How a Computer System Checks and Improves AI Text
This patent describes a method where a separate computer system, using structured knowledge, checks and corrects the output from a large language model to make it more accurate and consistent.
Original patent title: “Computer implemented methods for the automated analysis or use of data, including use of a large language model”
This patent describes a method where a separate computer system, using structured knowledge, checks and corrects the output from a large language model to make it more accurate and consistent. Granted to Unlikely Artificial Intelligence in 2024 with 31 claims and 4 forward citations, and it is expected to expire in 2043.
Coverage
What does this patent actually cover?
The patent describes a computer-implemented method to improve the output of a large language model (LLM). First, the LLM generates an initial response, called "first output" (ClaimclaimA numbered sentence at the end of a patent that legally defines what the inventor owns. The most important section.Read more → 1a). Then, a separate "processing system" steps in. This system uses a special, structured, machine-readable language where ideas are represented as "semantic nodes" and "semantic links" (Claim 1b). These nodes and links can represent specific meanings, reasoning steps, and even computation units. The processing system translates the LLM's initial output into this structured language and "verifies" it for accuracy and consistency using its internal knowledge (Claim 1c, 17). For example, it might check if facts are correct or if the logic makes sense. If issues are found, the processing system creates "second input data" which includes the verified or corrected information. This "second input data" is then sent back to the LLM, often as part of a new prompt, to guide the LLM to generate an "improved first output" (Claim 1d). This iterative process aims to make the LLM's final response more factually accurate, logically consistent, and less biased (ClaimsclaimsThe numbered statements at the end of a patent that legally define what the inventor owns.Read more → 5-8).
The gap
What does this patent NOT cover?
- Does not cover LLMs that improve their own output without an external "processing system" using a structured, machine-readable language.
- Does not cover systems where the external processing system does not represent semantic nodes, semantic links, reasoning steps, and computation units in its machine-readable language.
- Does not cover methods where the LLM's first output is not translated into the machine-readable language for verification.
- Does not cover systems that only verify LLM output without feeding verified or corrected information back to the LLM to generate an improved output.
- Does not cover basic prompt engineering or fine-tuning an LLM if it doesn't involve the specific verification and feedback loop with a structured knowledge system.
- Does not cover using an external system that merely filters or edits LLM output without representing it in semantic nodes and feeding it back to the LLM.
These exclusions are unique to PatentBrief — derived from the actual claim language, not patent-office boilerplate.
Key facts
What made this novel
The clever bit is using a separate, symbolic reasoning system, which understands structured, machine-readable knowledge and logic, to verify and correct the statistical, pattern-based output of an LLM. This combines the LLM's generative power with a robust, verifiable knowledge base, creating a feedback loop that improves accuracy and consistency rather than just filtering output.
The Patent Drawing
Schematic visualization of the patent's claim structure. Hand-drawn diagrams in progress for each landmark patent.
Where you've seen this
Real-world examples
AI assistants that provide fact-checked answers
Automated legal research tools verifying case summaries
Medical diagnostic aids cross-referencing patient data with established knowledge
Financial analysis tools validating market reports
Content generation platforms ensuring factual accuracy in articles
Why it matters
The bigger picture
Large language models sometimes "hallucinate" or generate factually incorrect or inconsistent information. This patent addresses a core problem in making LLMs reliable for critical applications. By introducing a structured, verifiable feedback loop, it aims to make LLM outputs trustworthy. This approach could be crucial for industries requiring high accuracy, such as scientific research, legal analysis, or financial reporting, where errors from AI could have significant consequences. It provides a mechanism to ground LLM responses in verifiable knowledge.
Filed
April 17, 2023
Granted
May 21, 2024
Market context
Who's building on this
Companies in this space
Companies developing enterprise-grade LLM applications are actively working on similar problems of factual accuracy and reliability. Major players like Google, OpenAI, Anthropic, and Microsoft are investing heavily in techniques such as Retrieval Augmented Generation (RAG) and symbolic reasoning integration to ground LLM outputs. Startups focused on AI safety and verification, as well as those building domain-specific LLMs for regulated industries, are also exploring methods to ensure verifiable and consistent AI responses.
