Turns out these AI Content Writing tools aren’t as random as us humans
So What is AI content?
Before we delve into how to detect AI content first we are going to have a look at what it is. AI content is anything that is created using Artificial Intelligence (AI). This type of content is starting to litter the internet in forms of articles for webpages. Other versions of this are images known as AI-Generated art like the images found on our website and videos.
The way AI content is created is by using complicated algorithms and datasets that are programmed by user input/data in the from of “prompts”. You may have noticed the term “prompt engineer” online, you can see more about prompts in our Stable Diffusion Cheat sheet.
Artificial Intelligence is all around here is a quick breakdown of places you can find AI used today:
- Text: Commonly found in articles and of course the well known chatbot responses (Like googles bard).
- Images: Photographs, illustrations and what is known as AI-generative Art
- Voice: Text to speech where AI is trying to sound more human like
- Video: Although in it’s early days AI can create some basic videos on a set of prompts.
- Code and Programming: AI has been helping some people create projects with no experience with coding before.
For this article our main focus is on AI generated Text and how you can notice someone is using it. Please be aware we are not against the complete use of AI when it comes to creating content. We however are against people who only use AI for content.
So let’s Ask A bot What they Think About Detecting AI Content
Ok, ok I know it sounds rather odd. Why on earth would we ask a bot to tell us what AI content is and how to detect it? Well let’s face it why not ask the tool that’s creating the content and find out what it thinks. Then we can poke holes in it’s response and see if we can come to some kind of level ground.
This response took only a few seconds to generate and to be fair the response form ChatGPTs bot is quite reasonable and does give some key points for us to compare when looking at content. Let’s break this down and analyse things a little more closely.
- Lack of personal touch: AI-written content may lack a personal touch, as it is generated by algorithms and does not reflect the individual writing style or voice of a human writer.
From experience with writing articles for the past few years, this one holds quite true. Majority of AI written content is quite cold. You don’t feel like your are engaging with someone that has written content. Content tends to be written in a statistically manner with very little human emotion. Of course over time this may improve but for now it’s something to look out for.
2. Consistency in style and tone: AI language models are designed to maintain consistency in style and tone throughout a piece of writing. This means that the content may lack the natural variations in language that are characteristic of human writing.
This a key method in finding out if something is written with AI. You will not notice any examples of anger, shock or surprise through an article written by AI. Not to mention not a whiff of dry humour which can be found in a large amount of articles written by humans. It’s a way to engaging with your audience and forming that contact. If we are not grabbing our readers attention then we have failed as writers.
3. Limited ability to engage in natural conversation: AI language models can be very good at generating text in response to a given prompt or question, but they may struggle with engaging in natural conversation or responding to unexpected follow-up questions.
This might be a little tougher to see depending on the article that is written. But a good rule of thumb is that if there is very little scope in a conversation and it tends to stick to the same point without expanding on the topic. Chances are there might be a little bit of AI behind that content you are reading.
4. Errors in logic or context: While AI language models can generate grammatically correct sentences and paragraphs, they may not always get the context or logic of a piece of writing right. This can result in errors or inaccuracies that a human writer would be more likely to catch.
We are human, we make mistakes. Simple spelling mistakes, typos or errors in our formatting. All Human. If you are familiar with a certain topic you might be able to spot some obvious inaccuracies that should have been caught out during the initial write up of the article. It’s not the number one thing to look for because even us humans make mistakes.
5. Repetitive phrases or ideas: Because AI language models rely on algorithms to generate content, they may sometimes produce repetitive phrases or ideas that are not particularly creative or unique.
Think to yourself. Is the article repetitive, does it keep repeating the same point over and over again. This might be acceptable for a young child but as soon as we learn and improve our writing skills we quickly realise it is not the ideal way of writing. Another key thing to look for is the use of Synonyms. When humans tend to write they don’t say everything is “good” they change it up a bit, “brilliant, excellent, amazing, great” etc. Look for the use of synonyms in the article if it just repeats the same words, chances are it’s a bot.
So let’s Break it Down for you
If you want to detect AI written content on your own you are going to want to look for the following:
- The article is written with no personality
- The Article is written in the same style and tone throughout no matter what the subject is.
- Lack of Natural Conversation
- Errors in the article. The logic and context do not make sense.
- Repetition in phrases no use of synonyms
Ok, this is not the end all be all way to detect AI but it’s a pretty good place to start. As AI improves it will be harder and harder to detect. But personally we believe that there will always be a place for writers in this world that create good engaging content.
There are also many tools online that can detect AI content for you, if you just want a quick analysis of someone’s page. Please be aware that some AI detectors may detect human written content as AI also, but this is normally down to a persons writing style being kind of robotic. For example this article hit 95% Human written content.