Top Generative AI Patent Trends in 2025

Key Findings: Patent Landscape Report on Generative Artificial Intelligence

A global patent overview on generative AI
In just two years, generative AI has shifted from early-stage experimentation to large-scale commercial deployment across industries. The 2026 WIPO Technology SPARK Report: Patent Trends Updates in GenAI reveals the latest patent trends in generative AI inventions in 2025.

Building on the 2024 WIPO Patent Landscape Report: Generative Artificial Intelligence, which captured the initial growth in generative AI inventions between 2014 and 2023, this WIPO Technology SPARK Report tracks the growth up to 2025. It highlights the most significant shifts in who's patenting, what they're protecting, and where the technology is heading.

#

Number of GenAI Inventions in the World in 2025

GenAI inventions roughly doubled from 18,862 in 2024 to 37,808 in 2025.

This acceleration builds on the post-2017 growth trend identified in the 2024 Patent Landscape Report, but the scale is new. The R&D wave triggered by ChatGPT’s release in late 2022 and the commercial explosion of large language models is clear from the global patent filings.

Who are the top patent owners in GenAI in 2025?

  1. SoftBank 
  2. Tencent Holdings 
  3. Ping An Insurance Group 
  4. Baidu 
  5. Chinese Academy of Sciences 
  6. State Grid Corporation of China (SGCC) 
  7. Alphabet/Google 
  8. Zhejiang University 
  9. Microsoft 
  10. IBM 

Where are the most GenAI technologies invented in 2025?

Seven of the top 10 GenAI-patenting locations have seen more GenAI patent families published in the last two years alone than the entire preceding decade.
China remains the dominant origin for GenAI patents but the United States and Japan are growing fastest. But, the geographic spread of inventions is widening.

Top GenAI-patenting economies (2025):

  1. China
  2. United States
  3. Japan
  4. Republic of Korea
  5. India
  6. Germany
  7. Canada
  8. United Kingdom
  9. Israel
  10. Switzerland

Which GenAI model drives the most GenAI-based patents in 2025?

A significant change in the updated data is that large language models (LLMs) have overtaken generative adversarial networks (GANs) as the largest GenAI model category by patent volume. Among the most important GenAI programs are:
  1. Large language models (LLMs)
  2. Generative adversarial networks (GANs)
  3. Diffusion models
  4. Variational autoencoders (VAEs)
  5. Autoregressive models
However, not all GenAI patents can be assigned to these five specific core models based on available information from patent abstracts, claims or titles.

Which are the main GenAI output and input modes?

The updated data shows that, among the different GenAI modes, image/video remains the largest category. Between 2014 and 2025, there were approximately 40,000 published patent families relating to image/video, with over 13,800 published in 2025 alone. The main GenAI data types include:
  • Image
  • Video
  • Text
  • Speech
  • Sound
  • Music
The remaining modes: 3D image models, chemical molecules/genes/proteins and code/software have far fewer patents so far.