Alex is the co-founder and CEO of Patentext. He’s spent over a decade drafting patents for startups, unicorns like Uber and Dropbox, and everything in between. When he’s not obsessing over Patentext or running his climate tech-focused IP firm, he’s likely training for a triathlon or chasing a very fast border collie.
When a startup founder tells me they've "filed a patent," I usually ask what their broadest patent claim covers. Nine times out of ten, they don't know. They may remember the technical details they explained in the specification or the drawings they spent hours perfecting, but they can't tell me the actual scope of protection they're seeking.
That's a problem, because everything else in a patent application is context. The claims are the only part that defines what you own, what competitors can't do, and what savvy investors actually evaluate when they assess the value of your patent portfolio.
Read on as we explore what patent claims are, how they work in practice, and why startups should think about them strategically rather than just technically.
What are patent claims?
Patent claims are the numbered sentences at the end of your patent application that define the legal boundaries of your invention. Everything else in the patent (including the background section, the detailed description, the drawings) exists to support the claims. The claims are what you enforce, what competitors design around, and what determines whether your patent is worth the paper it's printed on.
The part that often surprises most founders the most is that you don't “own” your invention as you've built it; you own what your claims define, and nothing more. If your product has ten features but your claims only cover three of them, you have no patent protection for the other seven. If your competitor builds something that falls outside your claim language but accomplishes the same thing your product does, they're not infringing, no matter how similar the underlying concept is.
The anatomy of a patent claim
Claims are written in precise legal language (typically single-sentence structures that can run 10+ lines with nested clauses) because they're legal instruments designed to be interpreted by patent examiners and judges, not engineering documentation for your team.
So, let's break down what claims actually look like. Here's a simplified example for a hypothetical AI-powered scheduling system:
Claim 1: A computer-implemented method for scheduling meetings, comprising:
- receiving availability data from a plurality of users;
- analyzing the availability data using a machine learning model trained on historical meeting outcomes;
- generating a proposed meeting time based on the analysis; and
- transmitting the proposed meeting time to the plurality of users.
This is an independent claim using the transitional word "comprising." Let me unpack what each piece means.
The preamble: "A computer-implemented method for scheduling meetings"
In most cases, the preamble tells you whether you're claiming a method (steps/process), a system (physical or logical components), or a composition of matter, and gives context for what follows. Most software/AI patents include both types of claims because different infringement scenarios call for different claim structures.
In many cases, the preamble isn't limiting; it doesn't add substantive requirements to the claim, but it establishes the frame for how the rest of the claim should be read. However, if the other limitations in the claim reference items in the preamble it can be limiting, so preambles are usually kept relatively short.
The transition: "comprising"
This one word has enormous strategic significance. "Comprising" means the claim covers systems that include at least the listed elements, but can have additional features. A competitor who builds a system with your four elements plus three more is still infringing.
The alternative transition is "consisting of," which means the claim covers only systems with exactly those elements and nothing more. This is much narrower and rarely used except in very specific contexts like chemical formulas. There's also "consisting essentially of," which is somewhere in between, but you'll almost always see "comprising" in software and hardware patents.
Most other transitional phrases like “including,” “having,” “defining,” and “configured to” generally take on their plain dictionary meaning or one commonly utilized in the technical field unless a more specific definition is provided in the detailed description of the patent application.
The body: The claim elements or “limitations”
These are the individual features or steps that define what you're protecting. Each element is a requirement; an infringing product must have all of them. Miss one and there's no infringement, no matter how similar the overall system is.
This is where claim strategy gets interesting. Broad claims tend to have fewer elements, which means it's more likely that a competitor's product will satisfy all the requirements. Narrow claims have more elements, making infringement less likely but also giving you protection in the specific territory you've defined.
Dependent claims: Narrowing the scope
After your independent claims, you'll see dependent claims that reference back to earlier claims and add additional limitations:
Claim 2: The method of claim 1, wherein the machine learning model is a neural network trained on at least 10,000 historical meetings.
Claim 3: The method of claim 1, wherein the availability data includes calendar entries, email response times, and time zone preferences.
Dependent claims serve three strategic purposes:
- First, they give you fallback positions if your broader independent claims get rejected during examination or invalidated during litigation. If claim 1 turns out to be too broad relative to the prior art, you might still get claim 2 or 3 through.
- Second, they can provide examples that give definitional meanings to the independent claims from which they depend. For example, Claim 3 above shows that the term “availability data” is broad enough to include calendar entries, email response times, and time zones.
