Weaknesses of AI-Generated Writing—and Why You Must Edit

1. Unnecessary Gerunds

People in corporate and academic settings often use gerunds (verbs ending in “-ing” and serving as nouns) as the subjects of sentences to sound more formal. This layer of distance and abstraction is seen as objective or authoritative. But these gerunds rarely add value and can usually be deleted. Removing them reveals the true subject, making the sentence clearer and more direct.

Here’s how WordRake handles the gerund trap:

Developing the The new data needed to support an acceptable non-experimental design would be costly.

Creating a A new appropriation for the Board will create greater visibility and transparency.

Surveys and questionnaires are subject to interpretation; having clear definitions is clear definitions are a key part of the research process.

Enhancing leadership Leadership development programs nurtures nurture talent.

Implementing new New recycling initiatives reduced trash production in the town.

Optimizing search Search functionality improves website navigation.

Participating in the The conference inspired me to try a new field of research.

It’s likely that LLMs learned to replicate the gerund sentence-starter pattern based on corporate, academic, and professional training material, in which this structure is common. The LLM appropriately mimics the tone it associates with formal writing styles, but it doesn’t see how gerunds can reduce clarity and unnecessarily increase word count. While gerunds are a safe choice by AI tools to maintain grammatical correctness, they’re clunky. When you get this output, be sure to cut the gerunds.

2. The “Not Only/But Also” Construction

In school, we were taught to create sophisticated arguments using cohesive devices like the “not only...but also” construction. Especially when spoken, this structure sometimes creates powerful rhetorical emphasis and forces the audience’s attention to the two ideas by either showing their equal value or elevating their clever contrast. This pattern can be a tool for persuasion. But when it’s used without carefully chosen rhetorical intent (and when the two ideas are not actually related), it falls flat and confuses readers. Simplify the sentence by removing the emphatic markers of “not only...but also” or, if the ideas aren’t closely related, split them into separate sentences. If they are related, find a more natural way to connect them.

Here’s how WordRake streamlines unnecessary cohesive devices in AI writing:

The president not only honors the Jones space legacy, but also and challenges the nation with his vision for the next era of exploration.

These varying definitions of the term meant to define a federal cross-waiver will undoubtedly complicate not only its interpretation but also and the analysis of preemption.

Hybrids not only provide improved fuel economy, but also and a corresponding reduction in emissions.

Any staff who accompanied the tourist back to the hotel would not only have solved the problem, but also would have and ended up with a huge tip.

The country has not only produced great works of art, but and is also the most industrialized country in the region.

Scientists also believed that impetus could be not only linear, but also and circular in nature, causing objects (such as celestial bodies) to move in a circle.

They focus not only on complete works but also and on the reserved spaces of freedom inside intellectual property.

It’s likely that LLMs learned to rely on the “not only…but also” construction to give the appearance of complex persuasive thought. LLMs use syntactic patterns seen in human writing such as sophisticated, persuasive scholarly papers. Since the “not only…but also” construction is statistically common in persuasive or argumentative writing, GenAI overuses it to mimic the training data that matches these goals, even when it isn’t necessary. And, without big-picture context, GenAI output matches the style but can’t match the content elements properly. Don’t let this formulaic structure mislead you. Look for these constructions and simplify them.

3. Superfluous Summaries and Obvious Conclusions

Phrases like “in conclusion” are often taught in school as a way to signal the end of an argument. This formulaic approach helps beginner writers create structure, but it should become unnecessary as writers improve. The same is true of generalizations and summaries. The writing need not announce itself with “in conclusion,” “in summary,” or “in general.” Trust readers to understand and cut these obvious phrases.

Here’s how WordRake deletes these needless phrases:

Overall, legal Legal professionals can leverage technology to enhance their career development.

In summation, Mr. Smith has made a modest effort to serve the communities in which he has lived.

Despite the extreme dangers to the camp and the staff, the nature preserve has, in summation, been improved by the fire.

In conclusion, journalists Journalists are taught to write differently than lawyers because they serve distinct purposes, cater to different audiences, and follow unique conventions.

For the reasons stated above, these These people were excluded from the analysis.

In summary, clear Clear and concise legal writing is essential to enhancing comprehension understanding, information sharing, and knowledge management in the legal profession.

In conclusion, effective Effective data management is crucial important to the success of the software company and of tech companies as a whole.

It’s likely that LLMs use these signals to create the appearance of structure and coherence. Desirable patterns found in the training data may come from scholarly academic articles where organizational markers like “in conclusion” are common and expected, even if unneeded. Since AI-generated output seems designed to follow the form of a standard five paragraph essay, it’s no surprise that we get the most elementary version. Try cutting these phrases so your main point stands out more clearly.

4. Empty Introductory and Transitional Phrasing

Good writing is often described as flowing. With flow, readers smoothly move from idea to idea without stumbling or doubling back. Strong writers organize their arguments by building on and referring to earlier ideas. Weak writers try to achieve flow through transition words. Clunky phrases like “that being said,” “with that in mind,” or “as mentioned earlier” are intended to link ideas, but they rarely add any real value. When combined with other dependent clauses, they also create run on sentences. Skip transitional phrases and pointless introductions; try tighter wording to keep ideas closer together.

