Can Grammarly Detect ChatGPT?

Grammarly and ChatGPT are two of the most commonly used writing aids in today’s world. While Grammarly focuses primarily on grammar correction and spell checking, ChatGPT is a powerful AI language model that can generate entire articles and essays.

As more and more people are using ChatGPT to create content in inappropriate contexts, there is a growing need for ways to detect AI-generated content.

Whether you’re a professor marking your student’s essays or a content marketing manager suspecting your freelance writers of using ChatGPT, you may be wondering if you can use Grammarly to detect AI content.

Using AI content on your website is a poor decision, as the generated content often sounds robotic, can contain factual errors, and may not offer the reader the explanation they are looking for. It also fails to adequately connect with the reader like a human writer could, as AI language models can’t express emotion or understand subtleties.

This article will explore Grammarly’s functionalities and features and whether it can detect AI content.

The short quick answer: No, Grammarly can’t detect AI content. Unfortunately, there is no foolproof method for detecting AI content, although various tools, such as Originality.AI, can offer some assistance.

Grammarly’s Capabilities

Grammarly is one of the most popular grammar correction and spell checker tools on the market. It’s used by writers, students, and professional journalists alike.

It does an excellent job of correcting syntax, picking up on minor and major grammatical issues, and providing suggestions to improve the natural flow of your content. Grammarly is one of the most powerful grammar correctors out there – in my many years as a writer, I have yet to see another tool as accurate and powerful as Grammarly.

Of course, Grammarly isn’t perfect, and it does get a lot of things wrong. It also misses some grammar mistakes, so you can’t rely on it fully.

Grammarly can’t replace a proofreader; it simply acts as an extra tool to help you enhance and refine your content.

Nevertheless, it is the best grammar tool on the market, in my opinion, and it uses powerful machine learning algorithms to analyze your text and generate suggestions.

With the recent trend of AI writing assistants, Grammarly has jumped on the AI bandwagon and is now offering a built-in AI writer that people can use to generate content ideas and even write entire paragraphs of text.

It is called GrammarlyGO, and it’s available in the Grammarly web editor.

You can choose from one of the pre-generated prompts or create your own prompt.

Some things the GrammarlyGO AI writer can do include:

• Drafting email responses
• Rewriting content to better fit a certain tone or style
• Generating blog post ideas
• Creating articles and essays

You can also customize the style and voice the AI writer will use.

For example, you can tell it to use a certain tone and adopt your profession when it generates content.

Grammarly is free to use, but it comes with a limited number of free prompts. If you upgrade, you will get access to additional prompts.

In this example, I told GrammarlyGO to adopt a confident but formal tone as an IT professional and gave it a prompt to explain what to do when a user experiences a BSOD.

BSOD stands for Blue Screen of Death – a common Windows error in which a blue screen appears, and the user can no longer use the computer.

Also Read: ChatGPT vs Stockfish

This is what GrammarlyGO generated:

One more common use case of Grammarly, in addition to spelling/grammar correction and AI content generation, is plagiarism checking.

Grammarly has a built-in plagiarism checker, but you need a premium plan for full access.

The plagiarism checker is very powerful, although it tends to bring up a lot of false positives, in my experience. I prefer using Copyscape for that reason.

ChatGPT and AI-generated Text

Now, let’s discuss ChatGPT, a powerful language learning model released in 2022 that uses advanced algorithms to create entire articles from scratch.

Using natural language processing, ChatGPT is able to create content that is informative and helpful. Natural language processing, or NLP, is a field of artificial intelligence that revolves around training AI models on human content so that it can detect patterns in human language and simulate natural language in its outputs.

ChatGPT has been the subject of much controversy lately, as a lot of people have been using it as a shortcut to generate content they should be writing themselves.

That includes students required to submit essays, writers who are being paid to write original content, and people tasked with writing reports or briefs.

ChatGPT is incredibly powerful, and when used correctly, it is a great writing aid. For example, you can use it to generate article outlines and blog post ideas.

However, using it to generate entire articles is a horrible idea. The reasons for that are many, but let’s go through some of the main ones.

