI models. The platform entered its alpha phase in March 2025 with over 200 users and is slated for a beta release later in the summer—which is now!
The current use cases being tested with SCOUT show a focus on specific examination tasks, including a "Developer Assistant" for software code analysis and an "Antecedent Check" feature to identify potential inconsistencies in patent claims under 35 U.S.C. § 112. There is expected to be an MPEP search tool in SCOUT, too.
The USPTO's Position on AI Tools
In an official paper from a few years ago, the USPTO framed the Similarity Search tool as a key part of its modernization efforts, designed to be an "enhanced replacement of PLUS [Patent Linguistic Utility Service] and provides examiners with optional new search capabilities to access prior art alongside traditional document retrieval approaches." The notice praises the tool for its speed, stating that AI-based searches "are completed in seconds, whereas PLUS searches historically can take hours to generate."
Anticipating the concerns of IP professionals, the USPTO’s notice directly addresses issues of data security and confidentiality. The agency assures the public that its "AI models supporting PE2E Search are trained using publicly available patent data" and that biases are mitigated by "excluding... applicant, inventor, and assignee information from the training data."
The USPTO stated it "has implemented National Institute of Standards and Technology (NIST) security controls and runs search queries using this feature within USPTO system boundaries using a Federal Risk and Authorization Management Program authorized cloud" to maintain confidentiality under 35 U.S.C. § 122(a).
To ensure transparency, the Office has established procedures for documenting the use of these tools. The notice further explained, "When an examiner uses the Similarity Search feature... the examiner’s PE2E search history will indicate that AI Similarity Search capabilities were used so the public receives clear notice that aspects of the examiner’s search were performed using AI."
Impact to the Examiners
A huge additional consideration is the practical impact on the patent examiners themselves. Examiners operate under significant time constraints, with production goals that demand high efficiency in searching and examination.
In theory, the similarity tool can be invaluable and, over the few years of use, it has proven to help many examiners. In practice, it is not always on the mark. The new mandatory nature of the reported email—and the limited time for the switchover—are a bit shocking given the long treatment of the AI-based PE2E tools as optional.
The mandatory use of an AI similarity tool, while intended to improve search quality, could paradoxically introduce new burdens. An examiner must now contend with a list of internally created AI-identified references, regardless of their actual relevance.
If the tool surfaces art that is tangentially related but ultimately not applicable for a rejection, the examiner will likely need to spend additional time analyzing and formally distinguishing those references in their search record to create a complete paper trail, thereby justifying why the AI-suggested art was not applied.
This adds a layer of documentary obligation to an already pressured workflow, potentially turning a tool meant for assistance into a mandatory procedural hurdle.
Cautious Optimism for the IP Community
The USPTO's direction offers potential benefits, including more comprehensive prior art searches—particularly of foreign patent documents—and faster examination cycles, contributing to the "goal of granting more robust and reliable patents" (p. 2).
However, the IP community should approach these developments with a degree of caution.
Transparency and Arguing Against the 'Black Box': Knowing that an AI tool was used is one thing; understanding how it arrived at its conclusions is another. Practitioners may face new challenges in responding to office actions based on prior art that was surfaced by an AI. If AI begins drafting portions of rejections, as suggested by the anonymous employee, arguing against its logic could become a complex new facet of patent prosecution.
Accuracy and Reliability: The delayed rollout of one AI tool suggests that even the USPTO recognizes the risks of moving too quickly. The potential for AI "hallucinations" or errors remains a valid concern. Quality control and the role of human oversight will be paramount to prevent a decline in examination quality driven by over-reliance on imperfect automation.
Confidentiality: While the USPTO's assurances about running AI tools within a secure, authorized cloud environment are welcome, the concept of feeding sensitive, unpublished application data into any AI system will continue to be a point of concern for inventors and their counsel. The pilot program described by The Washington Post, which involves emailing results to applicants, will warrant particular scrutiny regarding its security protocols.
The integration of AI into patent examination is no longer a future concept but a present reality. For patent attorneys, in-house counsel, and inventors, staying informed about these tools is essential. The focus must remain on ensuring that AI serves as a support for, rather than a substitute for, the nuanced judgment of human examiners.
Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.