AI chatbots The Internet now accounts for an incredible amount of content, including written text. Here’s the problem: text is one of the easiest things for chatbots to generate and often harder to pick out as fake. AI images or Videos (Especially if you Make your gestures precise). Still, capturing AI-generated text isn’t impossible. I spend a lot of time reviewing chatbots, so I’ve compiled seven common gifts of AI-generated text that I’ve encountered.
1. Undue excitement or enthusiasm
AI chatbots tell you what you want to hear. So, if your cue implies a certain point of view, the response you get will confirm that belief. This type of structure in written form is the biggest draw of AI-generated text, especially if the original context of the text in question doesn’t justify the level of motivation or excitement you see.
For example, I recently developed a skincare routine with AI.. When I send my routine description to ChatGPT, it tells me, “Overall decision first: Your routine is very well thought out and mostly solid.” (Yes, the bold text is part of the chatbot’s response.) ChatGPT then continues with its first big header, “What You Did Very Very,” beginning, “Honestly, you The Three Pillars of Anti-Aging Skin Almost exactly.”
2. Overly perfect grammar and punctuation
If you read a lot, you understand the concept of creative (or artistic) license. This is the idea that writers break grammatical conventions or rules of effect, especially in fiction or poetry. For example, a writer may use a run-on sentence for emphasis. or sentence fragment. Furthermore, most people do not have an encyclopedic knowledge of grammar rules, so they will make mistakes. If you look closely, you can find many examples of extra (or missing) commas in everyday writing. On the other hand, AI chatbots rarely make these kinds of mistakes, unless you explicitly tell them to.
In testing, I asked ChatGPT An independent poemwhich means he has full permission to violate every grammar rule under the sun. After all, most of my poetry reads like grammatically correct sentences with only line breaks. On the other hand, take a look A free verse poem by EE Cummingsnone of which read like normal prose. Of course, not every free verse poem is as experimental as Cummings’ poetry, and not every instance of AI chatbots has perfect grammar. However, if you notice that a long Reddit post doesn’t have so much as a grammar problem, for example, that’s a red flag.
3. Routine inclusion of M dashes
You use the em dash when you want to break up a sentence for emphasis, as opposed to the short n dash you use to describe a range like 5-10. No — there’s nothing necessarily grammatically wrong with using em dashes, but chatbots historically include them in their responses much more than the average person.
(Credit: OpenAI/PCmag)
When I asked ChatGPT if I could recycle my spent water filters at one point, I got the following response: “Short answer: No — don’t put Zero Water filters in your normal recycling bin.” Now, seeing a single m-dash doesn’t mean something is generated by AI (I like to use m-dashes sometimes), but if you see a lot of them, it’s a potential cheap one.
4. Repetitive phrases or words
AI-generated text is not new, so there are well-known phrases and words that AI uses. You can focus your conversations with chatbots to get a sense of their favorite sayings. Consult a listof which there are many.
Some examples of repetitive AI words are “harness,” “light,” “pivotal,” “realm,” and “underscore,” while examples of popular phrases are “at its core,” “create,” “what’s being said,” “simply put,” and more. Of course, these are very common words and phrases, so there’s no point in looking one up. But if you see many of them in one article or post, it’s a sign that what you’re reading might be from a bot.
Get our best stories!
Your daily dose of our top tech news
By clicking Sign Up, you confirm that you are 16+ and that you agree to our Terms of Use and Privacy Policy.
Thanks for signing up!
Your membership has been confirmed. Keep an eye on your inbox!
5. Over-reliance on lists, headings, and structure
Ask ChatGPT, or any other chatbot, a question, and the answer you get includes bullet points, various headings, and a summary or conclusion section, along with similar formatting elements. Such a structure makes sense in an article like this, but it feels too out of place in other contexts.
(Credit: OpenAI/PCmag)
For example, if I ask ChatGPT if Pokemon Pokopia Worth Buying, it begins with a short answer, then sets out the following topics (with bullet points within each): “What Pokémon Pokopia actually is,” “Reception (so far),” and “When it’s worth buying.” It then closes its response with a short concluding section. This is a very different type of writing than what you would get if you asked a friend or watched a video that covered the same topic.
6. Lack of unique content
AI chatbots run. Major language models (LLMs) that are trained on vast amounts of pre-existing data. Often times, chatbots also search the Internet for the latest information to inform their responses. As a result, when an AI tells you something, it simply reproduces what other people have already said, without forming an opinion of its own.
If you’re reading a blog post reviewing Pokémon Pokopia, but it only includes the general information you’d find when talking to ChatGPT about the game (which itself is compiled from various reviews and previews of the game), that could be a red flag. But if the blog post has personal stories or thoughts on the game, it’s less likely to be AI.
Recommended by our editors
If you don’t want to do the work yourself, no problem, as a ton of online tools can analyze text and tell you how likely it is to be AI-generated. As an example, I submitted the poem I created with ChatGPT above. Copy leaksan AI detection tool. He told me that there is a 100% chance of my poem being generated by AI.
(Credit: CopyLeaks/PCMag)
However, keep in mind that these tools are not perfect. I played the same poem Zero GPT And QuillBot’s AI detectorand both failed to detect that it was produced by a bot. If you’re going to trust these types of tools, I recommend using a few different tools to cross-check their results. Even then, you cannot fully trust these services.
Look for patterns, not a single thing.
Every single attribute I’ve identified above shows up in the legitimate writing of real people every day. Some people prefer grammatical perfection or highly structured writing. However, if writing something online does most or even all of the things I’ve called for, it’s highly likely that a human didn’t write at least parts of it.
However, don’t try to avoid the above features when you write something. Not only does it look potentially awkward, but it’s just unnecessary. It’s surprisingly hard to replicate the feel of AI-generated text, and the telltale signs change all the time. When you write, just focus on being the best writer you possibly can be.
Just like spotting AI images or videos, spotting AI text is a skill that takes practice. Additionally, AI models are constantly improving, so clocking AI-generated content will become increasingly difficult. Don’t sweat it if you mistake something for real for fake, because it’s just inevitable in the AI age. I will keep updating this article as new trends emerge.
About our expert
Ruben Cercelli
Author, Software
Experience
I’ve been writing about consumer technology and video games for over a decade in a variety of publications, including Destructoid, GamesRadar+, Lifewire, PCGamesN, Trusted Reviews, and What Hi-Fi? At PCMag, I review AI and productivity software—everything from chatbots to to-do list apps. In my free time, I’m probably cooking, playing a game, or tinkering with my computer.
Read full bio
