October 30, 2024

AI Patent Drafting: Patent Drafting Copilot vs GPT

Recent comparative testing reveals that Solve Intelligence's Patent Drafting Copilot consistently outperforms GPT-4o in European patent claim drafting. Here we explore the technical and practical limitations of using GPT-4o for drafting claims suitable for European practice, and how using Solve’s Patent Drafting Copilot can deliver substantial improvements.

Testing Methodology

At Solve, we routinely run complex evaluations of our products and their ability to provide quality outputs against established benchmarks. This in turn allows us to iterate and improve on existing features and test new features, such that they perform to the high standard patent practitioners expect. 

To illustrate the benefits of this approach, we conducted a test of our Patent Drafting Copilot, and specifically its ability to draft claims for a European patent application. We used the Patent Drafting Copilot to generate claims for the last 5 years of European Qualifying Examination (EQE) Paper A questions (2019-2024), and marked the independent claims according to the following criteria:

  • Clarity and conciseness
  • Novelty and Inventive Step over the prior art provided in each paper
  • Whether the essential features of the invention are included in the claims
  • Whether unnecessary limitations have been avoided in the claims
  • Compliance with European practice requirements

The inputs we provided to the Patent Drafting Copilot included the invention disclosure and the prior art provided in each EQE paper, with no further instruction or input.

For comparison, we provided GPT-4o with exactly the same information, as well as a simple prompt outlining the task of drafting European patent claims.

Comparative Performance

AI Patent Drafting: Patent Drafting Copilot vs GPT

As can be seen from the chart above, our Patent Drafting Copilot scored consistently better than GPT-4o across all of the previous 5 EQE papers. For the 2021 paper in particular, Patent Drafting Copilot scored 4x as many marks as GPT-4o. Across all five papers, GPT-4o scored an average of 38%, whereas Patent Drafting Copilot achieved 80%.

Differences in Output

From closely analysing the outputs of Patent Drafting Copilot and GPT-4o, some key differences emerge.

  1. Essential Features - Patent Drafting Copilot routinely identified more of the correct essential features of the invention from the invention disclosure, when compared to GPT-4o. The essential features are those that are required to achieve the technical solution the invention provides to overcome a technical problem. 

  1. Unnecessary Limitations - A common issue with the claims generated by GPT-4o is the presence of unnecessary features - those that overly limit the scope of the claim. In some cases, GPT-4o included more than three additional limitations in a single independent claim when compared to the model claim. In comparison, Patent Drafting Copilot rarely included very limiting features beyond those required to provide novelty and an inventive step.

  1. Clarity - There were further issues in the output of GPT-4o relating to clarity - notably antecedent basis and inconsistent use of prepositions. These issues were much less prevalent in the output of Patent Drafting Copilot.

Advantages of Patent Drafting Copilot

The above identified differences in output may be attributed to the fact that our Patent Drafting Copilot has been designed specifically for the task of drafting patent applications.

Unlike GPT-4o, which is built on a broad language model that relies on general linguistic patterns and an expansive, albeit nonspecific, knowledge base, our Drafting Copilot has been implemented with domain-specific knowledge, and is configured to draft sections such as the claims according to a methodology that mirrors how patent attorneys approach such tasks in reality.

At a high-level, Patent Drafting Copilot is built to follow detailed processes that include identifying essential features of the invention, identifying the objective technical problem to be solved, and iteratively working up independent claims, with several self-analysis steps along the way. We also ensure that these processes are kept up-to-date to accommodate any changes in practice requirements across several jurisdictions.

Ultimately this means that Patent Drafting Copilot is largely able to determine and distinguish the essential features of an invention from those which may be deemed to be unnecessary and overly limiting to the scope of the claims.  

Of course, the advantages of using Solve Intelligence’s Patent Drafting Copilot extend beyond claim quality; they provide tangible, practical benefits for patent professionals. In terms of, our Drafting Copilot can reduce the time needed to draft claims, by providing attorneys with a good starting point from which to build out suitable independent claims, or by providing multiple options for dependent claims for attorneys to edit or expand on, for example. More generally, drafting the claims using the Drafting Copilot can help patent attorneys get past the initial ‘blank page’ status of an application, to help move the drafting process along.

In our experience, we find that using Patent Drafting Copilot results in either or both of a reduction in time spent drafting and an increase in quality of drafted applications, as attorneys can focus more of their bandwidth on the parts of the drafting process that require the most complex analysis and work.

Future Outlook

While general-purpose AI will continue to evolve, specialised tools like our Patent Drafting Copilot demonstrate the crucial advantage of domain-specific optimization. We continue to update and build on our Patent Drafting Copilot to accommodate:

  • Refinement of our proprietary patent-specific algorithms
  • Integration of emerging drafting practices (taking into account new requirements)
  • Ongoing optimization processes

About Solve Intelligence

Here at Solve Intelligence, we are committed to building AI-powered platforms that can help practitioners draft in every technical field, including those relating to AI-inventions. Our platforms assist with every aspect of the patenting process while keeping patent professionals at the helm of these powerful tools. In this way, we give patent practitioners the control needed to reap AI's benefits while mitigating its associated challenges. Our Patent Copilot™ helps with patent drafting, patent filing, patent prosecution, office action analysis, patent portfolio strategy and management, and future patent infringement analyses. At each stage, our Patent Copilot™ works with the patent professional, keeping them in the driving seat, thereby equipping legal professionals, law firms, companies, and inventors with the tools to help develop the full scope of protection for their inventions.

Write Patents with AI

We use AI to help legal professionals write and prosecute high-quality patents efficiently. We do this by providing an in-browser document editor that you can start using straight away.