Key Takeaways
- Software patents are real, but the courts protect a specific technical improvement to how a computer works, not an abstract idea running on a generic computer.
- The strongest software patent examples, including *Enfish*, *DDR Holdings*, and *Finjan*, all claimed a particular technical mechanism rather than a desired business result.
- Run your software invention through four filters before filing: specific technical problem, particular claimed implementation, technical improvement in both spec and claims, and novelty over prior art.
- A provisional patent application locks in a priority date for 12 months, but only for subject matter it actually describes in enough detail.
- Global generative AI patent families climbed to more than 14,000 in 2023, so filing early matters more in AI-adjacent software than ever.
The Bottom Line
After Alice (2014), eligibility rejections for software patents rose 31%—but 10 court-tested patents show that claiming a specific technical mechanism (not a business result) still earns durable protection worth up to 20 years.
What You Need to Know
The Alice/Mayo two-step test is the central obstacle: examiners reject claims directed at abstract ideas unless the claim adds a specific inventive concept that transforms it into patentable subject matter. Simply adding the word 'computer' to a business process fails. The USPTO's 2019 guidance later reduced first-action eligibility rejections by 25% and cut examiner uncertainty 44%, but the bar remains high.
All 10 court-tested examples—from Enfish's self-referential database table to Finjan's security-profile attachment—succeeded because they claimed a particular technical mechanism, not a desired outcome. The same invention can pass or fail depending solely on claim language: result-oriented claims face far greater rejection risk than those specifying the exact computational steps and architecture.
What To Do Next
Jump to Section
Why most software patents fail the Alice eligibility test
10 court-tested patents and the exact feature that qualified each
The 4-filter checklist every eligible software claim must clear
4 categories of software that cannot be patented
Utility, design, and provisional: which patent type fits your software
How the AI patent surge is reshaping eligibility rules right now
*Written by Andrew Rapacke, Managing Partner, Registered Patent Attorney.* Andrew Rapacke is a registered patent attorney and the Managing Partner of The Rapacke Law Group, a full-service intellectual property law firm. He helps individuals and corporations across industries with the protection, prosecution, licensing, and enforcement of their intellectual property, with deep experience in patent, trademark, and copyright matters spanning software, AI and machine learning, blockchain, medical devices, and autonomous vehicle technology. A graduate of the United States Naval Academy, Andrew served as a Naval Engineering Officer before pursuing law and remains active in the startup and inventor communities throughout Florida.
Most software inventors guess wrong about whether their idea qualifies for a patent, in both directions. Some assume their clever feature is obviously protectable. Others assume software cannot be patented at all and never file. The reality sits in the middle, and it turns on a legal line the Supreme Court drew in 2014. In technologies affected by that decision, the probability of receiving a first-office-action eligibility rejection rose 31% in the 18 months that followed, according to the USPTO.
This article gives you 10 real software patent examples: three landmark software patents, four patents tested in Federal Circuit eligibility decisions, and three recent AI patents. For each, you will see what the software did, what the claims focused on, and why that drafting approach matters. You will then get a four-part qualification checklist, a list of what software cannot be patented, and the patent types that apply to software inventions. By the end, you can map your own invention against these examples and make an informed decision about pursuing a patent application. An issued patent is not a guarantee that every claim will survive a later court challenge, but it shows how the USPTO accepted a specific implementation rather than a result stated at a high level.
10 Software Patent Examples at a Glance
The strongest software patent examples do more than attach familiar goals to a generic computer. They identify a particular system, data structure, interface rule, model architecture, or processing sequence. This list combines recognizable foundational patents, claims examined in Federal Circuit eligibility decisions, and recent AI grants that show how sophisticated software is being claimed today.
1. OpenAI's Code-Aware Language Model — U.S. Patent No. 12,008,341
OpenAI's 2024 patent covers systems that use a machine-learning model trained on computer code to generate natural-language documentation, including docstrings. The claims describe a concrete pipeline involving code-aware training, candidate outputs, and selection or scoring, not the broad idea of using AI to explain software.