Here’s how WordRake deletes the unnecessary transitions and empty introductions:

In this day and age, Now, computer literacy is a necessity.

Moreover, the The location of witnesses and documents is not a paramount consideration in this day and age now.

Writers often become attached to their words, but it is important to recognize that cutting unnecessary words can strengthen the message.

Similarly, in In legal writing, engaging the reader plays a pivotal role in presenting persuasive arguments and supporting evidence.

It’s important to note that developmental Developmental editing does not typically involve rarely involves line-by-line copyediting or proofreading for grammar, spelling, or punctuation errors.

Additionally, by By insuring the majority of Americans, the legislation reduces the cost of caring for the uninsured.

Similarly, in In New York City, local historic district properties gained more value than non-designated properties.

It’s likely that LLMs, like humans, use these familiar linking phrases to create flow and smooth transitions between ideas. This may be necessary in longer texts, but not in the writing offered by most GenAI tools. These phrases also sound overly conversational for formal writing. If these phrases appear in your AI-generated text, you can delete them without losing meaning.

5. Wordy Prepositional Phrases

Both spoken and written English are plagued by wordy prepositional phrases that are better written as single words—or deleted. These phrases muddle sentences and delay value. So instead of “[verb] in a way that improves,” use the verb “improves,” or replace “in the manner that” with “as.” And try replacing phrases that describe how something happens with an adverb. These approaches reduce word count, improve readability, and make writing more direct.

Here’s how WordRake shrinks prepositions or converts them to adverbs:

The project was executed in a way that was systematic systematically, with each phase building on the success of the previous one.

The Department tries to operate in the most efficient way possible efficiently to keep costs low and ensure the best use of taxpayers’ dollars.

The team had to solve the puzzle using unconventional methods; they had to approach it in a novel and strange manner differently.

The textbook presents information in a logical and systematic manner logically so the reader understands the main points and arguments.

We look forward to executing the lease agreement in the very near future soon.

He responded to the client’s request in a prompt manner promptly.

These principles of contract construction were disregarded by the arbitrator and the trial court when they effectively rewrote the arbitration agreement in such a way as to deprive Ms. Anderson of rights granted both by statute and by the rules of civil procedure.

When the first of these lawsuits was filed in August 2008, the Corporation effectively functioned in a manner similar to like a workers’ compensation insurer for its employer-subsidiaries.

By utilizing Using editing software, you can streamline your writing process, eliminate errors, and refine your work for maximum impact.

When you’re using LLMs to generate text, you’ll see this padding often. It likely comes from GenAI’s inability to see the complete sentence. With a full sentence in view, it’s obvious that we don’t need “in the manner that X” or “of the type that Y” when a familiar adverb is available. But when generating a sentence word-by-word, LLMs need time to find the right descriptor, so the sentence meanders until the LLM reaches the next statistically relevant option.

6. Forgettable Nominalizations

Who hasn’t read a business article, a legal document, a technical discussion, or an academic paper filled with nominalizations? Nominalization—turning verbs into nouns and combining the noun with a generic verb (e.g., “engage in implementation” instead of “implement” or “perform decision-making” instead of “deciding”)—is a common problem that persists because it’s what we think professional writing should sound like. But nominalizations make sentences more abstract and less engaging, making the writing feel heavy or bureaucratic. Find the root verb and transform the sentence.

Here’s how WordRake transforms nominalizations:

She will advocate for the expansion of to expand the diversity and inclusion work underway at the school.

Journalists craft attention-grabbing leads to hook readers and provide a concise summary of summarize the story.

This new technique will have an important impact on impact society, influencing the treatment of both common and rare diseases.

An appraiser will do a thorough inspection of thoroughly inspect each property.

They are operating under the assumption assuming that the CEO will retire this year.

Photos give allow people the opportunity to capture a cherished part of life.

Scientists have come to the conclusion concluded that all chemical actions are electrical.

The intention of the legislators was intended to reduce drunk driving.

The witnesses gave a description of described the bank robber.

He had no knowledge did not know that the company was in dire straits.

They haven’t given any justification for justified their move.

Since LLMs are trained on available writing, flaws included, it makes sense that nominalizations would feature in the output at the same rate they appear in the training set, which is largely internet writing and technical and academic documents. Frequency aside, nominalizations also make complex, multi-clause sentences possible. If generating text while keeping options open is the goal, then nominalization is a great tool. It’s technologically clever but it harms readability and makes sentences more abstract. When you’re working from AI-generated text, search for words ending in -ion and -ment and convert them to their original verbs to improve pace. Read more about nominalization here.

7. Telltale AI Word Choices

A good writer will carefully select each word for its precise meaning. A human writer’s goal is not to maintain flexibility or ambiguity to account for many possible interpretations. Instead, the writer forms meaning, remembering her overall content goals, before recording the idea in text. With each word, the writer embraces the constraints of the text on the page. By embracing clear, specific diction, you can write vivid sentences that connect with your readers. If you’ve started with an AI-generated draft, cut bland words and needless helpers and hedgers.