- Third, dependent claims can be useful as a way to directly communicate with an examiner through functional limitations that may not be useful as a fall back position, but help the Examiner or a court understand the purpose of the invention.
The most important aspect of dependent claims to remember as a startup founder is that dependent claims are your hedge against uncertainty. You don't know exactly how your product will evolve or how competitors will respond, so you want claims at multiple levels of abstraction.
Independent vs. dependent claims
Most US patent applications include 20 claims total, structured as 3 or fewer independent claims with multiple dependent claims hanging off each one. The independent claims are your broadest assertions; they stand alone and define the core territory you're claiming. The dependent claims add specificity and create a defensive depth to your portfolio.
Generally, independent claims are more likely to be rejected during examination because they're broader and therefore more likely to overlap with prior art. But they're also the claims that matter most for enforcement. A broad independent claim that holds up under examination can cover a wide range of competitor implementations.
Dependent claims are easier to get through examination because they're narrower, but they're also easier to design around. A competitor who wants to avoid your patent can often engineer around one or two specific limitations in your dependent claims. They can't as easily engineer around a well-drafted independent claim that captures the fundamental innovation.
For startups, you want balance: independent claims broad enough to be commercially useful, with dependent claims at multiple levels of abstraction. That way if you aren’t able to achieve the broadest claim you were hoping for, you have multiple valuable options to fall back to.
If your claim set only includes broad independents and narrow dependents, you could describe it as a “brittle” claim set, which is essentially worthless as soon as the independent claim falls. These complexities are why experienced patent practitioners often spend as much time on claim strategy as they do on writing the specification.
Common patent claim types
The claim types in your claim set should target the infringer you’re actually interested in targeting in a way that is observable. If you want to target a company that sells heat pumps, then you should have a system claim on the heat pump, not a method claim on the installation method.
Similarly, when your invention includes both physical and process elements, it’s almost always a good idea to include both in your claim set, so that you have options to assert your claims. Even when your invention is process focused (like software), there’s always a physical system executing the method, so it can be beneficial to have both.
Here’s a quick overview of all the common claim types for startups.
Method claims
Method claims cover what someone does with your invention. They're often written as steps: "receiving data," "processing the data," "outputting a result." Method claims are useful when the infringement you're worried about is someone performing the process, like a competitor offering a service that uses your method.
The challenge with method claims is enforcement. To prove infringement, you need to show that someone is actually performing all the claim steps. If some of the steps happen on a user's device and some happen on a server, or if multiple parties are involved in executing the method, it can become difficult to prove that any single entity is infringing or inducing infringement.
System/apparatus claims
System/apparatus claims cover the physical or logical structure that performs your invention. They're written as components: "a processor," "medical device for stabilizing fractures," "a chain-driven vehicle." System claims are useful when the infringement you're worried about is someone making or selling a product that embodies your invention.
System claims are often easier to enforce because you can prove infringement by examining the product itself. If a competitor sells a device that includes all the claimed components, that's infringement, regardless of how end users actually use the device.
“Composition of matter” claim
A structurally similar type of claim is a “composition of matter” claim. These claims also cover physical things, but are specific to materials. They will often list ingredients or properties of the material instead of structural components as limitations.
For chemical and biological materials, oftentimes chemical classes, specific formulas, or biological sequences (protein, RNA, DNA) are included in the claim to define a composition.
“CRM” claims
For software inventions, you also might see “CRM” claims, which stands for computer-readable storage medium claims. They usually start with something like “A non-transitory computer-readable storage medium storing instructions that when executed by a processor…”
This is one way of claiming the actual medium that software is stored on, like a CD or a floppy disk. This claim type was more common when software was sold via physical media, and isn’t as useful for cloud-based software.
What makes a "good" patent claim for a startup?
This is the question that matters most, and it's where the gap between legal theory and business reality becomes obvious. Patent practitioners will tell you a good claim is one that's patentable (novel, non-obvious, adequately supported by the specification) and has broad scope. That's true but incomplete. For a startup, a good claim is one that's patentable and commercially useful.
Claims should cover what competitors will build, not just what you've built
The biggest mistake founders make is claiming their exact current implementation. You file a patent covering your v1 architecture, with claims that specify the exact device structure, the exact API calls, the exact ML architecture you're using right now. Two years later, you've re-engineered everything, your competitors have taken a different architectural approach, and your claims are commercially irrelevant.
Good claims are abstracted at least one level up from your implementation:
- They cover the method you're using to solve the problem, not the specific code or API calls you wrote.
- They cover the type of device structure you're using, not the particular hardware configuration you chose.
