AI trademark filings at the USPTO grew 86% in a single year. From 11,449 applications in 2024 to 21,286 in 2025, the rate of growth is accelerating, not plateauing. That works out to roughly 58 AI-related filings per day, every day, for the entire year.
This piece breaks down the shape of that boom: the growth curve, the class distribution, the geographic origins, and what the patterns tell us about where AI commercialization is heading. Every number comes from Signa's production database of 14M+ USPTO records.
AI Trademark Filings Reached 21,286 in 2025
The raw numbers tell a clear story. AI-related trademark filings (applications where "artificial intelligence" appears in the goods or services description) have nearly quintupled since 2020:
| Year | AI Filings | YoY Growth |
|---|---|---|
| 2020 | 4,385 | -- |
| 2021 | 5,262 | +20% |
| 2022 | 5,527 | +5% |
| 2023 | 8,394 | +52% |
| 2024 | 11,449 | +36% |
| 2025 | 21,286 | +86% |
AI-Related Trademark Filings at the USPTO (2020-2025)
Put this in context. The USPTO received 657,793 total trademark applications in 2025. AI-related filings represent 3.2% of that total. A small share on the surface. But the overall filing system grew at single-digit rates, while AI filings nearly doubled. The growth rate is multiples of the broader market.
The acceleration is what makes this notable. From 2020 to 2022, AI filings grew about 26% total across three years. From 2022 to 2025, they grew 285%. Something changed, and the timing is not a coincidence.
A brief note on methodology: this analysis defines "AI-related" conservatively, filtering for applications that include "artificial intelligence" in their goods or services text. Including machine learning trademark applications, neural network filings, and deep learning-related marks would produce higher counts. The numbers here represent a floor, not a ceiling. For a broader look at how Signa approaches USPTO filing data at scale, see the 14M USPTO filings analysis.
The Generative AI Inflection
The growth curve has a clear inflection point, and it lines up exactly where you'd expect.
From 2020 to 2022, AI trademark filings were essentially flat. Growth of 20% in 2021, then just 5% in 2022. Companies were building AI products, but the trademark filings reflected a steady, unremarkable pace.
Then 2023 happened: +52% growth. The year 2024 brought another +36%. And 2025 exploded with +86%.
The catalyst is obvious. ChatGPT launched in November 2022 and reached 100 million users within two months. But trademark filings don't respond instantly to a product launch. There's a 12-to-18-month lag between a technology entering mainstream awareness and the resulting trademark surge. Companies need time to build products, choose names, clear those names, and file applications.
The ChatGPT moment planted the seed. The harvest arrived in 2023 and 2024. The 2025 spike represents the second wave: companies that saw the first wave succeed and decided to enter the market.
This pattern has precedent. The "cloud" trademark wave followed a similar curve between 2010 and 2013, with filings for cloud-related marks surging roughly 18 months after AWS and similar platforms gained mainstream traction. The "blockchain" wave from 2017 to 2019 was sharper but shorter, tracking closely with the crypto hype cycle. AI is following the same general shape but at a larger scale. The 2025 AI filing count already exceeds the peak of both previous technology waves.
One subtler signal in the data: the goods and services descriptions in AI filings are getting more specific. Early filings used broad language ("software utilizing artificial intelligence"). Newer filings describe specific applications ("artificial intelligence-based tools for medical image analysis" or "generative artificial intelligence for content creation"). The growth in generative AI trademarks specifically reflects a market maturing past the "AI for everything" phase into specialized products with defined use cases. These are exactly the kinds of patterns explored in Signa's analysis of what AI trademark filings reveal about unreleased products.
Where AI Trademarks Cluster: The Class Distribution
Trademark applications are categorized using the Nice classification system, an international standard that groups all goods and services into 45 classes. Class 9 covers software and electronics. Class 42 covers scientific and technological services, including SaaS platforms. Understanding which classes attract the most filings tells you where commercial activity is concentrating.
Two classes dominate AI trademark filings, and their relative trajectories tell an important story.
Class 9 (software and electronics) remains the largest filing class at the USPTO. It received 90,630 filings in 2025, up 32% from 68,560 in 2018. This is the traditional home for software trademarks: the downloadable app, the hardware device, the consumer electronics product.
Class 42 (SaaS and technology services) is growing faster. It hit 66,694 filings in 2025, up 65% from 40,468 in 2018. Based on 47,585 filings through July 2026, Class 42 is on pace for roughly 82,000 filings this year.
Class 42 vs Class 9: USPTO Filing Trends (2018-2025)
The divergence matters. Class 9 covers products. Class 42 covers services. Class 42 growing at double the rate of Class 9 signals that AI commercialization is tilting toward platforms and services, not packaged software. This is the AI-as-a-service model: companies offering inference APIs, hosted models, and cloud-based tools rather than downloadable applications.
