For decades, the question of whether an invention is obvious has turned on a deceptively simple hypothetical: would a person having ordinary skill in the art (a PHOSITA) have found this invention obvious, given what was known at the time? It’s a legal fiction, but a consequential one. The PHOSITA sits at the heart of §103 obviousness determinations, shapes prosecution strategy, and anchors invalidity arguments across all technology fields.
Now, generative AI is quietly putting pressure on that foundation. Not by replacing the PHOSITA, but by changing what that person can do. And the IP community hasn’t figured out what to do about it yet.
A Quick Primer on Obviousness
Under 35 U.S.C. §103, a patent cannot be granted for an invention that would have been obvious to a person having ordinary skill in the relevant art at the time the invention was made. Courts evaluate obviousness using the Graham v. John Deere framework, which asks four questions: What is the scope and content of the prior art? What are the differences between that prior art and the claimed invention? What is the level of ordinary skill in the relevant field? And is there any objective evidence of non-obviousness, such as commercial success, long-felt need, or failure by others?
That third factor, the level of ordinary skill, is where AI is creating the most friction. The PHOSITA is not a genius and not a novice. As the Supreme Court noted in KSR International Co. v. Teleflex Inc., the PHOSITA is “a person of ordinary creativity, not an automaton.”1 The question now is whether that creativity has been fundamentally amplified by AI tools, and if so, what that means for what counts as obvious.
The USPTO Opens the Question
The United States Patent and Trademark Office saw this issue coming. In April 2024, the USPTO issued a formal Request for Comments on the impact of AI proliferation on prior art, the knowledge of a PHOSITA, and patentability determinations more broadly. Among the questions it posed: How does the availability of AI as a tool affect the level of skill of a PHOSITA as AI becomes more prevalent? Does AI’s ability to search across disparate art fields change how “analogous art” should be defined? And how should examiners evaluate the predictability of results given that AI systems can behave unpredictably?2
The comments it received drew a clear line between two schools of thought, and the stakes on either side are significant.
Camp One: AI Raises the Bar
A number of high-profile stakeholders argued that AI tools have materially elevated what a skilled practitioner can do, and that the PHOSITA standard should reflect that reality. IEEE stated that relevant factors should include “the extent to which AI makes innovation more rapid or technology more sophisticated in the field at issue.”3 The High Tech Inventors Alliance similarly argued that in fields where AI enhances innovation, the proliferation of AI tools will generally raise the level of skill of a PHOSITA.4
Writing in IPWatchdog, patent attorney Robert Plotkin put the argument in practical terms through a pharmaceutical research scenario. A researcher in 2015 might screen a few thousand compounds over months, relying on expertise and intuition. By 2025, that same researcher can use AI tools to screen millions of compounds in days, with the system predicting binding affinities, potential side effects, and novel molecular structures a human chemist might never have conceived on their own. As Plotkin asked: “Has the widespread adoption of AI tools fundamentally changed what constitutes ‘ordinary skill’ in drug discovery?” His answer was yes, and his position was that patent law’s existing frameworks are already flexible enough to accommodate that shift.5
The argument has an intuitive logic. Patent law has always expected the PHOSITA’s capabilities to evolve with the tools of the trade. Familiarity with computational modeling, laboratory automation, and database search has long been factored into how “ordinary skill” is assessed. AI, on this view, is simply the next step in that progression.
Camp Two: Raising the Bar Has Real Consequences
Others pushed back, and their concerns carry real weight for patent owners and practitioners.
Coleman T. Strine of Baker Botts laid out the stakes clearly in a July 2025 analysis: “If the PTO chooses to consider AI-enhanced capabilities and knowledge within the definition of a PHOSITA, the level of skill in the art will naturally increase. In this case, broad claims would be more susceptible to challenges during examination and subsequent invalidity challenges after issuance. Accordingly, the value of certain patents could decrease if owners settle for more narrow claims, while the expected costs to defend these claims from invalidity challenges rises.”6
That is not a theoretical concern. A higher PHOSITA bar means that genuinely creative inventions become more vulnerable to obviousness challenges. It also hands patent challengers a more convenient argument: if the hypothetical skilled person now commands AI tools capable of combining knowledge across fields in seconds, combinations that once required a creative leap can be reframed as routine.
Several stakeholders also cautioned that any AI-adjusted PHOSITA determination should be grounded in the tools that actually existed at the effective filing date, trained on data available at that time. It’s a sensible guardrail against hindsight analysis, the same type of hindsight bias courts have long tried to guard against in obviousness determinations.
Others took the position that AI doesn’t warrant special treatment at all. The Innovation Alliance argued that under existing law, a PHOSITA is already presumed to have access to all relevant prior art, so AI doesn’t change the calculus in a meaningful way. Some commenters framed AI as analogous to earlier generations of sophisticated tools, like computers or advanced search systems, that improved human capabilities without requiring a fundamental rethink of the standard.7
A Question Worth Sitting With
One of the more pointed observations to emerge from this debate: if AI cannot be named as an inventor (a principle firmly established through the DABUS cases), why should AI capabilities define what is obvious to a skilled person?
It’s a real tension. Patent law has clearly established that an inventor must be human. AI systems cannot be credited with conception. But if the PHOSITA is presumed to wield AI capabilities in their daily work, the standard is effectively measuring human inventiveness against a benchmark that is partly defined by what machines can do. The human inventor’s contribution gets evaluated against a bar that a human, working alone, may never have been able to reach.
Where Things Stand
As of now, the USPTO has not issued final guidance on how AI should factor into PHOSITA determinations. As Strine noted, “a review of the comments submitted in response to the April 2024 Request reveals a clear divide among stakeholders over whether the use of AI tools should affect the level of skill of a PHOSITA. This division stems partially from a disagreement over how much value AI can actually add to the innovation process.”6
What is clear is that the outcome will have material consequences either way. If the USPTO excludes AI from the PHOSITA analysis, the standard may drift from the actual state of practice in fast-moving fields. If it includes AI, the effective bar for patentability goes up, making prosecution harder, existing patents more vulnerable, and invalidity challenges cheaper to mount.
For IP counsel and patent professionals, this is a debate worth following closely. The resolution, whenever it comes, will affect how claims are drafted, how portfolios are valued, and how validity is argued in litigation. The skilled person has always been a fiction. The question now is which version of that fiction best serves innovation in an era where the tools available to practitioners are advancing faster than the legal frameworks built around them.
