Fixing AI’s Common Grammatical Errors

Fixing AI’s common grammatical errors would almost seem unnecessary. Why would artificial intelligence (AI) generate these errors in the first place?

With advanced language models and natural language processing capabilities, AI writing tools have revolutionized content creation and have become increasingly sophisticated since they were first introduced in 2022.

However, while these AI tools offer incredible efficiency and productivity gains, they have proven time and again that they are not infallible. They are prone to both very large mistakes like “hallucinations” and to very small mistakes, like grammatical errors. Even the most advanced AI models make critical grammatical and stylistic mistakes that an experienced human writer would not make – and that a good human AI-content editor would not miss. These types of AI-specific errors are telltale signs of machine-generated content.

Moreover, AI lacks the obviously crucial ability to fact check its own output. With that in mind, content strategists and content creators must stay on heightened alert for the grammatical errors that even the latest and most sophisticated AI chatbots can make.

Understanding the limitations of AI-generated content is crucial for maintaining high-quality content and preserving brand integrity while creating clear, concise, and compelling content that connects to its intended audiences – and moves them to act.

The Misplaced Comma

As an example of the severe consequences that grammatical errors can have, BBC Worklife told the story of the how for most people, a stray comma isn’t the end of the world, but in some cases, the exact placement of a punctuation mark can cost huge sums of money.

More than 150 years ago, for example, the United States Tariff Act included an unwanted comma that cost taxpayers nearly $2 million — the equivalent of $40 million today. As originally drafted in 1870, it allowed “fruit plants, tropical and semi-tropical for the purpose of propagation or cultivation” to be exempt from import tariffs.

However, when the Act when revised two years later, a stray comma sneaked in between the words “fruit” and “plants.” Suddenly all tropical and semi-tropical fruits – not just fruit plants — could be imported without any charge. Fruit importers quickly seized on this grammatical blunder and refused to pay the tax on their shipments. This was a little mistake that had large monetary consequences. In fact, members of the U.S. Congress actually debated the issue and the problem was fixed, but this was an expensive comma.

This is the kind of error that could have been generated by AI text, but that would not have been caught without an eagle-eyed human editor reviewing that text.

Human Writer vs. AI: Correcting Common Grammatical Errors in AI Content

Here are 15 common grammatical errors that have been found in AI content proofreading (and that should not have such dire consequences as the expensive comma).

Repetition of Words or Phrases

AI can repeat certain words or phrases within a short span of text. This can occur because the model tries to emphasize a point or fill in some more details, but instead repeats the same word again.

AI: “The project was successful because of the team’s execution of successful strategies.”
Human Writer: “The project was successful because of the team’s careful execution of well-thought-out strategies.”

Incorrect Use of Tenses

AI can struggle to maintain consistent tense throughout a passage, especially in more complex sentences where tense switches between the past, present, or future.

AI: “She was preparing the report yesterday, and she completes it now.”
Human Writer: “She was preparing the report yesterday, and she is completing it now.”

Overuse of Passive Voice

Overuse of passive voice can make the text feel impersonal or awkward. Passive voice is grammatically correct, but AI can default to passive constructions, which weaken the clarity and directness of the message.

AI: “The project was completed by the team in record time.”
Human Writer: “The team completed the project in record time.”

Agreement between Subject and Verb

Content generated by AI can confuse agreement between the subject and verb in a sentence. This happens especially when AI is dealing with more complex sentence structures or when multiple subjects are involved.

AI: “The group of students are excited.”
Human Writer: “The group of students is excited.”

Unclear or Ambiguous Pronouns

AI text can be unclear or ambiguous with its pronoun use. This causes problems for readers because they cannot determine to whom or to what the pronoun refers.

AI: “John met with Steve, and he said he was tired.”
Human Writer: (Who was tired, John or Steve? Is it: “John met with Steve, and John said he was tired.” Or is it: “John met with Steve, and Steve said he was tired.”)

