How AI Is Changing Trademark Examination at the USPTO

AI trademark examination is no longer a hypothetical. The United States Patent and Trademark Office has assembled an AI action plan with roughly two dozen initiatives designed to reshape how trademarks are filed, examined, and policed. From automated design code assignment to fraud-detecting account monitors, the agency is investing across every stage of the trademark lifecycle -- and brand owners, attorneys, and applicants need to understand what is coming.

"This is the thing that really excites us geeks in the IT business -- artificial intelligence," said Greg Dodson, Deputy Commissioner for Trademark Administration, during the October 2025 USPTO Hour webinar. The excitement is backed by concrete plans. The USPTO has parsed its initial list of initiatives into a strategic implementation plan, with some tools already in development and others deploying in the near term.

But the agency is also clear about what AI will not do. As Jason Lott, Managing Attorney for Trademarks Customer Outreach, told the webinar audience: "We're not necessarily replacing humans with AI. We're using AI to support the humans."

That distinction matters. The USPTO is building tools that augment examining attorneys, accelerate pre-examination processing, and detect fraud at scale -- not tools that make substantive legal decisions without human oversight. Here is what you need to know about each initiative, where it stands, and what it means for your filings.

The AI Action Plan: Two Dozen Initiatives and Counting

The foundation for the USPTO's AI push is a formal AI action plan developed in response to the executive order on removing barriers to American leadership in artificial intelligence. Dodson described the scope:

"We have an AI action plan that we've socialized both with the business unit and with the Under Secretary's office, had roughly two-dozen different initiatives on it, some of which were really AI and others were supporting activities to kind of help us bring AI into the fold."

Those roughly 24 initiatives span a wide range -- from near-term automation of manual tasks to ambitious, long-horizon systems that are still in conceptual stages. The agency is now distilling that list into a strategic plan that prioritizes what can be implemented quickly and what requires the cloud migration (targeted for June 2027) to be completed first.

The 10 Key AI and Technology Initiatives

Not all of the USPTO's initiatives are equally mature. Some are deploying now. Others are years away. The following table summarizes the current state of each major initiative as described by USPTO officials during the October 2025 webinar.

Initiative	Status	Lead Official
Design code / pre-exam automation	Near-term deployment	Greg Dodson
Image search resurrection	Reinitializing (vendor previously selected pre-COVID)	Greg Dodson
Account monitoring system	In development	Amy Cotton
Virtual assistant upgrade	Planned (includes rebrand)	Greg Dodson
AI-assisted examination	Exploring	Greg Dodson
Risk predictor for fraud	"Dream stage" -- conceptual only	Amy Cotton
Anti-fraud AI enterprise	Planning	Greg Dodson
Register Protection Office workflow automation	Funded -- 1-year project plan secured	Amy Cotton
Cloud migration (technology foundation)	In progress -- target June 2027	Greg Dodson
CRM for TMScams mailbox	In development	Amy Cotton

The most important takeaway from this table is the range of maturity levels. Some of these tools will affect your filings within the next year. Others are aspirational. Understanding the difference helps you plan accordingly.

Pre-Examination Automation: The Nearest Impact

The initiative most likely to affect applicants first is the automation of pre-examination tasks within Trademark Center (TM Center). Currently, the USPTO's Pre-exam team manually handles several steps before an application reaches an examining attorney:

• Design code assignment -- classifying the visual elements of a mark into the USPTO's design code system
• Pseudo marks -- generating the textual equivalent of design marks for searchability
• Classification -- assigning the correct Nice Classification for goods and services
• Mark description -- writing the description of the mark drawing

All four of these tasks are candidates for AI automation. Dodson described the goal as "making that a more seamless kind of experience between TM Center and then TM Exam once it gets inside of the workflow process."

For applicants, this means several things. Filing errors in classification and mark description -- which currently trigger examiner's amendments and office actions -- could be caught and corrected earlier in the process. Design code assignment, which affects how your mark appears in search results and how examining attorneys find potential conflicts, would become more consistent. And the overall time between filing and first examination could shrink as the pre-exam bottleneck narrows.

Image Search: A Long-Awaited Return

The USPTO's image search program has a complicated history. In early 2020, the agency had selected a vendor and was conducting early beta testing on an AI-powered image search system. Then the pandemic hit. The project was shelved.