WordRake changes these telltale words to more concrete and direct vocabulary:

In this section, we will delve dig into the valuable insights provided by from Ms. Lincoln in her article on bringing journalism skills into the legal writing class.

In the realm of fantasy writing, mythical creatures and enchanted lands become the backdrop of epic quests and heroic adventures.

The world we live in is getting more connected by the day.

The book was authored by a local reporter.

Visual aids may aid comprehension improve understanding of the lecture material.

The report identified demographic, socio-economic and political trends as being significant drivers causes of change in social work services since the 1968 legislation.

In a world where When technology keeps advancing, staying informed and vigilant is our best strategy.

With due consideration to Considering the competitive landscape, it is crucial to differentiate the product’s offerings by highlighting its unique features and benefits.

In conformity with Under the applicable regulations, the attorney must exercise due diligence in representing the client’s best interests throughout the legal proceedings.

The software offers a multitude of many benefits to writers of all levels.

This provides a visual guide to all alterations, allowing you to clearly see clearly showing what’s been adjusted, how it’s been adjusted, and where.

These individuals were tasked with scouting the market for novel original ideas and analyzing trends to identify find emerging market opportunities. 

Unlike human writers, it’s likely that LLMs resist constraints by using vocabulary that allows for multiple interpretations as each word gets added. The best way to do this is to use abstract words that can fit with any topic. But these words stick out in their unfamiliarity and blandness. In fact, researchers have found that AI-generated content has vocabulary that is statistically much less common in human writing.

8. Other AI Bloat

All first drafts include wasted words, but AI-generated text includes bloat that most writers would stop themselves from typing. This output is often filled with clichés, tautologies, and unnecessary words. It’s the most annoying part of merely average yet grammatically correct writing. It’s also what you might find from a content mill.

Because LLMs rely on frequent word associations to avoid generating nonsense, any word strings that appear together repeatedly in human writing are more likely to be generated together in the output. Set phrases like clichés and tautologies provide convenient patterns to mimic. But these patterns are formulaic and dull, so revise them to fit your purpose and never write about an “intricate tapestry” again.

Here’s how WordRake trims the fat from your sentences:

However, despite Despite widespread enthusiasm, this development is also creating challenges.

The rapid advancement of artificial intelligence has revolutionized various sectors, such as like healthcare and manufacturing.

These small business growth programs assist in helping help shepherd new products through the commercialization process.

Legal writing can benefit from the use of editing software, which can improve clarity and word count.

The City may in the future amend the City Map to reflect a “Public Place” designation on parts of Lincoln Avenue.

Effective writing requires an understanding of the target audience and the ability to tailor the message accordingly.

There was no claim made by any Respondent in response to the Court’s Orders to Show Cause that Mr. Smith had prior experience with bankruptcy stays.

The construction crew successfully rebuilt an exact built a replica of the historic bridge, restoring it to its former grandeur.

Firstly, First, a lawyer should understand the law.

Tacit knowledge is the knowledge that is difficult to articulate or formalize, and is typically acquired through personal experience, observation, or practice.

In this case, Here, the lawyer’s expertise is the key resource that they bring to the table offer.

Despite the fact that Although the court would rule in Defendant’s favor, Plaintiff pursued its case.

Why You Must Edit After Generating Text

Wordy writing wastes your reader’s time and makes it harder for readers to grasp your message. That’s why you should edit anything, no matter where it came from. But obviously AI-generated text does something worse: it offends your readers. If the writer didn’t invest the time to write the text, why should the reader invest their time to read it?

AI-generated text is a useful starting point, but it needs more to become clear, engaging, and effective. Take two steps when editing AI-generated text. First, use WordRake to tighten, tone, and humanize your writing quickly. By stripping away the unnecessary complexity, cutting redundant phrases, and tightening sentence structures, you can transform robotic text into something that reflects your voice and your message. Once you’ve decluttered the AI-generated draft, you’ll have the space to add the thoughts that matter. Then review the text yourself or work with a human editor to clarify your message and make sure it matches your intent. Writing is about creating meaning to share with other humans. Show respect for your audience by investing some effort in editing.

Ready to improve that AI-generated first draft? Try WordRake editing software free for 7 days

About the Authors

Danielle Cosimo is a Language Usage Analyst for WordRake. Before joining the team, she was a translator and editor for non-native English speakers applying to degree programs in the United States and the UK. Danielle is formally trained in linguistics and has a certificate in computer programming. She is fluent in English, Portuguese, and Spanish. She applies her interdisciplinary knowledge to create WordRake’s editing algorithms.

Ivy B. Grey is the Chief Strategy & Growth Officer for WordRake. Before joining the team, she practiced bankruptcy law for ten years. In 2020, Ivy was recognized as an Influential Woman in Legal Tech by ILTA. She has also been recognized as a Fastcase 50 Honoree and included in the Women of Legal Tech list by the ABA Legal Technology Resource Center. Follow Ivy on Twitter @IvyBGrey or connect with her on LinkedIn.