Run-On Sentences

This can be the bane of every writer and AI is no different. AI models sometimes generate long, convoluted sentences without proper punctuation or breaks. This makes the text difficult to read and understand.

AI: “The new software update provides several new features which will improve user experience and productivity but some users may find it difficult to adjust and might need some training to understand the changes which could affect overall satisfaction.”
Human Writer: “The new software update provides several new features which will improve user experience and productivity. However, some users may find it difficult to adjust. They might need some training to understand the changes that could affect overall satisfaction.”

Improper Use of Articles

The improper use or omission of the articles “a,” “an,” and “the” is a pretty common problem with AI-generated content. This is especially the case with noncount or uncountable nouns and abstract concepts. A noncountable noun is a noun that usually cannot be expressed in a plural form. It is not something that can be quantified, like “milk,” “water,” “air,” “money,” and “food.” You cannot say, for example, “He had many moneys” or “The airs smelled good this morning.”

AI: “She gave him advice on how to improve an productivity.”
Human Writer: “She gave him advice on how to improve productivity.”

Misuse of Prepositions

AI can use incorrect prepositions or place them in the wrong place, especially in idiomatic expressions or phrasal verbs.

AI: “She is interested for learning.”
Human Writer: “She is interested in learning.”

Misplaced Modifiers

AI-generated text can contain sentences with misplaced modifiers, which make the sentence unclear or altering the intended meaning.

AI: “Walking through the park, the flowers were blooming beautifully.”
Human Writer: “Walking through the park, I saw how beautifully the flowers had bloomed.”

Inconsistent Use of Tone and Style

Consistent tone can be a problem for any writer, and AI is no different. It can fail to maintain a consistent tone, inappropriately shifting from formal to casual or vice versa. This can happen particularly when transitioning between different parts of a text.

AI: “The company’s profits increased. Wow, that’s awesome!” (A shift from formal to casual).
Human Writer: “The company’s profits increased. That results were very satisfying.”

Unnatural Collocations

Collocations are word combinations that native speakers typically use together, and AI often produces unnatural or awkward collocations that aren’t proper to English readers.

AI: “She made a large success.”
Human Writer: “She achieved great success.”

Conjunction Overuse

AI-generated text can overuse conjunctions like “and,” “but,” or “so.” This results in long-winded run-on sentences that need to be broken down for better clarity.

AI: “He went to the store, and he bought some apples, and then he went home, and then he cooked dinner.”
Human Writer: “He went to the store and bought some apples. Afterward, he went home and cooked dinner.”

Improper Use of Comparative or Superlative Forms

AI can sometimes confuse the comparative and superlative forms, especially when comparing more than two items.

AI: “This is the most better solution.”
Human Writer: “This is the better solution” or “This is the best solution.”

Inconsistent Punctuation

AI often struggles with the proper placement of punctuation marks in a sentence. This leads missing commas, the incorrect use of semicolons, and run-on sentences.

AI: “I went to the store bought some milk then went home.”
Human Writer: “I went to the store, bought some milk, and then went home.”

Incorrect Word Choices between Homonyms and Synonyms

We have all had the problem of dealing with words that sound the same as other words, but have different spellings and meanings (homonyms) and two words that have the same or similar meanings (synonyms).

AI: “Their going to the park.”
Human Writer: “They’re going to the park.”

The Necessity of Human Judgment in AI Content Editing

These errors and the ones cited in my previous blog underscore the importance of expert human writers overseeing the generative AI content creation and editorial process. By leveraging AI as a starting point and applying our human professional judgment and expertise to review, refine, and perfect AI-generated content, we can create communications that are grammatically flawless. Having a human writer in charge of the editorial process can make all the difference in the world – a difference that can spell success and avoid embarrassment and potentially costly mistakes.

Have something valuable to say but need help shaping it for your audience?

Let’s talk about what you’re working on and what you need to accomplish.

Start a Conversation →