Check Out: ChatSonic vs ChatGPT

It May Contain Factual Errors

Photo by George Becker, Pexels

One of the problems of ChatGPT and other language learning models is that it can only pull information from the data in its database. This database is composed of sources from all across the web, but what if those sources are faulty?

Unlike actual humans, language learning models do not have the capability to determine right from wrong or differentiate what is real from what is not. It can only parrot what is being fed into it from its knowledge database.

That’s not even the worst of it. When faced with certain scenarios or prompts, ChatGPT has been demonstrated to completely invent facts from thin air.

A well-known case of this was when a lawyer representing a client who was suing Avianca (an airline) used ChatGPT to generate his legal filings.

The brief seemed pretty well put together, and the lawyer presented it in court. The filing cited prior cases, complete with quotes and citations, to argue the case.

The problem? Those cases didn’t exist.

Avianca’s lawyers and the judge searched high and low, but couldn’t find those cases anywhere in legal history. They just didn’t exist!

ChatGPT had fabricated those legal cases. The lawyer had even asked ChatGPT to verify that the cases were real, and it did so.

All those cases were fake, and facing what he called an unprecedented situation in which a brief was submitted citing bogus judicial decisions, the judge considered imposing sanctions on the lawyer, according to Forbes.

This is just one example of someone facing consequences for using ChatGPT and relying on it for information.

Now imagine different scenarios, such as someone using ChatGPT to generate user policies, legal compliance briefs for an organization, or market research reports.

If inaccurate information is present in the results, the organization could be in serious trouble.

Even if you are just a student or a freelance writer, if you use ChatGPT, it may generate articles or essays that contain factual errors. That could lead you to failing your course, and if the content is published, it could lead to your reputation being ruined.

It May Contain Plagiarism

Dahn, CC0, via Wikimedia Commons

ChatGPT and other language models rely on a database of articles to generate its outputs. A common misconception is that it copies and pastes from the internet; it does not.

That’s not how language learning models work. Language learning models rely on a database to collect knowledge and identify patterns in human writing and then generate their own content; they don’t copy any content.

However, there may be times when it uses phrasing that sounds suspiciously similar to an existing article, and depending on the plagiarism detector you are using, it may flag it as plagiarism.

Technically, using ChatGPT doesn’t count as plagiarism, since you’re not copying someone else’s article. However, your professor may count it as plagiarism, and if they discover that you are using ChatGPT to generate your essays, you may fail the course and even get banned from your school.

It’s Dishonest

Photo by Magda Ehlers, Pexels

Using ChatGPT to write an article that you were supposed to write yourself is simply dishonest.

If a client is paying you for your time and effort to write an article, you really should be writing it yourself. They could use ChatGPT themselves for free, so why would they need to pay you for it?

I would consider something like that stealing, since you’re taking someone else’s money for a service you are not providing.

If you are using ChatGPT to get you through school, it is also dishonest, and it won’t help you advance in your studies or career.

There are situations in which it is not dishonest, of course. If a client wants you to generate AI content and then tweak it to sound more human, that is perfectly acceptable.

A client might prefer that so that they can pay you less, since it would, theoretically, require less time to rewrite AI content than to write an entire article yourself from scratch.

Explore: Turnitin vs Grammarly

It Sounds Unnatural

One of the main reasons I discourage you from using ChatGPT is that it simply sounds unnatural. I have personally come across websites – including websites that are supposed to be authoritative, such as medical websites maintained by clinics – with content that just sounds weird.

This decreases the level of trust I have for these websites. If I suspect that an article has been written by ChatGPT, I will immediately leave the site and look for an article that was written by a human, which I trust more, especially if the writer has the appropriate credentials.

Grammarly’s Detection of ChatGPT

So, onto the big question: Can Grammarly detect ChatGPT?

Unfortunately, the answer is no. Grammarly is not designed to detect AI content.

Instead, it’s built to detect grammar and spelling mistakes and provide suggestions on how to correct them. It can also detect plagiarism, but since ChatGPT doesn’t copy and paste sentences from the internet, it’s unlikely for Grammarly to accurately detect content written by ChatGPT and flag it as plagiarism.

