💬 NLP

Unintended Negative Impacts of Promotional Language in Patent Evaluation

This large-scale study of 2.7 million USPTO patent applications reveals that, contrary to its role in science, the use of promotional language in patent filings significantly reduces the likelihood of grant, transfer, and successful appeal, despite objectively reflecting genuine novelty and future impact, with this "promotional penalty" being more pronounced among female and novice examiners.

Original authors: Bingkun Zhao, Chenwei Zhang, Hao Peng

Published 2026-05-07

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Bingkun Zhao, Chenwei Zhang, Hao Peng

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are an inventor with a brilliant, world-changing idea. You write a patent application to protect it. To make your idea sound exciting and important, you use big, enthusiastic words like "groundbreaking," "revolutionary," and "unprecedented." You think, "If I sound confident, the patent examiner will see how amazing this is and say 'Yes!'"

But according to this study, you might be doing the exact opposite.

Here is the story of what the researchers found, explained simply:

1. The "Hype" Backfire

In the world of science grants (where scientists ask for money), using enthusiastic, "promotional" language is like a magic key. It helps reviewers get excited and say "Yes, fund this!"

However, the patent office is a different beast. It's less like a science fair and more like a strict courtroom. The researchers looked at 2.7 million patent applications and found a surprising rule: The more "hype" words you use, the less likely you are to get your patent approved.

The Penalty: If you use a lot of promotional language, your chances of getting a patent drop. Your chances of someone buying your patent later also drop. Even if you get rejected and try to appeal the decision, using hype makes it harder to win the appeal.
The Size of the Problem: The difference is huge. Applications with the least hype were about 5.5% more likely to get approved than those with the most hype. In the world of patents, that is a massive gap.

2. The Great Misunderstanding: "Hype" vs. "Quality"

You might think, "Okay, maybe people who use hype are just trying to hide bad ideas. Maybe they are exaggerating to cover up weak inventions."

The researchers tested this, and it turns out that's not true.

The Reality: The patents with the most "hype" words were actually the most innovative and the most valuable. They combined old ideas in new, surprising ways (high novelty) and were cited by other inventors more often in the future (high impact).
The Paradox: Inventors use these enthusiastic words because their ideas are genuinely exciting and groundbreaking. But patent examiners hear those same words and think, "This sounds like an exaggeration. I need to be skeptical."
The Metaphor: Imagine a chef cooking a perfect, rare dish. They describe it as "the most amazing meal you'll ever eat." A food critic (the examiner) hears this and thinks, "They are just trying to sell me something; the food probably isn't that good." The chef is telling the truth, but the critic hears a sales pitch.

3. Who Falls for the Hype? (And Who Doesn't)

The study also looked at who is doing the rejecting. It turns out the reaction to hype depends on the person reading the application.

Experience Matters: New, inexperienced examiners are the most likely to reject hype. They are like strict students following a rulebook; they see "revolutionary" and think, "That's not allowed; be precise." But experienced examiners are different. They have seen thousands of applications. They can look past the "sales pitch" and see the actual invention underneath. In fact, the most experienced examiners were actually more likely to approve hype-filled applications than the novices.
Gender Differences: The study also found that male examiners were slightly more tolerant of promotional language than female examiners, though both groups generally preferred a more neutral tone.

The Bottom Line

The patent system has a strange blind spot. It is designed to reward the most innovative ideas, but it accidentally punishes the people who are most excited about them.

For Inventors: If you want your patent approved, don't try to sound like a movie trailer. Sound like a mechanic. Be precise, neutral, and stick to the facts. Let the invention speak for itself without the "hype" microphone.
For the System: The patent office is currently rejecting some of its best ideas because the inventors were too enthusiastic. The researchers suggest that examiners need to be trained to realize that "hype" doesn't always mean "fake"—sometimes, it just means "really good."

In short: In the patent office, being too excited can cost you your invention.

Technical Summary: Unintended Negative Impacts of Promotional Language in Patent Evaluation

Problem Statement

Patent examination is a legal-technical adjudication process where examiners determine if an invention meets statutory requirements of novelty, utility, and enablement based almost entirely on linguistic communication. While recent literature in scientific evaluation (e.g., grant reviews) suggests that "promotional language"—value-laden adjectives like "groundbreaking" or "transformative"—acts as a robust signal of novelty and increases funding success, it remains unclear how such rhetoric functions within the distinct, adversarial logic of the patent system. The central problem addressed is whether the linguistic strategies that aid the recognition of innovative ideas in collaborative scientific contexts inadvertently undermine their evaluation in the patent office, potentially creating a "promotional penalty" where high-quality, novel inventions are rejected due to their rhetorical framing.

Methodology

The study employs a large-scale computational linguistics approach on a comprehensive longitudinal dataset of 2,748,927 utility patent applications filed with the United States Patent and Trademark Office (USPTO) between 2010 and 2023.