2. Google's Cross-Application AI Agent — U.S. Patent No. 12,386,645
Google's 2025 patent covers automating semantically related tasks across different application contexts. The claimed system observes actions in one graphical interface, represents them through domain models and semantic or action embeddings, translates them for another context, and performs the corresponding task in a second interface.
3. Anthropic-Owned Multimodal Interface Agent — U.S. Patent No. 12,387,036
This Adept-origin patent, now assigned to Anthropic, covers a multimodal agent for image-and-text interface automation. Its claims focus on implementation details such as image patches, tokenized text and interface actions, projection into a common representation, and a decoder-only transformer that produces an action sequence. It should be understood as an Anthropic-owned patent, not as a patent on Claude generally.
4. Amazon's Single-Action Ordering System — U.S. Patent No. 5,960,411
Amazon claimed a client-server ordering flow that used a customer identifier and previously stored purchasing information to complete an order in response to a single action. The patent has expired, but it remains a useful example because the claims described a defined network process rather than merely claiming the goal of making checkout easier.
5. Google's PageRank System — U.S. Patent No. 6,285,999
Google's early PageRank patent ranked nodes in a linked database by using the relationships among those nodes. Its significance lies in the claimed computational method: it did not attempt to own the broad idea of ranking search results, but described how link structure could be processed to generate relative importance scores.
6. Apple's Touch-Screen Heuristics — U.S. Patent No. 7,479,949
Apple claimed particular heuristics for interpreting finger contacts on a touch screen as different commands, including vertical scrolling, two-dimensional translation, and transitions between displayed items. The example shows how a software patent can focus on functional interface behavior and input processing rather than the screen's appearance alone.
7. McRO's Automated Lip Synchronization — U.S. Patent No. 6,307,576
McRO used specific rules for morph weights and phoneme transitions to automate facial animation. The Federal Circuit found the claims eligible because they recited a particular process that improved computer animation, rather than claiming the general result of automated lip synchronization.
8. DDR Holdings' Hybrid Webpage System — U.S. Patent No. 7,818,399
DDR Holdings claimed a system that generated a composite webpage combining the visual elements of a host site with third-party product information. The Federal Circuit treated it as an Internet-specific solution to an Internet-specific problem, not simply a conventional commercial practice performed on a computer.
9. Enfish's Self-Referential Database Table — U.S. Patent No. 6,151,604
Enfish claimed a self-referential table that stored different kinds of information within a unified data structure. The Federal Circuit emphasized that the claims were directed to an improvement in computer functionality itself, including more flexible configuration, faster searching, and more efficient memory use.
10. Finjan's Downloadable Security Profiles — U.S. Patent No. 6,154,844
Finjan claimed attaching a security profile to downloadable content so a system could identify potentially hostile operations before execution. The Federal Circuit treated the behavior-based security mechanism as a specific improvement in computer security rather than the abstract idea of collecting and organizing information.
Why Software Patents Are Harder to Get Than Most Inventors Expect
Software patents fail more often at the concept stage than at the code stage. Understanding why these examples qualify starts with the legal test every software application must clear. For the practical version, our software patents checklist for founders walks through the same test with filing steps.

The Alice Problem Every Software Application Faces
In Alice Corp. v. CLS Bank International (2014), the Supreme Court held that implementing an abstract financial-settlement concept on generic computers did not make it patent eligible. The lesson is blunt, computer implementation alone is not enough. Your claim has to integrate the concept into a practical application or add significantly more. If your software solves a specific technical problem rather than automating something a person already does by hand, it has a real chance. This is the same eligibility hurdle covered in our guide on when software should be patented.
What "Abstract Idea" Actually Means to a Patent Examiner
Patent examiners apply a two-step test. Step one asks whether the claim is directed to an abstract idea, which the USPTO groups into mathematical concepts, methods of organizing human activity, and mental processes. Step two asks whether the claim adds an inventive concept that transforms it into patentable subject matter. The word "computer" appearing in your claims does not rescue an abstract idea; computer programs alone do not satisfy the inventive step, which must add meaningfully more than the abstract idea itself. The USPTO's 2019 guidance later reduced the likelihood of a first-action eligibility rejection by 25%, and the same USPTO study found examiner uncertainty for Alice-affected technologies fell 44% in the year after that guidance.
