USPTO expands AI tools across trademark examination after 250,000 application analyses

The U.S. Patent and Trademark Office has expanded the use of artificial intelligence across its trademark organization, saying all trademark work units achieved full adoption of the agency’s Scout LLM tool on July 1, 2026.

The agency also said a related classification system, known as the Classification Agentic Codification Tool, or Class ACT, had analyzed 250,000 trademark applications by July 13. The developments mark a broader deployment of AI in the federal process that determines how brands, products and services are classified and examined in the United States.

How the tools are being used

Scout LLM was made available to all trademark work units, including examining attorneys, according to the USPTO. The agency describes the tool as capable of summarizing information, analyzing data, researching topics and drafting selectable content for general audiences.

Class ACT is focused on trademark classification. Classification is a central part of an application because it identifies the goods or services associated with a mark. The USPTO’s reported total means the tool had been used to analyze 250,000 applications by July 13, although the agency did not characterize that figure as a measure of approvals, registrations or completed examinations.

Both tools are built on the USPTO’s Unified Generative AI Platform. The agency said the systems support trademark professionals rather than replace them.

Examiners remain responsible for decisions

The USPTO said examining attorneys continue to lead the work and remain responsible for decisions. That means the AI tools are being presented as assistance within the examination process, not as independent systems that approve or reject trademarks.

For applicants, trademark examination can affect when a business, creator or organization learns whether its application can proceed toward registration, or whether additional action is needed. The federal trademark register is used by startups, established companies and creators seeking protection for names, logos and other identifiers.

AI-assisted examination could influence examiner workflows and processing speed. It also raises questions about consistency and the reliability of information used in decisions. The USPTO announcement did not provide measured changes in examination speed, error rates or appeal outcomes, so the reported adoption and application-analysis totals do not by themselves establish that the process has become faster or more accurate.

Part of a wider USPTO AI effort

The trademark deployment follows earlier USPTO initiatives involving artificial intelligence in patent work. A separate agency program extended an Artificial Intelligence Search Automated Pilot Program for patent applications. That pilot involved at least 3,200 applications and was intended to gather evidence about AI-assisted prior-art searching.

The patent and trademark efforts involve different examination tasks. Prior-art searching addresses whether earlier disclosures may be relevant to a patent claim, while trademark examination includes issues such as classification and the examination of marks in relation to goods or services. The initiatives nevertheless reflect the agency’s broader effort to place AI tools inside its examination operations.

The next practical question is how the USPTO evaluates the tools after wider adoption. The agency has announced the July 1 milestone and the Class ACT usage total, while the public announcement did not disclose the model’s vendor or a detailed evaluation methodology. Continued agency reporting on processing times, accuracy, examiner oversight and outcomes would help show how the deployment affects applicants and the federal trademark register.