Skip to Content

AI Detection: Repetitive ChatGPT Words and Phrases

If you think the arrival of generative AI tools has made human-written content a thing of the past, you need to have a serious rethink.

Granted, AI platforms like ChatGPT have made life easier for content creators in many different ways, but they are not yet at the level where they can replace human input and oversight entirely.

Why? Because ChatGPT seems to draw from a limited pool of words and phrases, which is the case for both 3.5 and the enhanced 4.0 series, and therefore readers and search engines can spot AI-generated texts a mile off.

Let’s explore some examples of the words and phrases that ChatGPT likes to repeat to death, so you can reword or rephrase them should they appear in your own AI content.

Repetitive words and phrases in structured ChatGPT responses

You get a full sense of ChatGPT’s language limitations when you ask it to structure its writing as an article, guide or essay. This forces the chatbot to produce lengthier texts which have introductions and conclusions.

Commonalities in the introduction

When writing about specific tasks, ChatGPT will often introduce the topic with the phrase ‘when it comes to…’ e.g. “When it comes to baking cakes, preparation is key”.

For macro-level topics, the chatbot tends to kick things off with ‘in today’s digital world’—note that ‘digital’ is sometimes replaced with other strong adjectives like ‘dynamic’ or ‘fast-changing’, but the noun ‘world’ has a lot of staying power!

ChatGPT also has a habit of using redundant phrases like ‘in this article’ or ‘in this guide’ when setting out what the content is about. Worse still, it sneaks in bold adjectives like ‘intricate’, ‘ultimate, and ‘comprehensive’.

Bold being the operative word, because it nearly always over-promises and under-delivers. I can’t tell you how many times I have seen “The Ultimate Guide” used in the titles of AI-generated articles, only to be met with a body of text with less than 500 words.

Here’s a reminder of some of the most repetitive words and phrases in ChatGPT-written introductions:

  • When it comes to
  • In today’s digital world
  • Dynamic
  • Fast-changing
  • In this guide
  • In this article
  • Intricate
  • Ultimate
  • Comprehensive

Commonalities in the main body

The bulk of the text resides in the main body of the content, and this is where ChatGPT has to work quite hard to come up with ways to pad out its writing and satisfactorily answer the user’s query.

ChatGPT frequently uses the word ‘tapestry’ in a metaphorical sense to bring several elements together under one roof e.g. “Exploring the streets of Marrakech feels like unraveling a tapestry of vibrant colors, bustling markets, and intricate architecture”.

The chatbot often reaches for the word ‘resonate’ to bring more depth to its writing, as it opens up an opportunity to make emotional connections and highlight shared experiences, which in turn pushes up the word count.

It also likes to back up its statements with evidence by employing the phrase ‘a testimony to’ e.g. “His success in the face of adversity is a testimony to his resilience and determination.” Not only does this beef up the text a little bit, it enhances the quality of the writing overall.

Here’s a reminder of repetitive words and phrases in the main body of ChatGPT-written content:

  • A tapestry of
  • Resonate
  • A testimony to

Commonalities in the conclusion

As you would expect, ChatGPT aims to highlight the salient point(s) in its conclusions. To its credit, the chatbot does this well, but it relies on overused words and phrases to facilitate its concluding statements.

It does what many humans do when ending a paper and starts off the conclusion with common transition phrases such as ‘in conclusion’ and ‘in summary’, which weakens the originality of the writing.

It also repeatedly uses the verb ‘remember’ when it wants to remind the reader about a prevalent point or detail, usually with the intention to emphasize caution. If not ‘remember’, then ChatGPT will reach for ‘ultimately’ instead.

And finally, ChatGPT loves using the final line to end on a positive note by galvanizing the reader into taking action with phrases like ‘go ahead’ as in “So go ahead, grab your tools, and get ready to elevate your space with a perfectly placed shelf!”

Here’s a reminder of some of the most repetitive words and phrases in ChatGPT-written conclusions:

  • In conclusion
  • In summary
  • Remember
  • Ultimately
  • So go ahead

Words vs Phrases

From an AI detection perspective, it’s much harder to observe patterns among individual words than it is for phrases.

Think about it: a string of words in the same order will provide enough data points for search engines and AI content detectors to recognize a pattern.

Therefore, common ChatGPT phrases are a much bigger concern and you should rephrase them in the first instance or risk your content being distrusted, downgraded and deindexed.

That doesn’t mean you can freely publish content filled with overused ChatGPT words just because they are less susceptible to AI detection. It means you have a little more leeway in terms of the frequency in which they appear. And besides, some common words are not easily replaceable.

Why does ChatGPT repeat the same words and phrases?

The answer is twofold.

Firstly, ChatGPT received a substantial amount of training on the contents of the internet. More specifically, it was trained to recognize patterns in what the masses were writing and publishing online, and so the high frequency of certain words and phrases are the result of similar patterns in human writing.

Secondly, vague prompts will give rise to cookie-cutter AI-generated responses. If ChatGPT has nothing to work with, how can you expect it to output tailored information? You’d be surprised at the amount of users who are very lazy with their instructions, and most don’t ‘warm up’ the chatbot at all!

Can you block words in ChatGPT?

Unfortunately, ChatGPT is not able to block or filter certain words and phrases. You have to let the chatbot use an undesirable word and then you can ask it to regenerate the response with a different word.