Start a conversation
Menu
A hiker looks up a mountain path whose stone steps are labelled §101 abstract, §102 already disclosed, §103 obvious and §112 unclear, leading to a flag marked patented. Text: The cheapest part of a patent is the part founders budget for. Patent hurdles are predictable. Plan for the work beyond filing.
11 min read

The Cheapest Part of a Patent Is the Part Founders Budget For

Rejection is the default at the USPTO. The four kinds you'll face are predictable, and so is most of the fix.

For a large entity, the USPTO's basic filing, search, and examination fees on a utility application total $2,000. Small entities pay less.

That's the number most founders put in the plan. It's also the smallest number in the story.

How many times you'll be rejected

Rejection is the default at the USPTO. About 90 percent of applications are rejected at least once, about 60 percent reach a final rejection and about 20 percent go to appeal or RCE. Three stacks of paper show 1 round (about 10 percent allowed on first action), 2 to 3 rounds (most applications) and 4 or more rounds (not uncommon for complex technologies).

Prosecution runs in rounds, and most applicants go through more than one.

Rejected at least once: nearly everyone. Dennis Crouch at Patently-O looked at 1.5 million published utility applications filed from 2015 through 2018. First-action allowances ran about 10 percent. Recent first actions sit closer to 14 percent. Either way, roughly nine in ten applicants hear "no" first.

Rejected a second time: roughly two in five. About 42 percent of those same applications received a final rejection, which is usually the examiner's second "no."

Rejected three or more times: common in computing. After a final rejection, the usual way forward is a request for continued examination (RCE), which reopens the file and restarts the cycle. Each RCE tends to bring about two more office actions. In a Fenwick study of patents filed since 2000, the computer technology center had the lowest share of patents issued with no RCE at all, about 58 percent. More than four in ten needed at least one extra round after a final rejection.

And yet 76 percent of those 1.5 million applications became patents. Rejection is the normal route to a patent. What it costs you is rounds.

The fields we work in carry extra weight.

  • AI: among applications first examined in 2025, 42 percent of AI-related ones have drawn a §101 eligibility rejection, against 6.9 percent of everything else. Obviousness rejections run near 80 percent for both.
  • Cryptography and security: in USPTO Art Unit 2436, the average first action arrives around 27 months after filing, and final disposition around 44 months.

Each rejection starts the same cycle. Your attorney or agent reads the office action, decides how to answer, amends or argues, and files a response. Then you wait. The first rejection is close to certain. How many follow it is the part you can influence before you file.

What a round actually costs

Government fees at current USPTO large-entity rates:

  • Request for continued examination: $1,500 the first time, $2,860 for each one after.
  • Extension of time to respond: $235 for one month, rising to $3,395 for five.
  • Appeal: $905 for the notice, plus $2,535 to forward the appeal to the Board.
The fee ladder gets steeper after filing. Large-entity USPTO rates cited in the article: $2,000 basic filing, search and examination; $1,500 first RCE; $2,860 each later RCE; $235 to $3,395 for a 1 to 5 month extension; $905 plus $2,535 for an appeal notice and forwarding to the Board.

Professional time sits on top. In a July 2026 Federal Register notice, the USPTO cited AIPLA's 2025 Economic Survey for an average IP billing rate of $550 an hour across all firms.

A rough illustration, using only those figures: one continued examination request plus one eight-hour day of practitioner time comes to about $5,900. That's one round, before any extension, interview, or second response. Complex applications rarely stop at one.

A worked example: one prosecution round can reach about $5,900. A $1,500 first RCE fee plus $4,400 of practitioner time (8 hours at $550 an hour) equals about $5,900, before any extension, interview or second response.

Time has a price too. As of April 6, 2026, the USPTO's backlog of unexamined applications was 776,995, down from a peak of 837,928 in January 2025. Director Squires, announcing the drop, pointed to studies suggesting a one-week reduction in pendency adds roughly $50,000 to a US company's value on average. Whatever the right figure is for your company, every extra round moves your issue date, and investors, acquirers, and competitors are watching it.

