Foundations

How to read a patent claim

A patent has a 50-page specification, a fancy abstract, and a row of figures. Lawyers and courts look at none of that. They look at the claims.

5 min read · Updated Apr 29, 2026

A patent has a 50-page specification, a fancy abstract, and a row of figures. Lawyers and courts look at none of that. They look at the claims.

Claims are numbered sentences at the end of the patent that define exactly what the patent covers. If a competitor's product is described by a claim, the competitor infringes. If the product isn't, they don't. Everything else is context.

If you're a defendant, the first thing your attorney will do is read the claims and decide whether the asserted product fits within them. If you're trying to invalidate the patent, you're attacking the claims. If you're confused, you're probably reading the wrong part of the patent.

The structure of a claim

A claim has three parts:

  1. Preamble — a short phrase setting up what the claim covers ("A method for transmitting data over a network, comprising:"). Whether the preamble is "limiting" — i.e., whether it adds requirements — depends on context. Often litigated.
  2. Transition — a single word or phrase. The transition word matters enormously:
    • "comprising" — open-ended. The claim covers anything that has all these elements plus possibly more.
    • "consisting of" — closed. The claim covers only the listed elements; if a product has anything else, it doesn't infringe.
    • "consisting essentially of" — middle ground; the product can have additional elements only if they don't materially affect the basic invention.
  3. Body — the listed elements, called limitations. Each limitation describes a feature that must be present.

A toy example:

  1. A method for routing data, comprising:
    receiving a request at a first server;
    parsing a destination identifier from the request;
    transmitting the request to a second server identified by the destination identifier.

To infringe this claim, a system must do all three steps — receive, parse, transmit. Missing any one and there's no infringement.

Independent vs. dependent claims

  • An independent claim stands alone. Claim 1 is almost always independent.
  • A dependent claim references an independent claim and adds limitations: "The method of claim 1, wherein the destination identifier is a URL." Dependent claims are narrower than the claim they depend on.

When defending, you analyze each independent claim separately. Each independent claim is a different lawsuit, basically. Dependent claims add narrowing — if the independent claim is invalid, the dependents might still survive, but usually they fall together.

What infringement requires

For literal infringement of a claim, the accused product must have every limitation. Not most. Not the gist. Every one.

If a claim recites elements A, B, C, D, and the product has A, B, C, but not D, no infringement.

The doctrine of equivalents is a fallback — a product can infringe if its differences are "insubstantial," but courts apply this narrowly and most defenses focus on the literal claim.

This is why claim drafters fight over every word. A claim that says "a fastener" covers screws, nails, and rivets. A claim that says "a screw" covers only screws.

What invalidity requires

To invalidate a claim, you have to find a single piece of prior art that discloses every limitation (anticipation under § 102), or a combination of prior art references that together disclose every limitation with motivation to combine (obviousness under § 103).

The mirror image of infringement: missing a single limitation kills the prior-art argument too. Defense attorneys spend weeks building claim charts — spreadsheets that map each claim limitation to specific text in the prior art. If every limitation has a hit, the claim is in trouble.

Magic words to watch for

Patent claims are written in a particular dialect. A few phrases worth recognizing:

  • "means for [X]" — a means-plus-function limitation under § 112(f). Covers only the structures actually described in the specification (and equivalents). Often narrower than it looks.
  • "a plurality of" — two or more.
  • "substantially" / "about" / "approximately" — terms of degree. Often vulnerable to indefiniteness challenges under § 112.
  • "adapted to" / "configured to" — these often imply specific structure rather than mere capability.
  • "wherein" — adds a requirement on a previously-named element.

How to find the weaknesses

A defendant analyzing a claim looks for:

  • An overbroad limitation. Language so broad that it reads on prior art the patentee didn't disclose. Often the strongest invalidity argument.
  • A specific limitation absent from the accused product. Easy non-infringement.
  • Means-plus-function language. Limits the claim to structures in the specification, often missing from modern products.
  • Indefinite terms. § 112 challenges based on subjective or unclear language.
  • Generic-computer claims. Software patents that recite "a computer" performing routine steps are particularly vulnerable to § 101 challenges under Alice, often before discovery even starts.

A patent attorney's review of a claim takes hours. Most claims have weaknesses. The trolls' game is to assert claims that look broad in a demand letter and hope you settle before anyone reads them carefully. (See a worked summary of the claims in US Patent 8,069,073, the running example.)

Bottom line

Claims are everything. Specifications are background. If you're facing a patent assertion, start with the claims, read them slowly, and identify which limitations are present in the accused product and which are missing. Most patent assertions don't survive a careful read.

This article is for general education and is not legal advice.