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 in the fields of public relations, digital marketing and web content. These tools have become increasingly sophisticated since they were first introduced in 2022 as they are now capable of generating text across various genres and styles at impressive speeds.
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 not before the New York Times bemoaned the use of “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.
The Necessity of Human Writers for Fixing AI’s Common Grammatical Errors
In the previous blog on Writing For Humans™, we looked at 10 common grammatical errors that have been found in AI content proofreading. Now, let’s take a look at another five mistakes that we have seen (and that should not have such dire consequences as the “expensive comma.”
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.”
Human Writers for AI Content Editing
These errors and the ones cited in my previous blog underscore the importance of 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.
###