The Technical-Improvement Framing That Separates Winners from Rejections
Successful software patent applications show concrete technical improvements to the computer or a technical field, not just a useful result for the user. This matters most for computer programs, which the law treats as eligible only when they deliver more than the ordinary running of computer software on hardware. The European Patent Office frames a similar idea as a "further technical effect." Before drafting claims, articulate the specific technical problem your software solves at the system level, including the technical details of the mechanism involved.
What the Strongest Software Patent Examples Have in Common
The strongest examples above make four drafting principles visible. Run your own software invention through these filters before spending money on a patent application.

The Four Qualities Every Successful Software Patent Shares
First, the strongest claims identify a specific technical problem rather than stopping at a business objective. Second, they recite a particular method, architecture, data structure, or processing sequence rather than a general idea. Third, the specification connects that mechanism to a technical improvement. Fourth, the claimed invention must still be novel and non-obvious over the prior art. The Federal Circuit examples above provide the strongest guidance on eligibility under 35 U.S.C. § 101, while the recent AI grants show how applicants are currently tying model behavior to concrete system architecture. The USPTO MPEP §2106 explains the eligibility analysis examiners apply. Enfish did not claim "a faster database." It claimed a self-referential table structure. That specificity is the checklist in action.
What the Examples Reveal About Claim Language
The gap between a granted software patent and a rejected one often comes down to how the claims are drafted, not what the software does. The same invention can pass or fail the abstract idea test depending on whether the claims articulate the technical aspects of the improvement or stay at a functional, result-oriented level. Result-oriented claims that lack implementation detail face far greater eligibility risk. The USPTO instructs examiners to look for a technical explanation of the improvement in the specification. Evidence submitted later cannot repair a specification that never disclosed it. Describe the specific computational steps that computer programs execute, not just the outcome.
What Software Cannot Be Patented Under U.S. Patent Law
You need to see both sides to evaluate your own invention honestly. Some categories are off limits no matter how they are drafted.

Abstract Ideas, Math, and Mental Processes
The patent system does not protect an idea, a law of nature, or a pure abstract idea by itself. That includes mathematical formulas and human activities that could be performed mentally or with pencil and paper, such as methods of organizing or calculating. This is why many early fintech and e-commerce software patent applications failed. They automated a financial concept without a new technical process. The nuance matters, a mathematical concept or mental process may still appear in an eligible application when the claim as a whole integrates it into a practical technical application. If your software's core novelty is the logic or formula rather than the technical implementation, talk to a patent attorney before filing.
Open Source, Prior Art, and Public Disclosure Pitfalls
Software already disclosed in prior art under 35 U.S.C. 102 cannot receive patent protection. Building on an existing open source project does not automatically make your improvement unpatentable, but open source software code and documentation can count as prior art, and licenses may affect commercialization. A prior art search before filing is a highly advisable risk-reduction step, and the fastest way to learn whether your invention is genuinely novel. Public disclosure of your own software can fall within the U.S. 12-month grace period, but the same disclosure can immediately damage foreign rights. File before you demo, publish, or launch if international protection matters.
The Two Patent Types That Apply to Software Inventions
Many software developers do not realize more than one path can protect a single product. There are two patent types — utility and design — plus a provisional application filing option that lets you secure a priority date before you are ready to file a complete non-provisional application. For a full breakdown of costs, see our guide on software patent cost.
Utility Patents Protect How Software Works, and Provisional Patent Applications Let You File Before You Are Ready
A provisional patent application is not a patent and is never examined on its merits. Provisional patents are optional precursors to the non-provisional. They last 12 months, let you use "patent pending," and do not require formal claims. It secures a priority date, but only for subject matter it actually describes in enough detail. Your later claims receive that date only to the extent the provisional supports them. For SaaS founders still iterating on architecture, well-drafted provisional patent applications buy time while you build. This is the strategy we cover in our guide to patents for startups.
