Inside the USPTO’s AI Rollout: What IP Stakeholders Need to Know

By Anthony D. Marino

On April 29, 2026, the U.S. Patent and Trademark Office (USPTO) held a webinar providing a look at how the agency is integrating artificial intelligence (AI) across patent operations, trademark operations, and intellectual property (IP) policy. The session featured four senior leaders at the USPTO: Barry Schindler (Deputy Commissioner for Patents), Greg Dotson (Deputy Commissioner for Trademark Administration), Debbie Stevens (Acting CIO), and Stefano Karmis (Acting Deputy Chief Policy Officer for Patents and Innovation).

The takeaway for IP stakeholders is straightforward: the USPTO is moving deliberately to deploy generative AI tools that will change the way patent and trademark applications are examined, while simultaneously developing policy guidance on AI-related IP issues that directly affect how businesses protect and enforce their intellectual property. Whether you are in-house counsel managing a patent portfolio, a business owner filing trademarks, or an entrepreneur building your business, the developments discussed at this webinar have practical consequences worth planning for now.

The USPTO is not new to AI. The agency has used what it calls “traditional AI” or “classic machine learning” for automated classification, relevance ranking, and similar tools across the pre-examination, examination, and post-examination workflows for years. What is new is the deployment of generative AI for patent and trademark examination, internal operations, and policy development.

All AI initiatives at the USPTO are measured through KPIs that include reducing pendency, improving quality, driving efficiency gains (e.g., including lowering expenses and reducing fraud), and increasing incremental fee collections. The agency has also established an AI Policy Council, an AI Governance Council, a senior-leader AI working group with subordinate business-unit AI groups, a structured approval process for AI use cases, and an AI risk management process.

1. Generative AI Is Not Yet Being Used Widely for Patent Examination, but a Larger Rollout Is Planned

Patent examiners are not currently permitted to use generative AI for official examination. However, in Q4 of fiscal year 2025, the USPTO deployed a USPTO-approved generative AI tool to supervisory patent examiners (SPEs) to gather feedback and assess whether the tool can deliver measurable gains in patent quality and time savings.

The planned capabilities for AI tools span nearly every substantive area of patent examination. The tool would help examiners draft broadest reasonable interpretation analyses under claim construction, generate source-anchored invention summaries to guide §§ 101 and 112 analyses, identify priority chains and effective filing dates to reduce misapplication of references under §§ 102 and 103, assemble claim-to-reference mappings with pinpoint citations for novelty and non-obviousness analyses, generate reference combinations and draft obviousness rationales, provide step-by-step § 101 eligibility scaffolds, and flag potential clarity or support gaps under § 112.

Over the next 12 months, the patent side of the USPTO plans to conduct narrow, examiner-designed pilot projects in isolated environments covering claim construction, invention summaries, claim-to-citation mappings, and eligibility analyses. The agency plans to start with claim construction outlines, invention summaries, claim-to-citation mappings, and eligibility analysis, and support analysis. Critically, there will be no expansion without clear evidence of quality gains, measured through metrics including time saved on routine assembly, clarity of first actions, changes in after-finals and Requests for Continued Examination (RCEs), and examiner satisfaction. All output would remain subject to the examiner’s judgment and fully editable. Deputy Commissioner Schindler stated his guiding principle directly: “We will only deploy where the tools are demonstrably [helping] patent examiners deliver clearer, faster, and more defensible work, while preserving the patent examiner’s authorship and judgment.”

Four guardrails were described as “not negotiable” by stakeholders: (1) examiners remain the authors of record and final decision-makers; (2) every AI suggestion must be traceable to specific sources; (3) the tool must operate inside a secure, authorized environment; and (4) the system must be explainable and auditable.

2. Trademark Operations Have Already Deployed Multiple AI Tools

On the trademark side, the story is one of speed. Deputy Commissioner Dotson described an organization that was “doing basically nothing in the AI space” in the fall but has since launched several tools with more in development.

The Classification Agentic Codification Tool (Class Act) is a four-part AI tool that provides an international class for the mark, a design search code (if needed), a pseudo mark (if needed), and a mark description and/or color claim for the applicant to consider. Over 67,000 filings were processed through the tool since it was released last month, and the reduction in pendency is already being touted: “5 months into 5 seconds,” according to Dotson.

In early April of 2026, the USPTO also released an image search tool for trademarks. Users with a MyUSPTO account can upload a design into the search tool and see what exists within the corpus of federally registered marks, which is a capability the agency described as “a dream come true” for the trademark side of the USPTO.