- They cover the functional characteristics of your ML architecture, not whether it's a transformer vs. an LSTM.
This is a hard balance to strike. Too abstract and your claims become obvious or anticipated by prior art; too specific and they become easy to design around. The right level of abstraction depends on your technology area and competitive landscape.
Claims should be written with licensing in mind, not just litigation
Most startup patents never get litigated. What they do get used for is licensing discussions, acquisition negotiations, and venture due diligence. In those contexts, claim scope is a signal of the quality and value of your IP position.
Investors and acquirers don't just count how many patents you have. They look at what your broadest claims actually cover and whether that coverage maps to defensible market territory. Narrow claims that only protect your exact implementation suggest you have weak IP. Broader claims that cover the general approach to a problem signal stronger protection.
This doesn't mean you should write impossibly broad claims that will never survive examination. But it does mean claim strategy should be informed by how the claims will be read by business stakeholders, not just by examiners.
Claims should survive your product pivots
Startups pivot; your initial product might fail, your go-to-market strategy might change, your target industry might shift. But your patent filing happens at a specific moment in time, and the claims you write then are locked in (subject to amendment during prosecution, but you can’t add entirely new matter).
Smart claim drafting anticipates this. Instead of claiming "a thermal management system for electric vehicle batteries using liquid cooling channels," you claim "a thermal management system for rechargeable battery arrays using fluid circulation to regulate temperature." The first locks you into EVs. The second covers the underlying innovation in a way that remains relevant even if you pivot from automotive to grid storage to consumer electronics.
For founders, this means thinking about claims at the level of the core insight, not the current application. What's the fundamental thing you figured out that others haven't? That's what your broadest claims should target.
5 common claim mistakes that cost startups time and money
Mistake 1: Claiming the entire product instead of the invention
Your product probably does a lot of things, and most of them aren't novel. Your patent claims should cover the specific part that is novel, not every feature in your product.
I see this constantly with hardware startups. They want a patent that covers their entire device: the housing, the power supply, the sensor array, the processing unit, the output display. But only one piece of that system is actually inventive: maybe it's a novel sensor configuration, or a specific signal processing technique. Claiming the whole device just makes the claims longer and easier to design around.
Good claims isolate the inventive concept and claim that, with only as much additional context as necessary to make the claim coherent.
Mistake 2: Using marketing language instead of technical language
Your pitch deck says your product uses "AI-powered insights" or "next-generation optimization." That language is too vague to be used in claims. Claims need to be technically precise, using terms that can be assigned specific definitions with clear bounds.
This is a common problem with founder-drafted patents. The language feels like a product description rather than a technical specification. Claims should read more like dictionary definitions of the invention, meaning precise, bounded terms that can be objectively interpreted, not marketing copy or implementation-specific engineering documentation.
Mistake 3: Drafting claims around your implementation, not the underlying method
Perhaps you're using a specific titanium alloy (Ti-6Al-4V) for your medical implant, a particular polymer blend for your coating, and a commercially available piezoelectric sensor. None of that belongs in your claims (unless there's a very specific reason it does). Claims should abstract away implementation details and focus on the functional requirements.
A claim that specifies "a Ti-6Al-4V titanium alloy implant" is much narrower than a claim that says "a biocompatible metal alloy implant" or just "a biocompatible implant structure." Unless that specific titanium formulation enables something other alloys can't do, you're limiting yourself unnecessarily.
Mistake 4: Including too many elements in independent claims
Every element you add to a claim is a requirement for infringement. The more requirements there are, the easier it is for a competitor to avoid infringement by removing or changing just one element.
Founders often want to pack everything into their independent claims to make them feel comprehensive. But in the claims, less is more. An independent claim with four elements is broader and stronger than the same claim with an additional three elements, assuming both are patentable.
Mistake 5: Not claiming the parts competitors can't avoid
This is the strategic blind spot. You're claiming the parts of your system that you think are clever or novel, but you're not asking: what would a competitor have to do to solve the same problem? What are the unavoidable steps or components?
The best claims cover the narrow waist of your innovation, the parts that any solution to this problem must include. If there are multiple ways to implement your invention, but they all share one common step or structure, that's what your broadest claim should target. This strategy is called “bottlenecking” in the industry, and it can be extremely effective.
What claims mean for your fundraising and acquisition value
Claims are the first thing sophisticated investors look at when evaluating a startup's patent portfolio, and they're often the last thing founders think about when filing. That disconnect creates real problems during due diligence.
