The concept of the Person Having Ordinary Skill in the Art (PHOSITA), also referred to as Person of Ordinary Skill In The Art (POSITA), has long been a foundational construct in patent law, serving as a standard for patentability for determining obviousness under 35 U.S.C. § 103 and inventive step under Article 56 of the European Patent Convention (EPC) and similar statute in all or nearly all jurisdictions, which for efficiency will be referred to herein as “obviousness”.  The PHOSITA is a hypothetical person presumed to know all relevant prior art and to apply ordinary technical reasoning, which provides an objective standard of knowledge, but leaves the standard on the use of the knowledge as subjective.  As a patent practitioner it has always been a challenge contending with this mythical person that knows all and is on the side of the patent offices and companies seeking to invalidate a patent.

For business owners and their patent counsel, the rapid advance of artificial intelligence raises the question is the PHOSITA now real or at least approachable?  If so, what does that evolution mean for patents, past, present, and in the future.  Modern AI systems ingest massive volumes of patents, publications, and technical disclosures, identify relevant teachings across disciplines, and generate potential solutions by combining known elements.  In effect, AI enables a practical and useful approximation of the capabilities that patent law has always attributed to PHOSITA, i.e., having access to all prior art and determining obvious combinations thereof.  This can have immediate implications for how obviousness & inventive step are evaluated in both patent prosecution and litigation.

Traditionally, determinations of what a PHOSITA would have found obvious were set forth by a patent office or court based on advocacy-driven narratives.  Even under the structured framework of Graham v John Deere Co. 383 U.S. 1 (1966), these determinations often involve subjective interpretation.  AI introduces the possibility of computational repeatability: the same dataset can be analyzed multiple times, the same combinations of prior art can be tested systematically, and reasoning pathways can be reproduced.  This aligns with the principle that obviousness should be an objective inquiry, even if based on underlying factual determinations.  See MPEP § 2141.

Of course, the objectivity of the PHOSITA provided by AI is not absolute.  Two factors that may introduce subjectivity in the AI-based PHOSITA are training data and prompt design.  First, the scope and quality of training data determine the universe of prior art the AI can access.  If relevant references are missing, or if the dataset includes information that post-dates the invention, the resulting analysis may be incomplete or improperly influenced by hindsight.  While extremely important, this factor can theoretically be managed by proper definition of the prior art used for the reference for AI and disclosure of the dataset used for the analysis.

The second and more subjective factor is how the AI is prompted to perform its analysis.  While the AI prompt is clearly significant in the analysis that the AI performs, it may be possible to develop standard prompts that are generally agreed upon that make a reasonable request of an AI PHOSITA.  For example as starting point might be “Today’s date is April 20, 2026, please suggest improvements to a widget based on all references to widget available BEFORE today’s date and cite all references and how you arrived at each improvement without defying physical laws and limit the improvement to those with a reasonable expectation of success.”

For patent counsel, this shift also transforms how evidence is developed and presented.  Obviousness analysis has traditionally been argument-driven, relying on narratives about what a PHOSITA would have done.  AI introduces the ability to support these arguments with reproducible, data-driven evidence.  For example, AI can be used to demonstrate how specific prior art references can be combined, or to simulate the design pathways that would lead to a claimed invention.  USPTO guidance already requires examiners to provide articulated reasoning with rational underpinning for obviousness rejections, and AI is well-suited to generating such structured analyses.  See MPEP § 2143.

The AI PHOSITA is a double-edge sword as it may be configured to classify an invention as obvious OR not obvious.  If an AI system that is properly trained on the relevant prior art and configured to approximate PHOSITA is unable to identify or generate the claimed invention, that absence may be used to infer nonobvious.  Under KSR International Co.  v.  Teleflex Inc., 550 U.S. 398 (2007), obviousness depends in part on whether a PHOSITA would have had reason to combine prior art elements with a reasonable expectation of success.  If a system capable of systematically exploring the combinatorial space does not arrive at the claimed invention, it is reasonable to say that the pathway was not readily apparent and that the invention lies outside predictable design space.  While a finding of nonobvious by the AI PHOSITA does not presently provide a per se rule of patentability, it can serve as persuasive, corroborative evidence of non-obviousness, similar in function to secondary considerations recognized in Graham v. John Deere Co.

From a practical standpoint, these developments may provide more consistency in both obtaining and defending patents.  All parties may employ AI to identify more prior art and construct more robust obviousness arguments, consistent with the flexible approach endorsed in KSR.  At the same time, applicants can and should use AI proactively to assess patentability before filing, refine claim scope, and develop stronger arguments during prosecution.  The net effect can be a more rigorous and data-driven patent system, where outcomes are more predictable.

As patent practitioners know, but many businesses do not know, the preverbal deck is stacked in favor of patent challengers and against patent owners, because a patent challenger may spend millions of dollars and hundreds of hours trying to invalidate a patent, while a patent owner spends orders of magnitude less obtaining the patent and an patent examiner is afforded only a few hours to search and examine an application.  The AI PHOSITA may level that playing field at least somewhat by improving a businesses’ visibility of the prior art before filing a patent application since they have access to an AI PHOSITA without having to spend millions.  Patent offices will be able to perform more comprehensive reviews of patent applications.  Patent challengers benefit as well as the AI PHOSITA has the promise of reducing the number of 1) patent applications for inventions by businesses that are within the skill of the PHOSITA and 2) patents that issue as a result of missed prior art during examination.

For business owners, the strategic implications are several and generally beneficial.  AI can make the PHOSITA more accessible to their business, and, hopefully improve their resource allocation by allowing them to focus resources on inventions that are defensible and important to their business.  AI Tools are currently on the market and being continuously improved that enable business to integrate AI into their patent processes including invention discovery and harvesting, prior art search, drafting, prosecution, and portfolio management, that can enable businesses to apply their limited resources to developing a stronger patent strategy.  For more information on the use of AI in the patent lifecycle, please see the author’s article at https://ankar.ai/community/is-artificial-intelligence-becoming-table-stakes-for-patent-practice.

In short, AI has the promise of transforming the PHOSITA from a purely hypothetical subjective construct into a functional objective benchmark that can be approximated, tested, and leveraged.  For patent counsel and business leaders alike, the question is no longer whether AI will impact patent law, but how quickly and effectively AI can incorporated into their workflows.  Those who do will operate at the level the law has always assumed.  Those who do not may find themselves evaluated against a standard they are no longer equipped to meet.