Dotson made clear that the trademark side is not slowing down. The team is pursuing both small, quick-impact projects and larger initiatives with bigger returns on investment, with the stated goal of leading the rest of the world with trademark AI. Examining attorneys are expected to receive access to an internal generative AI tool in July 2026 after all guardrails, rules, and policies are in place.

Additionally, the USPTO has issued two Requests for Information: one seeking vendor solutions to fight trademark fraud (which the agency acknowledged is a “significant problem”) and another seeking vendors to assist with examination and quality review.

3. AI and IP Policy Guidance Continues to Evolve

On the policy front, Acting Deputy Chief Policy Officer Stefano Karmis highlighted a series of guidance documents the USPTO has issued over the past year at the intersection of AI and intellectual property.

Subject matter eligibility under § 101 has received the most attention over the past year on the policy side. Key developments include an August memo to examiners on examining patent applications involving AI, the September Ex parte Desjardins decision by Director Squires addressing eligibility of AI inventions, and a December 2025 guidance on using Rule 132 declarations to address § 101 rejections.

On inventorship, the USPTO issued guidance in November underscoring that the same legal standard for inventorship applies regardless of whether AI was involved in the inventive process.

Karmis flagged several policy issues that the USPTO is actively monitoring and on which further guidance may be forthcoming. These include disclosure requirements for AI tools, models, and training data used in the inventive process; the evolving capabilities of a Person Having Ordinary Skill In The Art (PHOSITA) now equipped with AI tools and the implications for novelty and obviousness; court case developments that may shape future guidance; and legislative activity, including the White House’s March 2026 national AI policy framework, as discussed in our March 26, 2026 legal alert.

For In-House Counsel

Expect Stronger and Faster First Office Actions. When generative AI tools are deployed to patent examiners, first office actions are likely to improve in quality: clearer claim constructions, better prior art citations, more thorough § 112 reviews, and (hopefully) better articulated § 101 analyses. This should ultimately reduce unnecessary back-and-forth with the examiner, but it also means that your claiming strategy may need to be more targeted and substantive from the outset.

Review § 101 Guidance Now. The Ex parte Desjardins decision and the Rule 132 declaration guidance have had meaningful effects on § 101 rejections in the tougher technology centers. In-house teams should review these documents and incorporate them into prosecution strategies for pending applications while keeping in mind that the gulf between the courts and the USPTO on § 101 analysis continues to widen.

Prepare for Potential AI Disclosure Requirements. The USPTO is considering how much information about AI tools, models, and training data may need to be disclosed in patent applications. In-house teams should begin documenting AI involvement in the inventive process now, even before formal disclosure requirements are adopted.

Prepare for a Change to the PHOSITA Standard. The question of what a person of ordinary skill in the art can accomplish with AI tools is one the USPTO is actively tracking. Changes to the PHOSITA standard could have significant implications for much of patent law across every technology area. It is safe to say that a PHOSITA’s capabilities are unlikely to be viewed as static once AI tools become part of the ordinary toolkit, so some patent law issues will get easier while others get more difficult.

Take Advantage of Trademark AI Tools. The Class Act tool and image search capability are already live. In-house trademark teams should familiarize themselves with these tools, particularly using the image search function for clearance work.

For Business Owners and Entrepreneurs

Trademark Filing Just Got Faster. If you are filing trademark applications, the Class Act tool can now suggest mark descriptions and color claims in the format the USPTO prefers. Accepting these suggestions where appropriate can streamline prosecution, reduce the number of office actions, and (in some situations) eliminate the need to hire a private attorney to file your trademark application.

Use the New Image Search Before You Launch. The USPTO’s new image search tool lets you search existing federally registered marks by image. Before investing in a new brand identity, use this tool (or have your counsel use it) to identify potential showstoppers early.

AI Inventions Are Patentable, but Strategy Matters. The USPTO has been issuing guidance designed to clarify the path for patenting AI-related inventions, particularly around § 101 eligibility. If your business is developing AI-powered products, consider working with your IP counsel to ensure your patent applications are drafted with this guidance and relevant court cases in mind.

Document AI Involvement in Your Inventions. With potential AI disclosure requirements on the horizon at the USPTO and elsewhere, entrepreneurs should begin keeping records of how AI tools are used in their research and development processes. Good documentation now can prevent headaches later.

The USPTO is moving with purpose on generative AI. Over the past year it has deployed tools on both the trademark and patent sides. For IP stakeholders, the message is clear: the agency is not waiting, and neither should you. Review the recent § 101 guidance, familiarize yourself with the new trademark tools, begin documenting AI involvement in your inventive processes, and prepare for the evolving PHOSITA standard. The businesses and practitioners who stay ahead of these developments will be better positioned to navigate the patent and trademark system as it evolves.