Time has a price too. Two bars compare the USPTO backlog: 837,928 unexamined applications at the January 2025 peak and 776,995 on April 6, 2026, a drop of 60,933. A callout notes an average company-value effect of about $50,000 per week of less pendency, from studies cited by Director Squires.

The four rejections

Rejections are not random. Juristat reviewed more than 3.4 million of them, on applications filed from 2014 to 2018, and found that §102, §103, and §112(b) alone made up more than 70 percent. Add §101 and you have the four that matter.

  1. Too abstract to patent (§101). The examiner says the claim is directed to an abstract idea, a law of nature, or a natural phenomenon, without enough beyond it. Software and AI claims are especially exposed. The usual cause is a claim that states what the system achieves rather than the specific technical way it achieves it.
  2. Already covered (§102). A single earlier document, whether a patent, a paper, or a product manual, discloses everything the claim recites. This is usually a search problem. The reference could have been found before filing, and nobody found it.
  3. An obvious next step (§103). No one reference shows the invention, but the examiner combines two or three and argues a skilled engineer would have put them together. It is the most common rejection and the hardest to argue past, because obviousness is a judgment call and the examiner makes it first.
  4. Unclear claim language (§112). Terms in the claim lack clear boundaries, or the claim reaches past what the application describes. These tend to come from drafting under deadline pressure, or from broadening claims late without revisiting the specification.

One office action often cites several at once. A 2017 USPTO working paper covering millions of office actions found an average of 1.8 rejection types per action. Expect rejections. The real planning question is which ones, and whether you saw them first.

The four rejections. Sections 102, 103 and 112(b) made up more than 70 percent of 3.4 million rejections reviewed by Juristat. §101 too abstract: a result without enough technical mechanism. §102 already covered: one earlier reference covers the claim. §103 obvious next step: the examiner combines two or three references. §112 unclear language: claim boundaries or support are insufficient. Average rejection types per office action: 1.8.

Four habits behind the extra rounds

Broad by default. Claims go in wide and get narrowed under pressure, one office action at a time. Each cut is a costly way to learn what a better search would have shown up front.

A thin search. The pre-filing search was light, or skipped because the team "knows the space." Examiners search patents, published applications, and technical literature, and they pull from adjacent fields. A competitor map covers far less ground than that.

A spec written for the product. The specification describes what you built rather than supporting the claims. When claims need to move during prosecution, there is nothing in the spec for them to move to.

No one argued the other side. The first serious attack on the claims came from the examiner.

Most avoidable prosecution cost starts before filing. Four numbered causes climb a mountain: 1, broad by default, claims go in wide and are narrowed by office action; 2, a thin search that misses the examiner's adjacent art; 3, a product-first specification without enough fallback claim positions; 4, no one argued the other side, so the examiner becomes the first serious adversarial reviewer.

The broad versus narrow trade

Broad claims are worth more. They reach design-arounds, future versions, and competitors who took a slightly different route. They are also the likeliest to be rejected.

Narrow claims get allowed, and can be close to worthless. A patent covering only your exact implementation invites a competitor to change one detail and walk around it.

The target is the broadest claim that survives examination. That call depends on the invention and the art unit, and no formula settles it. What you can choose is when to make it: before filing, with evidence, or one office action at a time.

The broad versus narrow trade. A route climbs a mountain to a summit marked survivable breadth. Narrow claims are easier to allow, with more limited coverage, less value and lower risk of rejection. Broad claims have greater commercial value, cover design-arounds and are more future-proof, but are harder to get and more vulnerable early on. The target is the broadest claim that survives examination.

The content of that first rejection matters more than its arrival. Crouch found that nearly 40 percent of first actions either approve the application or flag something as already allowable. Then he followed about 382,000 applications that got their first examiner review in 2023 and checked their status three and a half years later.

  • Approved on the first review: 98 percent had been granted.
  • Rejected, but some claims flagged as allowable: 93 percent granted.
  • Rejected, nothing flagged as allowable: 68 percent granted.