Utility patents are the primary vehicle for protecting software inventions and last up to 20 years from the earliest filing date, subject to patent term adjustment, extension, and special rules for certain older filings, per USPTO MPEP §2701. For most modern software applications, that term is measured from the earliest effective non-provisional filing date in the family, which is one reason continuation practice matters. Utility patents protect the functional method or process, not appearance. Nearly every example above is a utility patent. If your software does something new and technically useful, a utility patent application is the right starting point. To see how long that coverage runs, read how long software patents last.
If you file a provisional patent application first, your non-provisional patent application can claim the benefit of the provisional’s earlier filing date, provided the non-provisional is filed within 12 months. However, that earlier priority date applies only to subject matter adequately disclosed in the provisional application. Any new material added to the non-provisional application will generally receive the later non-provisional filing date.
Design Patents Protect How Software Looks
Design patents protect the new, original, ornamental appearance of an interface, icon set, or screen layout and last 15 years from grant for applications filed on or after May 13, 2015, with no maintenance fees. They do not protect functional aspects of the software. Use a design patent alongside a utility patent for UI-heavy software products, not instead of one.
What to Consider Before Filing a Software Patent Application
Is Your Software Invention Novel and Non-Obvious?
Patent law requires novelty, meaning no prior art, and non-obviousness, meaning the invention would not be obvious to a skilled software developer during software development. A prior art search covers USPTO databases, academic publications, open source repositories, and international patent filings. Many software patent applications fail on prior art grounds a basic search would have surfaced. The search reduces avoidable rejection risk, surfaces related inventions, and informs how broadly you can claim.
Does the Business Value Justify the Cost and Timeline?
As of the USPTO's January 19, 2025 fee changes, the undiscounted combined filing, search, and examination fees for a new utility application total $2,000 for a large entity, per the USPTO fee updates. Under the current 60% and 80% discounts, small entities generally pay about $800 and micro entities about $400 for these combined fees when filing electronically. Filing on paper adds a non-electronic surcharge of $400 for large entities and $200 for small and micro entities, which electronic filing through Patent Center avoids. Confirm the exact utility issue fee against the latest schedule because it is adjusted periodically. Professional fees are additional. The USPTO does not publish a software-specific timeline, so treat "two to three years" as a rough, non-guaranteed estimate that continuations, RCEs, and appeals can extend. Weigh that against the right to exclude competitors, licensing potential, and portfolio value for investors.
How a Patent Attorney Changes the Approach
The claims drafted in a software patent application define the entire scope of protection. Specialized patent attorneys who work with software inventions draft claims that anticipate the Alice challenge from the start, though no attorney can guarantee allowance. Andrew Rapacke is a Registered Patent Attorney whose practice focuses on software and technology filings. The attorney you choose should have a portfolio of granted software patents you can verify, which is why our guide on software patent attorney selection is worth reading before you commit.
How AI and Emerging Tech Are Reshaping Software Patent Eligibility
Artificial intelligence has become the busiest frontier in software patents, and the rules are shifting fast.

The AI Patent Filing Surge and What It Means
Global generative AI patent family publications rose from less than 800 in 2014 to more than 14,000 in 2023, according to WIPO. That surge means far denser prior art in machine learning categories. In AI-adjacent software, filing earlier matters more than in legacy categories.
What the USPTO's Current AI Rules Mean for Applications
The USPTO issued 2024 guidance adding three eligibility examples for AI inventions, and it did not create an easier test. AI must still claim a specific technical improvement to a computer process, not just "using AI" to produce a result. On November 28, 2025, the USPTO rescinded its earlier AI inventorship guidance in full. The current rule applies the traditional conception standard to every invention, only natural persons can be inventors, and AI, including generative AI, is treated as a tool. Document your human inventors' technical contributions throughout development, especially when AI tools were involved. For how these rules play out in recent filings, our 9 recent software patent examples from top companies and startups is a useful companion.
Frequently Asked Questions
Do software patents exist?
Yes, though "software patent" is a practical label rather than a separate statutory category. Software-implemented processes and systems are generally protected through utility patents when they meet eligibility, utility, novelty, non-obviousness, and disclosure requirements, per the USPTO. All 10 examples above are real issued patents. Four were examined in Federal Circuit eligibility decisions, while the three recent AI grants illustrate current claim-drafting approaches and have not been presented here as court-tested.
