While this sentence wasn't written by AI, would you be able to detect if it had been? Many industries are facing quite a commotion with the release of AI tools such as OpenAI's ChatGPT and few reliable means to detect AI-generated text.

Industries are debating how much these generative AI tools should be used to create content, while consumers of this content are unaware if what they are reading was created by AI and how much it should be trusted.

What is AI-generated text?

According to G2, AI- generated text is "automated content processed by artificial intelligence tools. It's reinforced with the generative pre-transformer algorithm and large language models." Essentially, this means that AI tools such as Bard and ChatGPT simulate human intelligence to produce text that is arguably indistinguishable from human text.

What industries are using AI-generated text?

Virtually every industry has taken an interest in AI-generated content, from science publishing to small business retail. It can be used to write a short reply to an email you've been putting off, generate a listing for a home you can't figure out how to sell or create that science fiction novel you've had writer's block about.

It seems that most industries will use AI tools in some form. However, the question is how much they will use these tools and what kind of rules they will put in place concerning AI.

What are potential problems with AI generated text?

Some people are concerned about how much authenticity will remain in written content if AI-generated articles become mainstream, while others appreciate the developments in natural language processing. Many individuals are worried about the potential of the proliferation of false information or deceptive articles intended to manipulate public opinion since it is difficult to distinguish AI-generated content from human-written language.

AI can plagiarize

AI-generated content has been known to plagiarize since it learns how to write from already written texts. This can lead to copyright issues for industries who employ AI-generated text. Not only can AI plagiarize, but it can also generate text that has not been backed up with data. AI tools are designed to be straight to the point with no "frilliness." While being clear and concise can be beneficial, AI is not designed to make inferences based on text, so it might miss major concepts and connections.

A lack of charm and humanity

Another potential issue with AI-generated text's conciseness is that it lacks basic human charm. The way AI-generated content reads like a robot wrote it (because, well…that's what happened) can make it difficult for people to want to finish reading an article. Our attention spans have become so short that articles need to have entertaining aspects to grab and keep our attention, which AI-generated content cannot yet accomplish.

How to tell if text is AI-generated

Online tools are continually being developed with the sole purpose of detecting AI-generated text. These automated systems look for things such as how often common words like "the" repeat in an article, punctuation and sentence length patterns and lack of typos.

While online tools are built and continue to evolve, people can learn to detect AI-generated text through their own eyes. Common ways to detect AI-generated text include:

  • Lack of emotion and personality – AI is concise and unemotional, unlike most humans. AI is also still learning and hasn't picked up on most of our slang yet. If an article lacks personality and doesn't pull on your emotions at least a little bit, it might have been written by AI.
  • Repetitive language and patterns – Just how automated detection systems looks for repetition of words and phrases, people can look out for these tell-tale signs as well. AI has been known to use "keyword stuffing." If an article reads like it has tried to fit as many keywords as possible into one article, there is a likelihood it was written by AI.
  • Inconsistency – Writing that has inconsistent style and tone throughout an article screams that it was written by AI.
  • Lack of typos – Hannah Montana's "Nobody's Perfect" states, "Everybody makes mistakes, everybody has those days." As much as writers try to achieve perfection, it is likely that we will always make a mistake. AI isn't known for making typos, so looking for the lack of these small mistakes can be key in identifying AI-generated text.

Detect AI-generated text like a pro

AI is getting better at mimicking human language, so utilizing more than one of these detection techniques is more reliable than relying on just one. While AI detection tools such as Originality.ai or GPTZero can be used to determine if something was written by AI, it is important for people to learn detection techniques for spotting AI writing as well.

It seems that AI-generated text is advancing faster than we can keep up. It is imperative that we continue to develop AI detection tools and continued learning of AI-generated text before it becomes impossible to keep up with AI's seemingly never-ending development.