Quality & Compliance6 min read
Built for Scrutiny
By Ian Schick, PhD, Esq.

A patent’s value is how well it performs under scrutiny, across its whole life, from every reader who will ever have a reason to look hard. Design for all of them.
The most consequential reader of a patent application usually arrives years after it was filed. It is a defendant’s litigation team, or an IPR petitioner, with a serious budget and one job: find the sentence that ends the case.
That reader is one of many. Over its life, an application is read by the inventor, the firm, the examiner, investors, licensees, courts and challengers. Each comes with a different job, and each tests the same fixed text.
Most applications are drafted for one of those readers: the examiner. That is understandable. Allowance is the first gate, and the only one with a docket date attached. But allowance is not the measure of a patent. The measure is performance under scrutiny, by everyone who will ever scrutinize it, for as long as it matters.
That turns drafting into a design problem. Name the readers. Name their jobs. Then build the document to perform for all of them.
Nine readers, nine jobs
An application meets at least nine kinds of scrutiny over its life. Each reader asks a different question of the same text.
| Reader | When | Their job | What the document must give them |
|---|---|---|---|
| Inventor and client | Before filing | Confirm it captures what they built and what matters commercially | The invention in their terms, tied to the product it protects |
| Firm reviewer | Before filing | Catch the errors before anyone else does | Consistent terms, antecedent basis, support for every claim |
| Examiner | Prosecution | Search the art and justify allowance or rejection | A clear technical problem and solution; claims that map to the specification |
| Foreign examiner | Prosecution abroad | Test every amendment against the application as filed | Fallback positions disclosed directly, not left to inference |
| Investor or acquirer | Diligence | Decide what it covers and what it is worth | Scope a non-specialist can follow in minutes |
| Licensee | Negotiation | Decide whether they practice it, and whether they can design around it | Claims that read onto real products |
| Court | Enforcement | Construe the claims; decide infringement and damages | Defined terms, a single infringing actor, claims at the unit that carries value |
| Challenger | Invalidity and IPR | Find the weakest claim and the best reference against it | Fallback claims and support that survive the best art |
| Your future self | Continuations | Claim what the market turned into | A specification deep enough to support new claims |
Most of these readers arrive years after filing, with more money at stake and more time to look. Several are adversaries. And every one of them reads the specification as filed: later arguments and amendments can only work with what it already says.
The jobs pull against each other
If every reader wanted the same thing, drafting would be easy. They don’t.
The licensee and the investor reward breadth. The challenger punishes it: the Supreme Court held in Amgen v. Sanofi that “the more a party claims, the broader the monopoly it demands, the more it must enable.” The diligence reader wants plain language. The court wants precision, because a claim that fails to inform skilled artisans “with reasonable certainty” about its scope is invalid under Nautilus v. Biosig.
The inventor wants the specification to champion the invention. That instinct is dangerous. Sitting en banc in Phillips v. AWH, the Federal Circuit reaffirmed that the specification is “the single best guide to the meaning of a disputed term.” In SciMed v. Advanced Cardiovascular, the specification called one lumen structure the basis for “all embodiments of the present invention,” and the court held the claims to that structure. A sentence written to explain the invention to one reader became a limit enforced by another.
The European examiner enforces a stricter version of a rule US examiners also apply. An amendment must stay within what a skilled person would derive “directly and unambiguously” from the application as filed, the “gold standard” of G 2/10. A fallback that was obvious to the drafter but never written down may not be available at all.
No single sentence can serve all of these readers. That is the design problem.
Six design principles
You can’t write one sentence for nine readers. You can write a document whose parts are assigned to them.
- Design for the hostile reader first. The challenger has the most time, the most money and the strongest motive. A document that survives that reader satisfies the examiner almost by default. The reverse is not true.
- Divide the labor. Tier the claims: a broad claim for licensing and diligence, mid-scope claims a court can construe and a jury can follow, narrow claims built to survive the best art. Layer the specification: a plain statement of the technical problem and its technical solution up front, then the detail that courts and foreign examiners will test. That same statement carries the eligibility argument in the US and the problem-solution analysis at the EPO.
- Write the fallbacks down, with reasons. Describe every alternative and every narrowing feature you might retreat to, and state the technical advantage each one delivers. That one habit supplies amendment support abroad, an answer to obviousness and raw material for continuations.
- Define; don’t celebrate. Under Phillips, where the patentee gives a term a special definition, “the inventor’s lexicography governs.” Use that power deliberately. Keep “the present invention is,” “critical,” “essential” and “must” out of the specification unless you mean them as limits.
- Claim the defendant. The en banc Federal Circuit held in Akamai v. Limelight that direct infringement of a method occurs where all steps are performed by or attributable to a single entity. Write claims from the perspective of the party you would actually sue, reciting what that party does and what an outsider can observe.
- Claim where the value is. In LaserDynamics v. Quanta, the court said royalties should generally be based on the “smallest salable patent-practicing unit,” not the entire product. A claim aimed at a minor component invites a royalty base the size of that component. Include claims at the level of the feature that drives demand.
Why this discipline has been rationed
None of these principles is new. Experienced litigators have preached them for decades. What has been missing is the time to apply all of them to every application.
By my company’s reading of the AIPLA survey medians, a relatively complex application has been priced at roughly 26 attorney hours, a figure that barely moved from 2014 to 2024. Those hours go first to the basics: a coherent specification and a claim set the examiner can allow. Second and third claim tiers, written-out fallbacks and a full read as a defendant would read it are the first things a fixed budget cuts.
AI-assisted drafting changes that arithmetic. Tiered claim sets, alternative embodiments and stated advantages become cheap to generate. Terminology, antecedent basis and claim support become cheap to check, line by line. The practitioner’s time moves from producing text to judging it, which is where senior review bandwidth belongs.
There is a catch. A model trained on decades of issued patents will reproduce their habits, including the ones that limit claims: “the present invention” framing, confident descriptions of one embodiment as the invention. A drafting system has to be designed against those habits, reader by reader. Speed alone just produces the old weaknesses faster.
The question to ask of every draft
Allowance tells you one reader was satisfied. It says nothing about the eight who come after.
The better test runs the whole life of the patent. Before filing, walk the draft past every reader in the table and ask the same three things. Who will read this? What are they trying to do? Does this document do it for them?
A patent that answers yes for all nine was not lucky. It was designed.
Sources
- Amgen Inc. v. Sanofi, 598 U.S. 594 (2023)
- Nautilus, Inc. v. Biosig Instruments, Inc., 572 U.S. 898 (2014)
- Phillips v. AWH Corp., 415 F.3d 1303 (Fed. Cir. 2005) (en banc)
- SciMed Life Systems, Inc. v. Advanced Cardiovascular Systems, Inc., 242 F.3d 1337 (Fed. Cir. 2001)
- Akamai Technologies, Inc. v. Limelight Networks, Inc., 797 F.3d 1020 (Fed. Cir. 2015) (en banc)
- LaserDynamics, Inc. v. Quanta Computer, Inc., 694 F.3d 51 (Fed. Cir. 2012)
- EPO Case Law of the Boards of Appeal, II.E.1.3.1, the “gold standard” of G 2/10
- Paximal, The Economics of Patent Drafting



