The claims are the invention. Everything else in a patent is background, drawings, and explanation; the claims are the numbered statements that legally define what the patent covers. This page teaches you to read them, with real granted claims pulled from our own USPTO corpus.
Every patent has independent claims and dependent claims. An independent claim stands on its own: it lists every element of the invention. A dependent claim refers back to an earlier claim and narrows it with one extra requirement. To infringe a dependent claim, a product must meet that claim AND everything it depends on.
Here is claim 1 of US06992474B2, a granted feedback-control patent:
INDEPENDENT CLAIM · US06992474B2, CLAIM 1
“A feedback control system, comprising: a flexible rod having an anchored end and a free end, the rod free end movable upon the application of electric potential to the rod; a current source coupled to the rod; a movement sensor disposed on the rod; and a circuit electrically coupled to the movement sensor and to the current source and having an external input.”
Claim 1 is broad on purpose: it requires a movement sensor, not any particular kind. That breadth is what makes an independent claim valuable. Now two of its dependent claims:
DEPENDENT CLAIMS · SAME PATENT
“2. The system of claim 1 wherein the movement sensor is a strain gauge.”
“4. The system of claim 1 wherein the rod comprises a shape memory alloy.”
Each dependent claim inherits all of claim 1 and adds exactly one requirement. Claim 2 covers only the version with a strain gauge; claim 4 only the version with a shape-memory-alloy rod. This is the standard structure: the independent claim stakes out the widest territory the examiner allowed, and the dependent claims build fallback positions inside it. If claim 1 is later narrowed in court, the dependent claims survive as separate, narrower fences.
Dependent claims can depend on other dependent claims, and the requirements stack. Claim 1, 2, and 3 of US07012011B2 (a wafer-manufacturing method):
A CHAIN · US07012011B2
“1. A method comprising: thinning a silicon wafer; processing a chemical vapor deposition diamond (CVDD) wafer; bonding the CVDD wafer to the thinned silicon wafer to form a bonded wafer; plating metallization on back side of the CVDD wafer; and reflowing the CVDD wafer to flatten the back side.”
“2. The method of claim 1 further comprising: singulate the bonded wafer into die.”
“3. The method of claim 2 further comprising: attaching the die to a package substrate; underfilling the die and the package substrate.”
Claim 3 does not say “of claim 1”, but it inherits claim 1 anyway, through claim 2. To read on claim 3, a process must do everything in claim 1, plus dicing into die, plus substrate attach and underfill. Each step down the chain is a narrower fence covering fewer products, which is why the last claim in a chain is the easiest to design around and the first claim is the hardest.
Claims read the same way in software patents. Claim 1 of US12574444B2:
INDEPENDENT CLAIM · US12574444B2
“A system for decoding data, the system comprising: browser code configured for execution on general-purpose processing circuitry; memory circuitry storing a plurality of accessors, wherein at least some of the plurality of accessors comprise platform-independent syntax implementing an algorithm for decoding encoded data; wherein the browser code is configured to access the memory circuitry automatically without user intervention and retrieve one or more of the plurality of accessors in response to retrieving encoded data; wherein the system comprises an operating system, a frame buffer, and a display for displaying decoded data; and wherein the platform independent syntax comprises WebAssembly programming language syntax.”
Every “wherein” clause is another limitation. A product only infringes this claim if it has ALL of them, including the WebAssembly requirement in the final clause. This is what a design-around looks for: the one clause your architecture genuinely does not practice. That is exactly what the design-distance check scores, limitation by limitation.
Scope is the territory a claim covers. The movement-sensor claim above covers any feedback system with a rod, a current source, and a sensor, whatever brand of sensor: a competitor using a piezoelectric sensor escapes claim 4 but still reads claim 1. Broader claims are worth more and are easier to infringe by accident; narrower claims grant more safely and survive examination more easily. When you screen a design, you are asking one question per claim: does my design practice every limitation? The free reverse search shows you the closest granted claims to your design in seconds.
Engineering screening aid, not legal advice. Claim construction is a legal question; for anything you rely on, talk to a patent attorney.