Access Official Technical PDF

Share Summary:LinkedInEmail

AI patent trends offer one way to track how artificial intelligence is spreading through invention. For industrial automation engineers, system integrators, and operations leaders, the significance is not simply that more AI-related patents are being filed. It is that AI is appearing across a widening range of technical fields, organizations, and locations—an indication that its role in innovation is no longer confined to a narrow group of technology specialists.

A U.S. Patent and Trademark Office (USPTO) analysis of patent documents from 1976 to 2018 used machine learning to identify AI-related inventions. Its findings show rapid growth in applications and broadening diffusion. Patent activity does not, on its own, prove that a technology has been adopted in production or delivered operational benefits. It does, however, provide a useful view of where inventors are applying AI and how that inventive activity is changing.

AI patent trends show sustained growth

Between 2002 and 2018, annual AI patent applications increased from about 30,000 to more than 60,000. AI’s share of all patent applications also rose, from 9% to nearly 16%. Looking at the share as well as the total helps distinguish growth in AI activity from an overall increase in patent applications.

The analysis identifies eight AI component technologies: machine learning, natural language processing, vision, speech, knowledge processing, planning and control, evolutionary computation, and AI hardware. A patent may involve more than one component. For example, an invention could combine machine learning with hardware designed to support computation.

Planning and control and knowledge processing were among the largest components in the analysis. These categories cover broadly applicable capabilities, including organizing information, identifying actions, and carrying out processes toward a goal. Since 2012, applications involving machine learning and computer vision increased noticeably. AI hardware applications also rose alongside computer vision, reflecting the relationship between computational demands and applications such as image recognition.

AI is spreading across technical fields

Patent classifications offer a way to see whether AI-related invention is reaching beyond its original technical domains. The USPTO analysis found that AI patents appeared in about 10% of technology subclasses in 1976. By 2018, they appeared in more than 42%.

Diffusion varied by AI component. Knowledge processing and planning and control spread across the largest share of technical areas, consistent with their general applicability. Vision, machine learning, and AI hardware also expanded, while speech, natural language processing, and evolutionary computation remained present in a smaller share of subclasses in the period studied.

Examples from the patent landscape show how varied these applications can be. One patent applied computer vision to images from a video camera to assess visibility under conditions such as fog, rain, and snow. Another used machine learning to analyze crop samples and identify whole or broken kernels, with the information intended to support adjustments to harvesting equipment. These examples illustrate AI appearing in fields beyond conventional software and computing.

More inventors and organizations are involved

Diffusion is also visible in the share of inventors and patent owners associated with AI patents. The percentage of inventor-patentees with at least one granted AI patent grew from about 1% in 1976 to 25% in 2018. The share of organizations patenting in AI followed a broadly similar upward trend.

From 2009 onward, the share of inventors with AI patents exceeded the share of organizations with AI patents. This points to AI-related invention spreading within organizations as well as across them: more inventors at organizations are contributing to patents that incorporate AI.

Information and communications technology companies made up most of the top 30 U.S. AI patent owners in the analysis. There were also notable patent holders from other sectors, including Bank of America, Boeing, and General Electric. The presence of organizations outside ICT is consistent with AI being applied in a range of industries and technical contexts.

AI invention reaches beyond established technology hubs

AI inventors were initially concentrated in large cities and established technology centers. The geographic pattern from 2001 to 2018, however, showed activity spreading across states and counties. The analysis identified AI-related inventions connected to areas including fitness equipment in Oregon, materials analysis in Montana, and agriculture in North Dakota.

Other examples included machine learning applications in medical imaging and disease prognosis, as well as systems for analyzing telephone calls or assessing cellular signal quality. These cases show that AI-related patenting can reflect local industry and research activity, rather than a single, uniform path of adoption.

How industrial decision-makers should read the findings

For automation and operations teams, the patent landscape is best treated as a signal of inventive activity—not a deployment roadmap. A patent does not establish that an invention is commercially available, integrated with a plant’s control architecture, or suitable for a particular process. Nor do patent counts establish productivity gains or return on investment.

The findings do provide useful context for technology planning. AI-related invention spans capabilities that may intersect with industrial systems, including:

  • Machine learning for models that learn from data.
  • Vision for extracting information from images and video.
  • Planning and control for selecting and carrying out actions toward defined goals.
  • Knowledge processing for representing and using information in automated systems.
  • AI hardware for computing components designed to support demanding algorithms.

These categories can help technical teams frame questions when evaluating emerging applications: what capability is involved, what data and computing resources might be required, and how does the proposed use fit the process being addressed? The patent evidence alone cannot answer those implementation questions, but it shows that AI-related invention is increasingly distributed across technical disciplines and organizations.

Conclusion: use patent activity as context, not a forecast

From 2002 to 2018, AI patent applications more than doubled, while the share of inventor-patentees with AI patents reached 25% by 2018. AI also appeared in a growing proportion of technology subclasses and spread across organizations and U.S. regions. Together, these patterns indicate that AI had become an increasingly broad part of U.S. invention.

For industrial decision-makers, the practical takeaway is to track AI as a cross-cutting area of innovation while evaluating applications against specific operational needs. Patent trends can help reveal the breadth and direction of inventive activity, but decisions about deployment still require attention to the technology, process, and implementation context.

Access Official Technical PDF

About The Author