Why Are AI Patents Getting Pushback? What Can I Do About It?
Dear Will & AiME,
We’ve filed several AI-related patent applications over the past year. Lately, we’ve noticed a pattern: 101 rejections are more common, and examiners are asking deeper questions about whether the invention is “abstract”. Even when we show real-world use cases, it feels like they want something more. Are we doing something wrong, or is this a broader trend?
— R&D Director of Early-Stage Robotics Company
Short answer 💡
AI patents are facing increased scrutiny under Section 101 as examiners look for clear technical improvements rather than abstract ideas. To improve outcomes, businesses should draft applications with detailed technical descriptions of how AI models function, integrate, and solve specific problems.
Dear R&D Director of Early-Stage Robotics Company,
AI is one of the most innovative spaces for IP. Because of the surge in patent applications in this area, patent examiners have applied what appears to be heightened scrutiny, particularly around patent eligibility under Section 101, which focuses on what subject matter is proper for a patent. This is part of a broader trend in U.S. practice.
The good news? We have experts to turn to. This week, Will & AiME worked with Robert Kalinsky, Shareholder at Merchant & Gould, to provide practical strategies to improve your odds—not just at filing, but at getting to allowance.
What’s Driving Increased Scrutiny of AI Patents?
Examiners are increasingly classifying AI-based innovations as “abstract ideas”, especially where the invention is not tied to a clear technical improvement. That’s true even when the AI model is novel or applied in a business-specific context.
What they’re looking for:
improvement to computer functionality or AI model performance, not just a use of AI in a process;
clear technical problem being solved—not just automation or decision-making; and
specifics, including how the model is trained, used, updated, and integrated into the system.
This mirrors patterns we’ve seen in fintech, software, and data processing patents. AI is now getting the same treatment: abstract unless clearly concrete.
USPTO Guidance vs. Examiner Reality
The USPTO has issued several recent AI memoranda aimed at clarifying eligibility. These generally favor applicants, noting that AI inventions are not inherently abstract. But in practice, examiners still demand strong technical anchoring.
If your invention uses AI but doesn’t improve it—or doesn’t show how it improves other systems—you may still face 101 challenges.
Practical Strategies for Drafting Stronger AI Patents
We’re seeing stronger outcomes when applications build in more technical detail up front, especially in the following areas.
1. Clarify What the AI Is Actually Doing — Go beyond “the model predicts…” Break it down:
How is it trained?
What data is used?
What’s the input/output structure?
How does it operate on that input/output?
What happens downstream?
Think of this as the anatomy of the model’s function, not just its output.
2. Frame the Technical Problem — Don’t say “the invention improves user experience” in the specification. Instead, say:
“The invention improves processing speed in low-memory environments.”
“The model enables accurate detection in noisy data sets where rule-based systems fail.”
“The invention improves operation of the model itself.”
Include metrics if possible: accuracy gains, reduced latency, better precision. This grounds the invention in technical reality, not abstraction.
3. Explain Implementation & Integration — Where does the model sit in the system?
Is it behind an API?
Does it talk to a database?
What preprocessing or postprocessing is needed?
Are there validation or fallback mechanisms?
Describe the system like a machine, not just a thought process.
4. Disclose Multi-Model Interactions (if any) — If your system uses more than one model, explain:
How they interact.
Why multiple models are needed.
What challenges are solved through this architecture (e.g., load balancing, data transformation).
AI inventions are patentable, but not if they look like ideas floating in the air. To overcome 101 rejections (they are unavoidable in this climate), businesses must draft patent applications with precision, highlight the technical gains, and connect the dots between innovation and implementation.
Think like a technologist, write like a systems engineer, and disclose like an inventor.
— Robert Kalinsky + Will & AiME
Three Takeaways:
AI-related patents are facing more Section 101 scrutiny—especially when the invention lacks a clear technical improvement.
Drafting strategies matter: details about model training, implementation, and system integration can make or break a case.
Treat the AI as a technical component of a machine—not just a decision engine—to strengthen eligibility and survive examination.