Multi-class filing patterns reinforce this. AI companies rarely file in a single class. The typical pattern is to file across Classes 9, 42, and 35 (advertising and business services). A company building an AI writing tool might file in Class 9 for the software itself, Class 42 for the SaaS platform, and Class 35 for the marketing automation services it enables. Multi-class filings are more expensive and more complex, which means the companies filing them are serious enough about their brands to invest in broad protection.
For context, the overall filing picture in 2026 shows Class 9 at 52,133 filings through July and Class 42 at 47,585 through the same period. The gap between the two largest technology classes continues to narrow.
The congestion in these classes has practical consequences. More filings in the same classes mean more potential conflicts, more likelihood of refusal based on existing registrations, and a harder clearance process for anyone entering the market today. The fights over AI brand names (covered in depth in the AI trademark battle) are a direct consequence of this crowding.
Who Is Filing: Geography and Company Origins
Trademark owner data shows where AI development activity is concentrated, at least as reflected in USPTO filings. For filings from 2024 onward, the owner country breakdown across all classes (not AI-specific) looks like this:
| Country | Filings (2024+) |
|---|---|
| United States | 1,033,110 |
| China | 391,017 |
| United Kingdom | 24,579 |
| Canada | 22,384 |
| South Korea | 19,742 |
USPTO Trademark Filings by Owner Country (2024+)
An important caveat: these numbers represent total USPTO filings by owner country for 2024 onward, not AI-specific filings. The USPTO does not provide a clean way to filter by owner country and goods/services content simultaneously at the aggregate level. These figures show the overall filing picture, within which AI filings are a growing subset.
The US dominance is expected. Companies headquartered in the US file the majority of USPTO applications. China's position at nearly 400,000 filings reflects both the scale of Chinese technology development and the strategic importance of securing US trademark rights for international expansion.
South Korea's presence in the top five is notable. Samsung, LG, and Naver are among the most active AI trademark filers globally, and South Korea's national AI investment strategy has translated directly into IP activity. The country's filing volume at the USPTO exceeds several larger economies, pointing to a deliberate focus on the US market for AI-related products and services.
The geographic data also connects to the monthly AI trademark filings column, which tracks individual notable filings and often flags non-US filers expanding into the American market.
What AI Trademark Trends Signal About the Next Phase
Three patterns in the data point to what comes next.
The clearance problem is getting worse. With 21,286 AI-related filings in 2025 alone, the namespace for AI brands is crowding fast. Every filing that matures into a registration narrows the range of available names for the next entrant. Companies that wait to file are competing against an ever-growing wall of prior registrations and trademark squatters who race to register popular names. For startups still choosing a name, the practical window for clean, distinctive marks in AI-related classes is shrinking quarter by quarter.
Class 42 congestion is the sharpest pressure point. The fastest-growing major filing class is also becoming the hardest to clear. A 65% increase in Class 42 filings since 2018 means more registered marks, more pending applications, and more potential conflicts for anyone seeking to protect a SaaS or technology services brand. AI companies that default to Class 42 without a thorough clearance search are significantly more likely to face an office action or opposition.
The services shift reveals AI's commercial model. Class 42 growing faster than Class 9 is not just a data point. It reflects a fundamental shift in how AI reaches the market. The dominant model is not "download this AI software" but "use this AI platform." This has implications beyond trademarks: it suggests the AI market is consolidating around hosted services, APIs, and subscription platforms rather than standalone products.
For companies entering the AI market, filing strategy matters more now than it did two years ago. Multi-class filings across Classes 9, 42, and 35 are increasingly standard.
The choice between filing on an intent-to-use vs. use-in-commerce basis is particularly relevant for AI startups that have a brand name chosen but a product still in development. An intent-to-use (ITU) filing lets an applicant reserve a mark before the product launches, while a use-in-commerce filing requires the mark to already be in active commercial use. Many AI companies file on an ITU basis to secure protection early, given how quickly the space is filling up.
Consult a trademark attorney for legal guidance specific to your situation. Filing strategy depends on factors that go beyond what aggregate data can tell you.
About the Data
Source. Every number in this analysis comes from Signa's production database, which indexes 14M+ USPTO trademark records. The data includes applications, registrations, amendments, and status updates across all active and historical filings.
AI filing definition. "AI-related" means the application's goods or services description contains the phrase "artificial intelligence." This is a conservative filter. A broader definition including "machine learning," "neural network," "deep learning," and "natural language processing" would yield substantially higher counts. The numbers reported here represent a lower bound.
Time period. Calendar year data for 2018 through 2025. Partial 2026 data covers January through July.
Owner country data. The geographic breakdown (US, China, UK, Canada, South Korea) reflects total USPTO filings by owner country for 2024 onward, across all classes. These are not filtered to AI-specific filings. This limitation is noted in the geographic section above.
Legal note. This analysis is informational. It does not constitute legal advice. Trademark filing decisions should be made in consultation with a qualified trademark attorney.
The AI trademark boom is still accelerating. Whether you are building an AI product, investing in one, or advising companies that do, the filing data provides an early and quantitative signal of where the market is headed. Signa's trademark database powers every number in this analysis. Explore the data at signa.so.