Artificial intelligence has undergone a remarkable transformation from a niche academic pursuit to one of the most active domains in the US patent system. This chapter examines the trajectory of AI-related patents -- from early expert systems and symbolic reasoning through the machine learning revolution to the current era of deep learning and generative models.
Growth Trajectory
AI Patent Activity Over Time
Annual count and share of utility patents classified under AI-related CPC codes (G06N, G06F18, G06V, G10L15, G06F40), 1976-2025.
The exponential growth in AI patents mirrors the broader AI boom, driven by advances in deep learning frameworks, GPU computing, and large-scale data availability.
AI Patent Share of Total Patents
Percentage of all utility patents classified under AI-related CPC codes.
AI's growing share of all patents demonstrates that the AI boom is not merely tracking overall patent growth — it represents a genuine reallocation of inventive effort toward AI technologies.
AI Subfields
AI Patent Activity by Subfield
Patent counts by AI subfield over time, based on CPC classifications. Neural networks/deep learning and machine learning have driven the recent surge.
The shift from expert systems to deep learning reflects fundamental changes in AI methodology — from hand-crafted rules to data-driven pattern recognition.
Organizations
Leading Organizations in AI Patenting
Organizations ranked by total AI-related patents, 1976-2025.
The dominance of large tech firms in AI patenting reflects the resource-intensive nature of AI R&D, which requires massive datasets, computing infrastructure, and specialized talent.
Inventors
Most Prolific AI Inventors
Primary inventors ranked by total AI-related patents, 1976-2025.
The concentration of AI patenting among a small cohort of prolific inventors mirrors the broader 'superstar' pattern in innovation, where a few highly productive individuals account for a disproportionate share of output.
Geography
AI Patents by Country
Countries ranked by total AI-related patents based on primary inventor location.
The US lead in AI patenting reflects its concentration of major AI research labs and tech firms, but the strong showing of Japan and South Korea highlights Asia's significant investment in AI-driven electronics and consumer technology.
AI Patents by US State
US states ranked by total AI-related patents based on primary inventor location.
California's dominance in AI patents reflects powerful agglomeration effects — proximity to talent pools, venture capital, and established AI research communities creates a self-reinforcing concentration of innovation.
Quality Indicators
AI Patent Quality Over Time
Average claims, backward citations, technology scope, and team size for AI-related patents by year.
Rising backward citations and technology scope suggest AI patents are becoming more interconnected and interdisciplinary, reflecting AI's expanding role as a general-purpose technology.
AI Patent Team Size Over Time
Average number of inventors per AI-related patent by year.
Growing team sizes reflect the increasing complexity of AI research, which now requires expertise spanning machine learning, domain knowledge, hardware optimization, and software engineering.
AI Patenting Strategies
The top AI patent holders pursue very different strategies. Some focus deeply on neural networks and deep learning, while others spread their portfolios across computer vision, NLP, and other sub-areas. Comparing AI sub-field portfolios across major players reveals where each company is placing its bets and where white spaces exist.
AI as a General Purpose Technology
A hallmark of general purpose technologies (GPTs) is that they diffuse into many sectors of the economy. By tracking how often AI-classified patents also carry CPC codes from non-AI technology areas (excluding Section G which contains AI), we can measure AI's spread into healthcare, manufacturing, chemistry, and other domains.
AI Patent Co-Occurrence with Other Technology Areas
Percentage of AI patents that also carry CPC codes from each non-AI section (G excluded). Rising lines indicate AI diffusing into that sector.
AI's presence across multiple CPC sections confirms its status as a general-purpose technology, similar to electricity or computing in earlier eras. The rising co-occurrence with healthcare and manufacturing signals AI's expanding real-world impact.
AI's Inventor Problem
AI patents increasingly involve larger teams and corporate assignees, reflecting the capital-intensive nature of modern AI research. The average AI patent lists more inventors than non-AI patents, and the gap has widened over time — raising questions about individual attribution in an era of large-scale collaborative AI development.
Team Size: AI vs. Non-AI Patents
Average number of inventors per patent for AI-related vs. non-AI utility patents, 1976–2025.
AI patents consistently have larger teams than non-AI patents, and the gap has widened since 2010. This reflects the increasing complexity of AI systems, which require expertise spanning machine learning, domain knowledge, hardware, and software engineering.
AI Patent Assignee Types Over Time
Distribution of AI patent assignees by type (corporate, university, government, individual) over time.
Corporate assignees have dominated AI patenting throughout its history, but the corporate share has intensified since 2010 as large tech companies built massive AI research divisions. University AI patenting has grown in absolute terms but declined as a share.
Having charted the rise of artificial intelligence in the patent system, the next chapter shifts from structured classification data to the raw text of patents themselves. By applying natural language processing to millions of patent abstracts, we can uncover the hidden thematic structure of American innovation and trace how the language of invention has evolved over 50 years.
AI patents are identified using CPC classifications: G06N (computational models including neural networks and machine learning), G06F18 (pattern recognition), G06V (image/video recognition), G10L15 (speech recognition), and G06F40 (natural language processing). A patent is classified as AI-related if any of its CPC codes fall within these categories. Subfield classifications are based on more specific CPC group codes within G06N. AI patenting strategies show patent counts per AI sub-area for the top 20 assignees. AI as GPT measures co-occurrence of AI CPC codes with non-AI CPC sections (Section G excluded since it contains AI classifications).