What followed was a series of higher-priority technology projects: the retirement of TRAM (the Trademark Reporting and Monitoring system that had been in service for 42 years, finally retired in May 2024) and the ongoing cloud migration. Image search kept getting pushed back.

Now, the executive order has provided new momentum. "We're looking at reinitializing that and reimagining where it is because technology has certainly changed in the last several years," Dodson said.

This matters because image search is one of the most significant gaps in the current trademark examination process. Examining attorneys searching for conflicting marks must rely primarily on text-based searches -- word marks, pseudo marks, and design codes. An AI-powered image search would allow the USPTO to find visually similar marks directly, regardless of how they were coded or described. For applicants with design marks or logos, this could mean more thorough searches that surface conflicts earlier, but also more accurate results that reduce false positives.

The specific technology and timeline remain undefined, but the fact that the USPTO previously had a vendor selected and a beta underway suggests this initiative is closer to reality than many of the others on the AI action plan.

Account Monitoring and Fraud Detection

Two of the most concrete AI initiatives target the growing problem of trademark fraud. Amy Cotton, Deputy Commissioner for Trademark Examination Policy, described both during the webinar.

Account Monitoring System

The first is an account monitoring system designed to detect anomalous filing behavior in real time. The tool watches USPTO.gov accounts for suspicious patterns -- particularly sudden spikes in filing volume that suggest bulk fraudulent submissions.

"If all of a sudden the account is just sending in bulk tons of applications in a short amount of time, this tool will identify that, ping us, we can hit 'suspend' and then take a look at what that account is doing," Cotton explained.

The system is designed to work as an early warning layer. When it flags an account, the fraud team can suspend it, pull the filings out of the examination queue, and route them to the Register Protection Office for investigation. This is a meaningful upgrade from the current process, which relies heavily on manual detection and victim reporting through the TMScams@uspto.gov mailbox.

Risk Predictor

The second fraud-focused initiative is more ambitious: an AI risk predictor that would assign a fraud probability score to individual applications based on filing behavior, account data, and application attributes.

Cotton was candid about the maturity of this tool: "We're in the dream stage trying to figure out exactly what those risk predictors are and how it would work and then how it would implement into our workflow."

If developed, the risk predictor would route high-risk applications to the Register Protection Office or into a slower examination track for closer scrutiny. The specific risk signals, threshold percentages, and workflow integration have not been defined. This is an initiative to watch, not one to plan around today. For more on how the USPTO combats trademark fraud, see our guide to USPTO fraud enforcement.

AI-Assisted Examination

The most potentially transformative initiative -- and the one with the least detail -- is what Dodson called "AI-assisted examination." The concept is straightforward: use AI to help examining attorneys research prior marks, identify potential conflicts, and prepare office actions more efficiently.

The USPTO has provided no specifics on what AI-assisted examination would look like in practice. It could mean AI-generated prior art searches that surface relevant marks for the examiner to review. It could mean automated drafting of routine office actions. It could mean analysis tools that help examiners identify issues in goods and services descriptions.

What is clear is the boundary the agency has drawn. Lott's statement -- "We're not necessarily replacing humans with AI. We're using AI to support the humans" -- signals that examining attorneys will retain decision-making authority. AI tools will handle research and analysis; humans will make the legal judgments.

For trademark applicants and attorneys, AI-assisted examination could eventually mean faster examination times, more consistent results across different examining attorneys, and more thorough prior art searches. But those benefits depend on implementation details that have not yet been determined.

The Cloud Migration: Why It All Depends on June 2027

Many of these AI initiatives share a common prerequisite: the USPTO's cloud migration, which is targeted for completion on June 1, 2027.

The agency is migrating its trademark systems from legacy on-premises infrastructure to cloud-based platforms. Four trademark products are scheduled to complete the migration in the third quarter of fiscal year 2027. As Dodson explained, the cloud provides "cost savings, security, resiliency and adaptability for things that are cloud-native."

That last point -- adaptability for cloud-native tools -- is the key connection to the AI roadmap. Modern AI systems require scalable compute resources, large-scale data processing, and the ability to deploy and iterate quickly. Legacy on-premises systems make all of that harder. Until the cloud migration is complete, the USPTO's ability to deploy sophisticated AI tools at scale will be constrained.