Market impact
This type of technology addresses the "hallucination" problem, which has been a significant barrier to wider LLM adoption in critical business functions. By offering a structured way to verify and improve LLM outputs, it enables the deployment of AI in areas where accuracy is paramount, such as legal, medical, and financial services. This could accelerate the market's trust in generative AI, potentially leading to new product categories focused on "verifiable AI" or "grounded LLMs" and expanding the overall market for AI solutions.
Claim 1 — Plain English
What this patent covers
The patent describes a computer-implemented method to improve the output of a large language model (LLM). First, the LLM generates an initial response, called "first output" (Claim 1a). Then, a separate "processing system" steps in. This system uses a special, structured, machine-readable language where ideas are represented as "semantic nodes" and "semantic links" (Claim 1b). These nodes and links can represent specific meanings, reasoning steps, and even computation units. The processing system translates the LLM's initial output into this structured language and "verifies" it for accuracy and consistency using its internal knowledge (Claim 1c, 17). For example, it might check if facts are correct or if the logic makes sense. If issues are found, the processing system creates "second input data" which includes the verified or corrected information. This "second input data" is then sent back to the LLM, often as part of a new prompt, to guide the LLM to generate an "improved first output" (Claim 1d). This iterative process aims to make the LLM's final response more factually accurate, logically consistent, and less biased (Claims 5-8).
The clever bit
The clever bit is using a separate, symbolic reasoning system, which understands structured, machine-readable knowledge and logic, to verify and correct the statistical, pattern-based output of an LLM. This combines the LLM's generative power with a robust, verifiable knowledge base, creating a feedback loop that improves accuracy and consistency rather than just filtering output.
What it does not cover
- Does not cover LLMs that improve their own output without an external "processing system" using a structured, machine-readable language.
- Does not cover systems where the external processing system does not represent semantic nodes, semantic links, reasoning steps, and computation units in its machine-readable language.
- Does not cover methods where the LLM's first output is not translated into the machine-readable language for verification.
- Does not cover systems that only verify LLM output without feeding verified or corrected information back to the LLM to generate an improved output.
- Does not cover basic prompt engineering or fine-tuning an LLM if it doesn't involve the specific verification and feedback loop with a structured knowledge system.
- Does not cover using an external system that merely filters or edits LLM output without representing it in semantic nodes and feeding it back to the LLM.
Patent timeline
Application submitted to the patent office
Application published, typically 18 months after filing
Patent officially issued
Patent enters public domain
PatentBrief Score
Impact Score
Strong
Citation count
14/40
Early citations
Claim breadth
20/20
Very broad protection
Recency
20/20
Granted within 5 years
Assignee scale
20/20
Major company or institution
PatentBrief Impact Score — based on citation count, claim breadth, recency, and assignee scale. Not a legal assessment.
Heuristic Value Estimate
What this patent might be worth
$115K – $369K
Midpoint $230K · 16.6 yr remaining · industry ×1.6
Heuristic only — blends forward/backward citation counts, claim scope, time remaining, litigation history, and CPC-derived industry baseline. Real valuations need a professional appraisal.
Claim text not yet imported for this patent
The original legal language
Original claims
31 claims as filed with the patent office.
Concepts involved
Citations
Patent lineage
Cite this patent
Heywood, R., BENN, P., REYNOLDS, D., Zhu, Z., WARREN, S., Shah, A., Tunstall-Pedoe, W., & KRNIC, L. (2024). How a Computer System Checks and Improves AI Text (U.S. Patent No. 11,989,527). U.S. Patent and Trademark Office. https://patentbrief.org/patent/us/11989527/computer-implemented-methods-for-the-automated-analysis-or-use-of-data-including-11989527
Auto-generated from the patent record. Double-check author order and the issue date against the official USPTO document before submitting.
Embed
Add this patent to your site
Drop this plain-English patent card into any blog post or article — free, no signup. It always links back to the full breakdown here.