Will Grammarly ever release a tool that can detect AI content? The answer is that it is unlikely.

Remember, Grammarly now has its own AI writing assistant called GrammarlyGO. If it were to release a tool aimed at detecting AI content, it would be shooting itself in the foot – it’s just a horrible business decision that doesn’t make any sense.

Alternative AI Detection Tools

While Grammarly can’t detect AI content, there are other tools that are designed to detect content written by ChatGPT and other language learning models.

However, I want to stress that these tools can’t really be relied on as the final word as to whether content is AI-generated or not.

Tools that are designed to detect AI content use AI themselves to do that. They work by detecting patterns common in content produced by chatbots.

Some of these patterns may include:

• Repetitive words or ideas
• Lack of slang
• Lack of grammar or spelling mistakes
• Too many connecting words (such as “ultimately” at the beginning of a sentence)

While these patterns are common in AI-generated text, their presence doesn’t mean that the content is actually generated by AI. You may find that you can actually bypass these AI detectors by simply inserting a few grammar mistakes and slang words, which is actually crazy when you think about it.

In fact, these AI detection tools have incorrectly flagged the US Constitution as being generated by AI, which is ridiculous. That just shows how inaccurate they are – it’s basically AI trying to detect AI.

Nevertheless, these tools do have some functionality, but you should take their results with a grain of salt.

They are often good at detecting AI content.

The problem lies in the false positives that these tools tend to produce, which have led professors to ban students or give them a failing grade for no good reason. You can not rely on these tools to give you authoritative results.

Instead, you must use your own intellect. Much of the time, by taking a good look at the article, you can see for yourself whether it has been generated by AI or not.

AI content tends to be repetitive and use syntax that sounds a bit off. If an article is unnatural and uses phrases like “one can do it” instead of “you can do it,” it might be written by AI.

Sometimes, an article just flows wrong and doesn’t sound like it was written by a real person. It fails to provide real opinions and emotions like a human writer would.

So, do use these tools when appropriate, but don’t rely on them entirely and use your best judgment.

With that said, let’s talk about some tools you can use to detect AI content.

Check Out: Best Turnitin Alternatives

1. Originality.ai

Originality.AI is one of the top AI detectors on the market. It uses advanced algorithms to detect patterns in AI content and determine how likely it is for an article to have been written by AI.

Originality.AI will highlight portions of a text that it thinks were likely to be written by AI. Some people try to get away with using ChatGPT by only inserting some paragraphs of AI content into their articles.

One thing that sets Originality.AI apart from other tools on the market is its Chrome extension. The Chrome extension tracks your writing in real time in Google Docs and generates an AI report based on that.

People who use AI tend to copy and paste entire articles or sections of articles, and if you are updating your document word by word in real time, it is unlikely you are using AI.

This isn’t foolproof either, as technically, someone could rewrite the entire ChatGPT article instead of copying and pasting it to get around this detection tool. Still, it’s a lot more reliable than most other AI detection tools on the market, in my opinion.

Explore: Best Quillbot Alternatives

2. GPTZero

Another tool that is gaining popularity is GPTZero, not to be confused with ZeroGPT, a separate tool with a similar name.

GPTZero, despite its name, can detect AI content not only written by ChatGPT itself but also content written by other AI language models, such as Bard.

3. Turnitin

Finally, we have Turnitin. Turnitin is a similarity detection service aimed at ensuring students aren’t plagiarizing from other articles.

Essays submitted by students are stored in a database, and future submissions are compared against this database to ensure there is no copying of other students’ works.

Turnitin has now launched an AI detection service as well, which is aimed at helping professors detect students who are using ChatGPT or other AI language models to generate content.

Also Read: How To Copy & Paste Without Plagiarizing?

Conclusion

Grammarly, unfortunately, is unable to detect AI content.

However, there are other tools that are designed to do so, and while they are not yet perfect, they are getting better, and there may come a time when they could be relied on to detect AI content authoritatively.

In either case, I highly discourage you from using AI generated content unless you were specifically asked to do so. It’s dishonest, could land you in trouble, and the results just don’t sound natural.