Data Integration: The dataset integrates full-text specifications, administrative examination records, and bibliographic metadata. Unlike prior studies relying solely on granted patents, this sample includes both granted and rejected applications to mitigate selection bias.
Lexicon Construction: The authors utilized a validated, domain-diagnosed lexicon of 135 promotional words. This list was refined from an initial 139-word scientific lexicon by removing terms inconsistent with the patent context (e.g., "first," "key") to improve Cronbach's Alpha reliability from 0.18 to 0.60.
Key Variables:
- Independent Variable: The percentage of promotional words in the full-text application.
- Dependent Variables: (i) Patentability (grant vs. rejection), (ii) Ownership Transfer (signal of market value), and (iii) Appeal Success (reversal of initial rejection).
- Quality Proxies: To test if promotional language masks weak technology, the study measures Combinatorial Novelty Score (based on atypical CPC code pairings) and Citation Velocity (forward citations).
Statistical Analysis:
- Fixed-Effects Regression: Multivariate logistic and OLS regressions were used to estimate the association between promotional density and outcomes, controlling for technical novelty, linguistic features (readability, concreteness), inventor/attorney characteristics, and examiner demographics.
- Propensity Score Matching (PSM): Used to construct matched samples comparing applications with similar observed and unobserved characteristics but varying promotional intensities, isolating the causal effect of language.
- Heterogeneity Analysis: Interaction terms were used to examine variations across eight Cooperative Patent Classification (CPC) sections and examiner characteristics (gender and experience).

Key Results

1. The Promotional Penalty

Contrary to findings in scientific grant reviews, a higher frequency of promotional words is negatively associated with all three patent evaluation outcomes:

Patentability: Applications with higher promotional density are statistically less likely to be granted ( $\beta = -20.65, p < 0.001$ ).
Ownership Transfer: Higher promotional density correlates with lower probabilities of ownership transfer ( $\beta = -24.82, p < 0.001$ ).
Appeal Success: Promotional language reduces the likelihood of a successful appeal against initial rejection ( $\beta = -28.53, p = 0.001$ ).
Magnitude: In matched samples, moving from the lowest (Q1) to the highest (Q5) promotional density quintile results in a 5.5 percentage point decrease in grant probability, a 5.9 point decrease in transferability, and a 5.3 point decrease in appeal reversal.

2. Robustness Across Technological Fields

The negative association holds across seven of eight major technological sections (CPC Sections A–G).

Exception: In Section H (Electricity), which encompasses rapidly evolving fields like semiconductors and communications, promotional language shows a slightly positive total effect. The authors suggest this may be due to the high uncertainty and rapid recombination of knowledge in this sector, where promotional framing serves as a cognitive shortcut for examiners.
Traditional Sectors: In mature fields (e.g., Mechanical Engineering, Chemistry), promotional tone triggers stronger negative reactions, likely perceived as rhetorical overstatement obscuring technical substance.

3. Promotional Language Reflects Genuine Novelty

A critical finding challenges the intuition that promotional language is used to mask technical defects. The study finds a positive association between promotional density and objective quality metrics:

Novelty: Higher promotional density predicts a higher novelty score percentile ( $\beta = 0.37, p < 0.001$ ).
Impact: Higher promotional density predicts higher forward citation counts (3-year: $\beta = 1.45$ ; 5-year: $\beta = 2.47$ ).
Implication: Inventors using promotional language are often truthfully signaling genuinely high-value, disruptive innovations, yet the system penalizes them for it.

4. Examiner Heterogeneity

The "penalty" is not uniform but is moderated by human factors:

Gender: Male examiners show a higher tolerance for promotional language than female examiners, though both groups generally exhibit a negative reaction.
Experience: The penalty is disproportionately concentrated among novice examiners (lowest experience quintile). As examiners gain experience (Q2–Q5), their tolerance increases significantly. The most experienced examiners (Q5) even show an overall positive response to promotional words, suggesting they can look past the "linguistic packaging" to evaluate the technical substance.

Significance and Claims

The paper claims to reveal a fundamental paradox in the patent system: linguistic tools that successfully signal novelty in collaborative scientific peer review (grant funding) function as institutional noise in the adversarial legal context of patent examination.

Theoretical Contribution: The study challenges the assumed universality of linguistic signaling in innovation. It posits that the efficacy of communication strategies is context-dependent, shifting from a "promotional reward model" in open science to a "promotional penalty model" in competitive legal evaluation.
Practical Implications:
- For Inventors/Attorneys: The findings offer a counter-intuitive guide to avoid over-reliance on novelty-signaling rhetoric, as it may inadvertently increase rejection rates even for high-quality inventions.
- For Patent Offices: The results highlight an institutional blind spot, particularly among novice examiners who may rely on formalistic skepticism as a cognitive shortcut. The authors suggest targeted training and procedural adjustments could reduce this unintended bias.
- For Policy: By penalizing the linguistic markers of high-novelty inventions, the current system may inadvertently slow knowledge diffusion and reduce incentives for radical innovation.

The authors conclude that while their observational design cannot fully rule out all residual confounding, the robustness checks (PSM, field variations, and quality metrics) strongly suggest that the patent system is currently misaligned with the linguistic strategies used to communicate breakthrough technologies.