Can People Tell the Difference Between Human and AI-Generated Text?

The development of artificial intelligence has revolutionized the world, and its impact on creativity has been significant. AI text generators are now capable of producing content that is nearly indistinguishable from human-written text.

The development of artificial intelligence has revolutionized the world, and its impact on creativity has been significant. AI text generators are now capable of producing content that is nearly indistinguishable from human-written text. I did an experiment to explore how far AI-generated text has come and whether people can differentiate between AI-generated and human-written content. The results of the survey, which included ten questions, proved the hypothesis that most people cannot consistently tell the difference between AI-generated text and text written by humans. The implications of this are significant as it raises questions about the authenticity of content created by AI and the need for measures to help people identify AI-generated content. This experiment demonstrates the potential of AI to generate quality content, but it also highlights the need for greater awareness and regulation.


Hypothesis

I hypothesized that the majority of people cannot tell the difference between human and AI-generated text. I believe this because, after using AI-text-generation tools, I have noticed that the information they output is quality and often indistinguishable from the human-written text.


Results and Analysis

The results of the experiment proved my hypothesis. After surveying 205 people with ten questions (six written by ChatGPT 3.5 and four written by a human), the average score was 44.2%, a failing grade. The highest score was 80% or a B. There was also a significant difference between personal and non-personal texts. Every personal question (for example, my day at the beach with my dad or my experience with migraine headaches) had twice as many responses saying they were human-written than non-personal questions (for example, how to protect your iCloud data and a recap of the 9/11 attacks). In short, most people cannot consistently tell the difference between text written by artificial intelligence and by humans.


Real-World Applications

This experiment demonstrates how far artificial intelligence has come. Artificial intelligence can generate text, images, and even audio. This is particularly powerful as now it is feasible for anyone, no matter their experiences or education, to create authentic, quality content. It is key to familiarize people with identifying AI-generated content. We can accomplish this through government regulations or companies creating software that scans for this content.


Graphs

You can read the questions associated with each of these graphs using the button below!

Artificial Intelligence-Generated Questions

9/11 Attack News Story
Incorrect Haiku – When asked to create a haiku, ChatGPT messed up the syllable count.
Motivational Quote
The Struggles of Chronic Pain
COVID-19 Breakthrough
Beach Trip With Dad

Human-Written Questions

Haiku
iCloud Account Data Protection – This Question Is Actually an Excerpt From One of My Articles!
Migraine Struggles – This Question Is Actually an Excerpt From One of My Articles!
Drag Race Death News Story

 I highly recommend clicking the button below to read more about the survey questions.