Can SMEs use AI to write their own patent applications?

You’ll save time, avoid explaining your invention to a patent attorney and will get your IP rights granted for a fraction of the cost: on the face of it, using AI to create your patent applications has a lot of benefits. 

In fact, if you’ve ever entered your idea into an LLM such as ChatGPT or Claude and asked it to generate a patent application, you’ll no doubt have been impressed by its formatting, significant word count and how fluently it sets out key points. 

As a patent attorney firm, you’d expect us to say that using LLMs for this purpose is inherently risky, but it’s also the truth. Having received AI-generated applications from clients and prospects, we’ve spotted a few pitfalls. Here, we cover the drawbacks and potential risks, but also look at where AI can bring benefits. 

Why AI-generated patent applications don’t meet the mark

On first glance, an AI-generated patent application might look great. Even when you start reading it, it could seem to hit all the key points. However, from reviewing these documents, we’ve uncovered a few common faults. 

Inadequate content

When a patent attorney takes an application brief from a client, they’ll often ask questions you wouldn’t think were relevant, but are actually key to persuading the patent office that your claim is worthy of a patent. Beyond including the technical detail of the invention, an application needs to argue that it is new, inventive and useful. 

Writing a patent application is a form of advocacy. We need to convince a (human) examiner that your invention is brilliant. If your patent ever ends up being litigated, the court will also need to be convinced. 

Patent attorneys know how to harvest and frame the right background and context, alongside any details about an invention, into a compelling argument for an examiner. An AI tool will struggle to do this if you don’t cover all these aspects in your prompt. 

Trying to get to the answer too quickly

I have seen claims written by AI tools that are far too narrow. The first application (the priority application) needs to press the boundaries of what is new and inventive. 

We are looking to get as broad protection as possible, and we do that by filing broad claims with solid fallbacks. That allows us to get a meaningful search report back and keep our client’s options open to secure the best possible protection.

AI tools don’t tend to see it like that (unless prompted to). They will default to writing a narrow claim. Although defensible, is too narrow to be of real commercial value.

Not taking into account the commercial background

Writing a patent is not an academic exercise. Taking into account your commercial circumstances is absolutely key. This is an area that’s often overlooked in self-filed applications. 

One example is claim scope. When an attorney drafts an application they always have an eye on the market in which the patentee operates. What are their competitors likely to actually try and do? For example, if your competitors manufacture equipment, claims focused on the use of that equipment are less useful in many cases. 

Your business plans may also play into how the patent is written. A key question any good patent attorney will ask is why you want a patent. If you only want a patent for tax relief purposes, the claims may look very different to an application designed to shut your competitors out of the market.

It’s not your fault, but as a business owner (even one who’s filed a patent application before), you might not be aware of all the elements you need to include. This means you’ll struggle to cover all the information needed in a prompt for an AI tool to create a thorough, effective patent application that is likely to be granted.

National bias

Each patent office has different rules and guidance about how to structure and format applications. More significantly, some language and phrasing is unwelcome and will cause objections. 

AI tools tend to draft documents in a US-style by default. If you’re filing to the UK or European patent offices, this can cause problems during examination, increasing cost and complexity. 

You can also miss out on scope. For example, in Europe we can get broad protection with “means plus function” language in the claims. By contrast, in the US this can be interpreted narrowly and should be avoided.

If you’re planning on filing to multiple patent offices across different countries, then you’ll need to reformat and potentially “translate” the applications. This is fraught with problems for the untrained.

A lack of clarity

Patent attorneys are really good at asking stupid questions. We make sure we completely understand the invention. LLMs can fall short in asking those questions. 

Whether they’re hallucinating, caveating or not able to understand the technical aspects of your invention, LLMs can often produce applications that have inconsistent language. The phrasing hedges its bets too much or tends to be overly verbose. This can obscure the true nature of the invention or lead to fees for excess pages and claims. 

What are the risks of using AI for patent applications? 

Beyond an inaccurate, unclear document, there are some other risks to using AI tools to create a patent application.

Confidentiality 

It’s possible that any information you’re inputting into an LLM is being used to train the model. This is potentially disastrous. You may have undermined your own patent application by making an unintentional public disclosure. Although settings can be adjusted to protect the information in many paid-for tools, if you use free plans, models will have access to it by default. 

Privilege

As a Chartered and regulated patent attorney, whatever you discuss with me is protected by privilege. That means our discussions can’t be used against you in court (if you are ever unlucky enough to find yourself there). 

Your discussions with AI tools are not privileged. This means any information you share with a tool could be disclosed in court. The tool is more of an “untrustworthy advisor” than a patent attorney. 

Deadlines

It’s also important to note that AI tools won’t file and prosecute a patent application for you. This means you’ll need to keep track of any deadlines, fee payments or responses during the filing process, which can take years to complete. For some deadlines, you’ll only be told once. For others, you won’t be told at all. Some deadlines can be extended, others can’t. It’s a minefield!

Patent attorneys use a specialist software to keep track of these deadlines and carefully monitor the progress of your application. If you don’t have access to this, you could miss key deadlines, risking your rights both in the UK and overseas. This could mean your patent application is withdrawn or abandoned, meaning it’s never granted. 

Things go wrong

We all make mistakes - even pedantic patent attorneys. Fortunately we are set up to rescue those deadlines that go awry. Our systems are designed to make sure we have the best chance of meeting the requirements for reinstatement or restoration if a deadline is missed. We also have to carry mandatory PI insurance to be regulated.

If your AI messes up, the buck stops with you. I doubt any tribunal will be convinced that putting the key rights that support your business in the hands of an AI meets the “all due care” standard required.

Should you use AI for any part of your patent application? 

AI tools do bring a lot of benefits in terms of research, prior art analysis, exploring alternatives, efficiency in writing, brainstorming, checking facts and testing how robust a patent application really is. Using them effectively to create a patent application involves putting the right guardrails and human oversight in place. 

At Grey Wolf, we use high level paid-for AI tools as part of a blended approach. With strong security protocols, we guide the tool to help us produce quality patent applications efficiently. 

Each word is still reviewed by an attorney to make sure we’re working in line with guidance from the Intellectual Property Regulation Board and the Institute of Professional Representatives before the European Patent Office (EPI). 

We also have extensive professional indemnity insurances and are fully regulated. So if something does go wrong, SMEs know that they are fully protected. In short, our clients can enjoy the benefits of AI tools, without having to worry about potential risks and consequences.