An approval still needs the issue fee paid and the grant completed, which is why even the top group falls just short of 100.

A first rejection with a road map is nearly as good as an approval. One without a road map costs 25 points of odds.

That gap is the case for filing claims that give the examiner something clearly allowable, while still covering ground worth owning.

Red-team your claims before the examiner does

That's the work Kenorai does. Nobody can promise to prevent the first rejection. What we can influence is which kind you get: one that already concedes allowable ground, with fewer surprises behind it. We do it by running the examiner's process on your claims before you file. For AI, blockchain, and IoT security inventions, that starts with §101, tying each claim to a specific technical mechanism, because these fields draw eligibility rejections far more often than other technologies.

Search like an examiner. Go beyond competitor products. Cover patents, published applications, and technical literature, including the adjacent fields an examiner is likely to combine.

Attack each claim position on all four grounds. Is it tied to a concrete technical mechanism, or could it read as an abstract result? Does any single reference already cover it? Which two or three references would an examiner combine, and what is your answer to that combination? Is every term clear, and does the application support every limitation?

Rank what survives. Some positions will collapse. Some will survive only narrowly. A few will hold at meaningful breadth. Score each on two axes, likelihood of surviving examination and commercial ground covered, then lead with the positions that do well on both. Know your fallbacks before you need them.

Hand your practitioner a better starting point. The deliverable here is a sharper set of claim positions, the prior art that shaped them, and counterarguments already worked through. Drafting and filing still belong to your practitioner, who now starts ahead.

This will not guarantee an allowance. Examiners differ, and obviousness always involves judgment. It removes the rounds you could have predicted, so the ones left are about genuinely hard questions.

Where your patent attorney or agent fits

Preparing and prosecuting an application on a company's behalf is the job of a registered patent practitioner, and it stays that way. Red-team work improves what they receive.

Incentives are worth a moment. Prosecution is commonly billed per application and per response, and most practitioners do excellent work however they bill. A strategy view from someone with no stake in how many applications you file, or how many rounds they take, still gives you a second opinion the billing model has not touched.

Before your next filing

Answer these three before you spend on an application:

  1. What will an examiner cite against this, and have we searched for it the way they will?
  2. What's the broadest version of this claim we believe survives, and why?
  3. If our lead position is rejected, what's the fallback, and is it still worth owning?
Three questions that expose the expensive gaps early, on three cards: 01, what will the examiner cite, and have we searched the way they will? 02, what is our broadest survivable claim, and what evidence supports it? 03, what is the fallback, and is it still worth owning if the lead position is rejected?

Clear answers to all three remove some of the back and forth before it begins. If any of them is fuzzy, start there.

We have expanded these into a six-question pre-filing checklist you can work through with your team: The Pre-Filing Red-Team Checklist.

If this was useful, follow Kenorai for the rest of the series on patent strategy for deep tech founders.

About Kenorai

Kenorai is a patent strategy consultancy for deep tech companies, founded by Jan Gerards, Managing Partner, and Seth Everson, Co-Founder and Principal Technical Partner. Together, our founders bring 16 years of patent experience.

We work alongside any patent attorney or agent you choose. In certain technology areas, our founders also have a 12-year working relationship with a preferred patent agent we can bring in when it fits. That relationship is non-exclusive, and we receive no referral fees or other compensation from any patent agent or firm.

We don't file patents. Kenorai is not a law firm and does not provide legal advice.

Sources

Updated September 16, 2026: added data on repeat rejections and AI eligibility rejections, and clarified the first-action outcome figures.

Kenorai - Patent strategy for deep tech companies - Kenorai is not a law firm and does not provide legal advice.

Kenorai is not a law firm. No Kenorai personnel are licensed attorneys or registered patent agents. We do not practice law, do not practice before the USPTO, do not give legal opinions, and do not make filing or claim decisions. We provide technical and business consulting that helps companies prepare for their work with a licensed practitioner. Licensed practitioners are engaged for each matter, and they make every filing decision and give all legal advice.

No comments yet

Continue reading