Substantive patent due diligence starts with your broadest independent claim
When an investor or acquirer reviews your patent portfolio, their IP counsel will read claim 1 of each patent. They'll evaluate whether it's broad enough to create meaningful barriers to competition, whether it's well-supported by the specification, and whether it's likely to survive examination (if pending) or hold up under challenge (if granted).
If your claim 1 is narrow, implementation-specific, or written in vague marketing language, that's a red flag that signals that either your IP isn't as strong as you've represented, or that you haven't been strategic about how you're protecting it.
This is why claim strategy matters from day one. You don't get to rewrite your independent claims during due diligence. What you filed is what gets evaluated, and if the claims are weak, the entire patent becomes a weaker asset.
Claims define your defensive moat (or lack thereof)
Investors at Series A and beyond increasingly care about whether your IP creates a defensible position. The question isn't just "do you have patents," but "would those patents actually stop a competitor from building something similar?"
The challenge is that broad claims are harder to get granted. They're more likely to be rejected for prior art, and they require more sophisticated drafting to survive examination. This is where working with an experienced practitioner makes a material difference in the value of your IP assets.
Claim amendments during prosecution affect portfolio value
Over 70% of patent applications are amended after the first office action. Amendments often narrow the claims to get around prior art or address examiner objections. That's normal prosecution, but how the narrowing occurs matters.
If your independent claims start out relatively narrow and then get narrowed further during prosecution, you can end up with claims that have very little commercial value. If they start out broader and get narrowed to a defensible middle ground, you've still got commercially useful protection.
Alternatively, if adding a step that is logically needed for the rest of a process to work gets you past the prior art, then the claim is nearly as valuable as it was before, despite technically requiring one more step
Investors and acquirers often review the prosecution history (the back-and-forth between your attorney and the examiner) to understand how much your claims were narrowed and why. Significant narrowing, especially if it was driven by prior art the examiner found that your team missed, is a negative signal about the strength of the patent.
How startup founders should approach claim drafting
Now that you understand what claims are and why they matter, the practical question is: how do you actually get good claims written? There are several paths, each with different trade-offs in cost, quality, and risk.
Draft pseudo-claims to guide your practitioner
Some founders, especially those with technical backgrounds, want to be involved in shaping their claims before engaging an attorney. After all, you know your technology better than anyone.
That said, don’t try to write real patent claims using templates from Google Patents. The language is precise, the conventions are specific, and the strategic choices have downstream consequences that aren't obvious until examination. A founder-drafted claim might look reasonable on the surface but fail in ways you won't discover until you get an office action 18 months later.
Instead, consider drafting "pseudo-claims" in plain language that capture what you think is novel about your invention. Write out the key steps in your process, the essential components of your system, or the functional requirements that competitors would have to replicate. This gives your patent practitioner a clear starting point and ensures the final claims reflect what you actually think matters.
Use a patent attorney or agent (traditional approach)
This is the standard path. You hire a patent attorney or agent, explain your invention, and they draft claims based on their understanding of your technology and the prior art landscape. At a traditional law firm, this is usually bundled with the full application drafting, and the total cost typically runs $12,000 to $20,000+ for a non-provisional patent.
The advantage is quality and experience. A good patent practitioner knows how to structure claims to survive examination, how to balance breadth and specificity, and how to position claims strategically based on your technology area and business model.
The disadvantage is cost and transparency. At $400-1000/hour for attorney time, the claim drafting portion alone can run several thousand dollars, and most founders don't see drafts or have meaningful input until late in the process. You're trusting the attorney's judgment on claim scope, which is fine if you've hired well, but risky if you haven't.
Use Patentext
Patentext is built specifically to solve the claim drafting problem for startups, and handles everything that comes before and after it.
We start with your business context and prior art analysis to understand your space and where the opportunities lie, then guide you through describing your invention in a way that helps us identify the strongest claim positions. Our AI drafting tools generate initial claims at multiple levels of abstraction, and a registered patent agent reviews and refines them to ensure they're strategically positioned for both examination and commercial use.
But we don't stop at claims. We handle the full application (specification, drawings, filing), so your claims are properly supported and your entire patent is built to survive prosecution. You get law-firm-quality work at a fraction of the cost, with full transparency into what you're claiming before you file.
If you're a founder looking to protect your IP without the $15K+ law firm price tag, join our waitlist (and lock in 25% off when we launch).
Disclaimer: This article is for informational purposes only and does not constitute legal advice. Patent laws are complex and vary by jurisdiction. For personalized guidance, consult a qualified patent attorney or agent.