This timeline helps explain why some initiatives are in the "dream stage" while others are deploying soon. The pre-exam automation tools (design codes, classification, mark descriptions) can likely operate within the current infrastructure. The more ambitious systems -- image search at scale, real-time risk prediction, AI-assisted examination -- may need to wait for the cloud foundation.

Virtual Assistant and User Experience

Not all of the AI initiatives target the examination process. The USPTO is also planning to upgrade its virtual assistant -- the tool that helps users navigate the agency's web resources and find answers to common questions.

Dodson acknowledged that the current virtual assistant is limited and even joked about the branding: "We're going to try to rename that thing too from 'virtual assistant,' which isn't cool."

The upgrade connects to a broader goal of reducing the volume of calls to the Trademark Assistance Center (TAC) by making self-service tools more effective. For practitioners who regularly navigate the USPTO's web ecosystem -- TSDR, TEAS, Trademark Center, the ID Manual -- a more capable AI assistant could meaningfully reduce the time spent searching for information and troubleshooting filing issues.

What This Means for Trademark Filers

The USPTO's AI initiatives will not change trademark examination overnight. But they signal a clear direction. Here is how to think about the impact across different time horizons.

Near Term (2026)

• Pre-exam automation will begin affecting how design codes, classifications, and mark descriptions are processed. Filers may see fewer pre-exam delays and more consistent coding.
• Account monitoring will make it harder for bad actors to file fraudulent applications in bulk. Legitimate filers should see fewer fraudulent registrations cluttering the register and blocking their applications.

Medium Term (2027-2028)

• Image search could return once the cloud migration is complete, fundamentally changing how the USPTO searches for conflicting design marks.
• AI-assisted examination tools could begin augmenting examining attorneys, potentially reducing pendency times and improving search thoroughness.

Long Term (2028+)

• Risk prediction systems could automate the initial triage of applications, routing suspicious filings for additional scrutiny before they enter the examination queue.
• Anti-fraud AI enterprise could integrate account monitoring, risk prediction, and cross-agency data sharing into a unified platform.

What You Should Do Now

While the USPTO builds its AI infrastructure, there are practical steps brand owners and attorneys can take today:

1. Monitor the register proactively. AI tools will improve the USPTO's ability to catch conflicts and fraud, but they will not replace your own vigilance. GleanMark's AI-powered search scores similarity across 13.9 million trademark records, helping you identify potential conflicts before they become office actions.

2. Prepare for more thorough searches. As AI-assisted examination matures, examining attorneys will likely surface more prior art. Conduct comprehensive clearance searches before filing, not after receiving a Section 2(d) refusal.

3. Clean up your filings. Automated pre-exam tools will be less forgiving of sloppy classifications and vague goods descriptions. Use the USPTO's ID Manual, file with precise language, and review your applications before submission.

4. Secure your USPTO accounts. With account monitoring detecting anomalies, ensure your own filing patterns do not trigger false positives. Use multi-factor authentication, maintain consistent filing behavior, and report any unauthorized access immediately.

5. Stay informed. The USPTO is making these changes publicly and iteratively. Follow the agency's webinars, subscribe to trademark notices, and track implementation timelines. If you are new to the trademark system, our Trademark Basics guide covers the fundamentals.

The Bottom Line

The USPTO is not experimenting with AI as a novelty. It is building a systematic technology infrastructure -- cloud migration, automation, machine learning, and artificial intelligence -- designed to make trademark examination faster, more accurate, and more resistant to fraud. The roughly two dozen initiatives in the AI action plan represent the most ambitious technology investment in the history of the trademark office.

But ambition and execution are different things. Some of these tools are deploying now. Others are years away. And the agency has been transparent about that distinction, with officials like Amy Cotton openly describing certain initiatives as being in the "dream stage."

For trademark professionals and brand owners, the right response is informed readiness. Understand what is coming. Prepare your filing practices for a more automated examination process. And monitor the register yourself, because even the most sophisticated AI tools cannot replace your own knowledge of your brand landscape.

GleanMark tracks new filings, status changes, and potential conflicts across the entire USPTO trademark register -- 13.9 million records monitored continuously so you never miss a threat. Start monitoring your brand today.