\nrequirements for\u00a0patenting Graphical User Interfaces (GUIs) at the
\nEuropean Patent Office (EPO)<\/a>.\u00a0<\/p>\n
Since then, the landscape has continued to evolve, particularly
\nwith the rapid integration of Artificial Intelligence (AI).<\/p>\n
While the fundamental principles of assessing inventive step
\nremain consistent, the growing role of AI demands a more nuanced
\napproach to demonstrating a technical contribution.<\/p>\n
The enduring principles: Technical contribution remains
\nkey<\/p>\n
The core of the EPO’s assessment for GUI-related inventions
\nhas not changed. The consistent application of the problem-solution
\napproach means that patentability still hinges on two fundamental
\nquestions:<\/p>\n
\nDoes the GUI solve a technical problem?
\n
\nDoes it involve technical considerations that go beyond mere
\nuser preference or aesthetic design.\u00a0
<\/p>\n
As established in landmark case law, such as COMVIK (T 0641\/00<\/a>), features of an invention are
Innovations like predictive text input or gesture recognition
The new frontier: How does AI impact inventive step in
The central question now facing applicants is: Does the
The AI model itself is typically viewed as a non-technical
However, if the AI is used to improve the internal functioning
Key principles for patentable AI-enhanced GUI claims<\/p>\n Based on the EPO’s approach, applicants seeking to patent
Practical examples of patentable AI-enhanced GUIs<\/p>\n To illustrate these principles, consider these examples:<\/p>\n AI-powered image retrieval<\/p>\n A method where a trained neural network identifies semantically
AI-assisted control of a robotic arm<\/p>\n A system where a user indicates a target object on a real-time
AI-enhanced medical image analysis<\/p>\n A method where a user selects a region of interest on a medical
Statistical trends: A look at the EPO’s activity
The latest data reveals clear trends in this space:<\/p>\n GUI applications are growing<\/p>\n Both publication and grant numbers for general GUI cases have
Source – ipQuants<\/p>\n GUI & AI is an emerging field<\/p>\n While the numbers are smaller, the growth is significant.
Source – ipQuants<\/p>\n Source – ipQuants<\/p>\n Appeals remain challenging<\/p>\n
\nassessed by whether they contribute to solving a technical problem.
\nFor GUIs, this means distinguishing between features that offer
\ngenuine technical innovation and those that are purely decorative,
\nsubjective, or relate to non-technical business methods. The main
\nhurdle remains demonstrating an inventive step under\u00a0Art 56 EPC<\/a>.<\/p>\n
\nPresentation of Information (Output): How information is
\ndisplayed must serve a technical purpose. For example, a specific
\nlayout that enables a surgeon to position an implant more precisely
\n(as noted in\u00a0T 0336\/14<\/a>) has a technical effect. In contrast,
\nsimply colour-coding data based on user preference does not.
\n
\nMechanisms for User Input: Input features are often more likely
\nto have a technical character because they require compatibility
\nwith a machine’s predetermined protocols.
<\/p>\n
\nthat allow for faster, more accurate input or reduce the
\ndevice’s processing load are considered technical.<\/p>\n
\nGUIs?<\/p>\n
\nincorporation of AI really change anything? The answer is yes, but
\nonly if the AI’s contribution is correctly framed.<\/p>\n
\nmathematical method. Therefore, the technical contribution must
\ncome from its specific application to solve a technical problem via
\nthe GUI. Simply using AI to enhance user experience in a subjective
\nway\u2014for instance, to arrange icons or change colours based on
\npredicted user preferences\u2014will not be considered a technical
\ncontribution.<\/p>\n
\nof the computer or to control a technical process, it can form the
\nbasis for an inventive step. A clear example is an AI system that
\nanalyses user interaction to predict and pre-render data, thereby
\nreducing CPU usage and latency. This is a tangible technical effect
\non the system’s performance.<\/p>\n
\nAI-enhanced GUIs should focus on the following key principles:<\/p>\n
\nFocus on a Specific Technical Purpose: The AI-driven GUI must
\nserve a clear technical goal. This could be enhancing the quality
\nof a displayed image, improving the efficiency of data retrieval,
\nor controlling an external technical system. A GUI that assists a
\nuser in performing a technical task through a guided human-machine
\ninteraction is explicitly mentioned in the EPO Guidelines as having
\ntechnical character.
\n
\nDemonstrate a Credible Technical Effect: The effect must be
\nobjective, credible, and causally linked to the claimed features.
\nIf the presentation of information enables a user to perform a
\ntechnical task more efficiently or accurately (e.g., medical image
\nanalysis), it is considered technical. An effect that is merely
\nmore aesthetically pleasing is not.
\n
\nHighlight the “How” over the “What”: Claims
\nshould focus on how the GUI and the underlying AI algorithm achieve
\na technical result, not just what the result is. Instead of
\nclaiming a GUI that displays “more relevant” information,
\nthe claim should specify the technical means, such as a specific
\ndata processing method that reduces computational resources.
\n
\nLink to a Technical System: The GUI’s functionality should
\nideally be linked to the control or operation of a technical
\nsystem, such as industrial machinery, a medical device, or the
\ninternal workings of the computer itself.
<\/p>\n
\nrelated images from a database. It then generates low-resolution
\nthumbnails and arranges them in a grid based on a dimensionality
\nreduction of their feature vectors, clustering similar images
\ntogether. This actively guides the user, facilitating a more
\nintuitive and efficient browsing experience, which is a technical
\nsolution to the problem of retrieving images from a large database
\n(inspired by\u00a0T 0643\/00<\/a>).<\/p>\n
\nvideo feed. A convolutional neural network (CNN) determines the
\nobject’s 3D coordinates, and a reinforcement learning model
\ngenerates a collision-free trajectory for the robotic arm. Here,
\nthe AI and GUI are directly controlling a physical object, which is
\na clear technical application.<\/p>\n
\nimage. A generative adversarial network (GAN) processes this region
\nto generate a synthetic image highlighting potential anomalies not
\nreadily visible to the human eye. This provides a clear technical
\ncontribution by revealing hidden information, credibly assisting a
\nmedical professional in the technical task of diagnosis.<\/p>\n
\n(2019-2025)<\/p>\n
\nseen a steady increase, with over 1,700 applications published and
\nnearly 1,000 granted in 2025 alone.<\/p>\n![\"1747534a.jpg\"\/](\"https:\/\/www.newsbeep.com\/uk\/wp-content\/uploads\/2026\/02\/1747534a.jpg\")
<\/p>\n![\"1747534b.jpg\"\/](\"https:\/\/www.newsbeep.com\/uk\/wp-content\/uploads\/2026\/02\/1747534b.jpg\")
<\/p>\n
\nPublished applications for GUIs involving\u00a0AI<\/a>\u00a0have quadrupled from 2019 to 2025.
\nGrant numbers have also risen sharply, indicating the EPO is
\nawarding patents in this area when the criteria are met.<\/p>\n![\"1747534c.jpg\"\/](\"https:\/\/www.newsbeep.com\/uk\/wp-content\/uploads\/2026\/02\/1747534c.jpg\")
<\/p>\n![\"1747534d.jpg\"\/](\"https:\/\/www.newsbeep.com\/uk\/wp-content\/uploads\/2026\/02\/1747534d.jpg\")
<\/p>\n