Can you patent an idea for software?
Not the idea alone. The application must describe a definite technical invention in enough detail to show how to make and use it. A general goal such as "use AI to detect fraud" is insufficient without the claimed implementation. A provisional application can secure a filing date while you develop the full specification.
What qualifies for a software patent?
Software has a stronger eligibility position when it improves computer functionality or another technical field through a specific claimed mechanism that clears the inventive step, such as a new database structure or defined interface behavior. To patent software successfully, merely performing a business or mental process faster on a generic computer is usually insufficient.
What are the requirements for a software patent?
The invention must be patent-eligible subject matter, useful, novel, non-obvious, and supported by a written description that enables a skilled person to make and use it. Meeting these patent requirements is essential for allowance. For software, the claims and specification should identify the particular technical implementation rather than the desired result.
How much does a software patent cost?
Under the USPTO's 2025 fee schedule, the undiscounted combined filing, search, and examination fees total $2,000 for a large entity, with small and micro entities generally paying about $800 and $400 respectively. Confirm utility issue fees against the current schedule at the US patent office. Attorney, drawing, office-action, and maintenance costs are additional.
Can ChatGPT write a patent?
ChatGPT can generate draft language, but it cannot be named as an inventor. Current USPTO guidance treats AI as a tool and asks whether a natural person conceived the claimed invention. Claims-drafting strategy still needs human patent attorneys.
How do I patent my software idea?
Document the technical problem, architecture, data flow, and the specific mechanism that makes your system different. Search relevant patents, related patents, publications, and products. File a complete non-provisional, or an adequately detailed provisional followed by a non-provisional within 12 months, and be prepared to address eligibility and prior-art rejections.
How do I enforce a software patent once it is granted?
A granted patent gives you the right to exclude others, and enforcing it can involve cease and desist letters, licensing negotiations, or litigation. Our guide on how to enforce software patents walks through the practical options.
Your Next Steps to Software Patent Success
The 10 examples above share a consistent pattern. What patent offices grant is technically specific, process-oriented, and demonstrably different from an abstract idea running on a generic computer. A software invention without patent protection can be copied once it gains traction, though copyright law, trade-secret, and contract claims may still protect parts of your code. The Federal Circuit examples show why specific claim language matters under the Alice framework, while the recent AI grants show how companies are claiming model architecture, representations, and action pipelines. A patent grants a right to exclude others, not guaranteed market exclusivity.

The bottom line: A weak software patent claims a desired business result and stays at a result-oriented level, which examiners and courts reject as abstract. A strong software patent claims the particular technical mechanism that distinguishes your system from conventional ones and describes it fully before you file. The strongest evaluation question is whether you can identify that mechanism, and whether it is documented in enough detail to support your claims.
Waiting to file carries real business consequences. In AI-adjacent categories, prior art density is climbing every quarter, public disclosure can start a clock that erases your foreign rights, and a competitor who files first can lock you out of your own market. The cost of moving early is small next to the cost of losing the right to protect what you built.
Here is how to move forward:
- Schedule a Free IP Strategy Call to discuss your specific software invention and get a direct answer on whether it qualifies.
- Run a prior art search before any public disclosure, demo day, or app store launch.
- Document each human inventor's technical contribution, especially if AI tools were involved.
- Decide between a provisional and a complete non-provisional based on how settled your architecture is.
Rapacke Law Group works with SaaS founders and software developers on a fixed-fee model, handling everything from trademark registration to patent prosecution, so you know the cost before you commit.
Our Rapacke Law Group Guarantee: if our patentability search concludes that your invention is not novel, we will refund 100% of your fee — no questions asked. A well-drafted software patent is a durable competitive advantage, and getting the claims right the first time is the highest-ROI decision you can make.
Connect with Andrew Rapacke on LinkedIn or follow the firm on Twitter/X and Instagram.
To Your Success,
Andrew Rapacke Managing Partner, Registered Patent Attorney
Andrew Rapacke Managing Partner, Registered Patent Attorney Rapacke Law Group


