How to Avoid Sounding Like AI in Your Writing


Key Takeaways

  • Making small tweaks to your writing style can prevent AI detectors from flagging it as non-human.
  • Avoid repeated patterns in your sentences to sound more human-like and improve your writing.
  • Using personal anecdotes, contractions, and varying sentence lengths can add authenticity and engagement to your writing.



Whether you use AI to support you when writing your work is entirely your choice. However, very little is more frustrating than your human-written work being flagged as potential AI content when, in fact, you wrote every word yourself.

As an academic proofreader and ex-English teacher, I have become familiar with various ways you can adapt your writing to avoid sounding like an AI bot, here’s how you can do it too.


How Does AI Detection Work?

AI detectors often look for patterns that make writing sound like it’s not written from a human stream of consciousness. To do this, they apply classifiers (which categorize text into things determined by usage, grammar, style, and tone), embeddings (which determine how each word relates to others), perplexity measures (which assess the text for randomness), and burstiness measures (which evaluate sentence structures and lengths).


So, making small tweaks in these areas will help us to make sure we’re not writing in a way that AI detection software would highlight as being non-human.

Disclaimer: The key thing to note when considering whether your writing sounds like it’s produced by AI is that AI detectors don’t all necessarily work the same way, and
there’s no evidence that any are reliable
. The hints in this guide will go a long way to making your writing sound more human-like, though following the tips will help improve your writing quality in general.

Avoid Repeated Patterns

Repeated patterns can be useful in certain scenarios, such as for emphasis. For example, notice how politicians often use three-part slogans to add gravitas to their speeches. However, modifying the sequences of your words will help you to sound more human.

One way to do this is to use different words at the start of your sentences. For example, if you repeatedly start with ‘The’, it’s likely that the next word will be a noun or an adjective, and that will largely dictate how the rest of your sentence will be structured. Instead, start some sentences with verbs or adjectives—indeed, this is a handy tip for more interesting writing more generally.


Avoid too much of this…

Try more of this…

The lion is a large cat with a rounded head and a hairy tuft at the tip of its tail. It is native to Africa and India.

Native to Africa and India, the lion is a large cat with a rounded head and a hairy tuft at the tip of its tail.

Overall, focusing on how you start your sentences will go a long way to letting AI detection know that you are, indeed, a human.

Use Synonyms

If you use the first word that comes to mind, it’s likely that many other people have used the same word before, so the AI detection software is more likely to recognize that word’s relationships with others.

According to British lexicographer Susie Dent, the average active vocabulary of an adult English speaker is 20,000 words, but when we really think about using alternatives, our lexicon doubles to 40,000 words. This means that if you take a moment to think of an alternative word, you could be doubling your chances of avoiding a false positive.


The key is not to overdo it. If you use too many synonyms, your work will sound unnatural, and you might lose your intended meaning. Similarly, you can right-click a word in Microsoft Word to see suggested alternatives, but make sure you choose a word that accurately fits the context.

A Word document with the Synonyms option displayed under the word 'tells.'

Rectifying this as you type might end up taking too long and might disrupt your flow. If this is the case, wait until you proofread your work. You’re more likely to spot mundane or repeated words when you read your work from start to finish. Even more usefully, you can get your computer to read your document back to you, which might help you spot other things you want to tweak.


Vary Sentence Lengths and Structures

AI-produced content usually consists of sentences that are similar in length and structure. Take this example, which was written by AI:

A dog is a domesticated mammal and a member of the Canidae family, scientifically known as Canis lupus familiaris. They are known for their loyalty, companionship, and diverse range of breeds, each with unique characteristics and traits. Dogs have been bred for various purposes, including herding, hunting, guarding, and companionship. They are often referred to as “man’s best friend” due to their strong bond with humans and their ability to understand and respond to human emotions. Dogs communicate through vocalizations, body language, and facial expressions, making them highly social animals.

Of these five sentences, four of them are 13 to 19 words long. What’s more, notice how they all read monotonously, with limited variation in their structures. You’ll also spot the same words being used at the start of the sentences, something we touched on earlier.


This simply isn’t how most humans communicate, whether in speech or writing. We use varied sentence lengths, naturally structure things in certain ways to add emphasis or provide more detail, and generally add more interest to what we output. It comes as no surprise that, when the paragraph above is re-inserted into AI detection software, it shows as being 100% machine-produced.

A paragraph of text showing as AI-produced when inserted into AI detection software.

So, reorganizing your sentences into shorter and longer ones will not only help you avoid false positives, but also make your writing sound more interesting.


Include Personal Anecdotes

AI writing tools use large language models (LLMs) to produce their output. LLMs contain parameters and are trained on large data sets, such as books, websites, articles, journals, and others, to “learn” how language works. As a result, including anecdotes, lived experiences, or individualized language adds a personal touch to your writing that AI would struggle to replicate, which, in turn, means that it’s less likely to be flagged by detection software as artificially generated.

Even if whoever is reading or assessing your work doesn’t plan to run it through a detector, adopting a personal approach has other benefits. For example, take Google’s EEAT (Experience, Expertise, Authoritativeness, and Trustworthiness) principle, which helps the search engine determine which search results to prioritize. If you’re writing online, sharing unique experiences and incorporating authenticity into your writing will boost your EEAT rating, meaning you’re more likely to get more hits on your website. On a more basic level, if your writing is more personal, it’s more likely to be more engaging.


Use Contractions

Shortening some of your words with apostrophes—like using “she’s” instead of “she is”—adds a more conversational tone to your writing and, as a result, will read less robotically.

Whether you choose to follow this tip depends on the context of what you’re writing, however. If you’re crafting a formal academic thesis, for example, you might be advised to avoid shortening your words in this way. On the other hand, contractions will appear less out of place in a blog post, an informal website article, or an email—in fact, these are all genres of writing where being more chatty could be deemed more favorable. Either way, if you do choose to employ contractions, do so sparingly.



As well as making small tweaks to your writing style to prove that you are, indeed, a human, there are other steps you can take. For example, sharing earlier drafts of your work, keeping a record of your sources, screen-recording your work process, and using tools that track version history will back up your authenticity and help you counteract claims that your work isn’t your own.



Source link

Previous articleBytes: Week in Review — Dems’ tech platform, Google’s legal troubles and OpenAI’s newest partnership