<div data-patentlens-widget data-patent-number="US11989527"></div> <script src="https://patentbrief.org/embed.js" async></script>
Stay in the loop
Get a weekly digest of new patents.
One email per week. No spam. Unsubscribe anytime.
Keep exploring
Related patents you should know
US 4683195 · 1987
How to Make Billions of Copies of a DNA Segment
This patent describes the Polymerase Chain Reaction (PCR), a method to rapidly create many copies of a specific piece of DNA or RNA, enabling its detection and analysis.
Cetus Corp
US 8697359 · 2014
How to Edit Genes in Human Cells Using an Engineered CRISPR System
This patent describes an engineered CRISPR-Cas9 system for precisely cutting DNA in eukaryotic cells to change how genes work, opening the door for gene editing in complex organisms.
Massachusetts Institute of Technology
US 7657849 · 2010
How the iPhone's Slide-to-Unlock Gesture Works
Apple's 2010 patent describes unlocking a device by dragging a specific graphical image across the touchscreen along a predefined path, a gesture that became iconic with the original iPhone.
Apple Inc
US 4733665 · 1988
How Doctors Implant a Permanent Stent Using a Balloon
This patent describes the method for placing a permanent, expandable wire mesh tube inside a blood vessel or other body tube using a balloon-tipped catheter to widen it and keep it open.
Expandable Grafts Partnership
US 4965188 · 1990
How to Make Many Copies of a DNA Piece with Heat
This patent describes the Polymerase Chain Reaction (PCR) method, a technique to make millions of copies of a specific DNA segment using a heat-resistant enzyme and repeated temperature changes.
Cetus Corp
US 4235871 · 1980
How to Encapsulate Active Materials in Lipid Bubbles Efficiently
This patent describes a method for trapping biologically active substances inside tiny, multi-layered fat bubbles called liposomes, using a specific water-in-oil emulsion and gel-forming process to improve how much material gets captured.
Individual
Semantically similar
You might also find these interesting
US 12073180 · 2024 · Unlikely Artificial Intelligence
How a Computer System Fact-Checks AI Language Models
US 12067362 · 2024 · Unlikely Artificial Intelligence
How a Computer System Improves Large Language Model Output with Structured Data
US 12353827 · 2025 · Unlikely Artificial Intelligence
Using Non-AI Systems to Improve AI Text Generation
US 12299406 · 2025 · Casetext
How AI Checks Documents for Policy Rules and Compliance
More to explore
More in AI & Machine Learning
US 10452978 · 2019 · Google LLC
How AI Models Understand Language Using 'Attention'
US 6523026 · 2003 · Huntsman International LLC
How Computers Find Hidden Connections Between Different Fields of Knowledge
US 11615208 · 2023 · Capital One Services LLC
How Cloud Systems Automatically Create and Train AI Data Models
US 10402750 · 2019 · Facebook Inc
How Facebook Uses Deep Learning to Predict What You Might Like
New to patents?
Common Questions
Frequently Asked Questions
What does How a Computer System Checks and Improves AI Text cover?
This patent describes a method where a separate computer system, using structured knowledge, checks and corrects the output from a large language model to make it more accurate and consistent.
Who owns patent US 11989527?
Unlikely Artificial Intelligence owns this patent, granted in 2024.
When does this patent expire?
This patent is expected to expire on April 17, 2043, when the invention enters the public domain.
What is patent US 11989527 cited by?
This patent has been cited by 4 later patents that build on its ideas.
What problem does this patent solve?
Large language models sometimes "hallucinate" or generate factually incorrect or inconsistent information. This patent addresses a core problem in making LLMs reliable for critical applications. By introducing a structured, verifiable feedback loop, it aims to make LLM outputs trustworthy. This approach could be crucial for industries requiring high accuracy, such as scientific research, legal analysis, or financial reporting, where errors from AI could have significant consequences. It provides a mechanism to ground LLM responses in verifiable knowledge.
What does this patent NOT cover?
Does not cover LLMs that improve their own output without an external "processing system" using a structured, machine-readable language.
Same assignee
More from Unlikely Artificial Intelligence
Patent monitoring