How to draft patent claims for a solo inventor (2026)

The claims at the end of your patent application are the only part that matter legally. Everything else — the specification, the figures, the abstract — exists to support them. The claims define the exact scope of the right you're asking the USPTO to grant, and the difference between a defensible claim and a worthless one often comes down to ten words. This post is a practical playbook for solo inventors writing their first set of claims.

What a claim actually does

A patent claim is a single sentence that defines, with legal precision, what your invention is. When someone infringes, the court compares their product to your claims word-by-word. Every element in your claim must be present in the accused product for infringement to exist. This is why broader claims are more valuable — they cover more potential infringers — but also more likely to be rejected at examination or invalidated later by prior art.

Claims come in two structural flavours that sit at the end of your application:

• Independent claims stand alone. They define the invention in full with no reliance on any other claim. Most applications have one or two independent claims.
• Dependent claims reference an earlier claim and narrow it with additional features (“The method of claim 1, wherein X…”). They're cheaper fallbacks if the independent claim gets rejected or invalidated.

The anatomy of one claim

Every well-formed claim has three parts:

1. Preamble — the grammatical subject. “A method for…”, “A system comprising…”, “An apparatus for…”. The preamble establishes whether the claim covers a method (a process someone performs) or an apparatus / system (a physical or software structure). Don't mix the two in one claim — it's called a “mixed-class claim” and it's a common §112 indefiniteness rejection. You can (and should) have separate method and apparatus claims for the same invention.
2. Transition — one of three specific legal words:
   • comprising — open-ended. The claim covers anything that has these elements, regardless of what else it has. Use this unless you have a specific reason not to.
   • consisting of — closed. The claim covers only things that have exactly these elements and nothing else. Rare — mostly used in chemistry.
   • consisting essentially of — semi-closed. Covers these elements plus anything that doesn't materially change the invention. Also rare.
3. Elements (or limitations) — the substantive list of features that make up the invention. Each element typically gets its own numbered sub-part in the claim, separated by semicolons.

Writing your first independent claim: a five-step process

1. Pick your claim type. Software / business methods → method claim. Physical device → apparatus claim. Most software inventions actually deserve both, filed as separate independent claims.
2. List the minimum viable elements. Start with your disclosure in front of you. Write down every feature your invention has. Now aggressively cut: which ones could a competitor design around? Which ones are table stakes for the category? Keep only the features that, together, define what makes your invention novel.
3. Order them by dependency. Elements that reference each other (“the widget of step (a) is…”) need to be introduced in the correct order. Introduce each new noun with “a” on first mention and “the” thereafter. This is the antecedent-basis rule (§112).
4. Pick each element's abstraction level carefully. “A cryptographic hash function” is broader than “SHA-256”. Go broad by default, but not so broad that you walk into prior art. The prior-art playbook informs this trade-off: if SHA-256 is unique to your disclosure but SHA-2 family is already claimed elsewhere, say “cryptographic hash function” — broader than SHA-256, narrower than any hash.
5. Read it out loud. Claims are one sentence. If you have to gasp for air mid-sentence, rewrite. Clear claims are easy for examiners to allow; tangled claims invite §112 rejections.

A worked example

Say you've invented a novel authentication flow. A first-pass independent claim:

1. A method for authenticating a user to a remote service, the method comprising: receiving, from a client device, a biometric input; computing a local hash of the biometric input using a cryptographic hash function; transmitting the local hash, but not the biometric input, to the remote service; and granting access to the user upon a match between the local hash and a previously-stored reference hash.

Elements: receive → compute → transmit → grant. Preamble: “A method for authenticating…” (method claim). Transition: “comprising” (open-ended). Every noun is introduced once and referenced with “the” thereafter.

A dependent claim narrowing it to a specific hash:

2. The method of claim 1, wherein the cryptographic hash function is SHA-256.

Now Claim 1 is your broad coverage; Claim 2 is a narrower fallback if Claim 1 gets rejected over prior art that uses, say, BLAKE3 — Claim 2 still covers SHA-256 users.

Dependent-claim patterns that add defensive value

• Specific-implementation narrowing: “wherein the hash function is SHA-256”, “wherein the communication protocol is TLS 1.3”
• Parameter-range narrowing: “wherein the key size is between 128 and 512 bits”
• Optional-step add-on: “further comprising encrypting the hash before transmission”
• Alternative-embodiment claim: “wherein the biometric input is a fingerprint” — paired with another dependent claim for face recognition, etc.

Rule of thumb: 1 independent claim + 8–15 dependents is about right for a provisional. More than 20 claims total triggers USPTO excess-claims fees at examination.

The five mistakes we see most

• Using “may” or “preferably”. Claims are declarative. “Wherein the hash may be SHA-256” is indefinite — does it claim SHA-256 or not? Write “wherein the hash is SHA-256” in dependent claims.
• Using “approximately” without an anchor. “approximately 1 micrometer” fails §112 (definiteness) unless the specification defines what approximate means. Put the range in: “between 0.9 and 1.1 micrometers”.
• Missing antecedent basis. Claim 3 references “the widget” but no earlier claim introduced one. Instant §112 rejection. Always re-introduce nouns explicitly or trace them back through claim dependency.
• Functional claiming without structure. “A means for authenticating” without a clear structural description in the spec falls into means-plus-function territory (§112(f)) and gets interpreted narrowly as only the structure actually disclosed. Either avoid “means for…” or accept the narrow interpretation.
• Mixed-class claims. “A system comprising: a processor; and a method performed by the processor” — §112 rejection. Split into separate system and method claims.

When to bring in a practitioner

For a provisional, DIY claims that avoid the five mistakes above are fileable. For a non-provisional, practitioner review is the difference between a $3,000 application that issues in 2 years and a $3,000 application that stacks 4 rounds of office actions and ends in final rejection. One hour of a registered practitioner's time ($250–$500) specifically on claim-scope review is the highest-leverage spend in the whole patent process — see the cost breakdown for the all-in math.

If you're using 50search to generate a provisional draft, our output includes 1 independent + 8–15 dependent claims by default, with an adversarial review step that flags claim-scope and antecedent-basis issues before delivery. The claims are still worth a practitioner eye before you file, but you're starting from a disciplined first pass instead of a blank page.