OpenAI released ChatGPT last week and opened it up for people to play with. Their intention was not for it to be a free AI article generator but comparing its output to some of the most popular paid AI article writers shows that it might just be the best one!
ChatGPT is positioned, as the name would imply, as a chatbot. It is capable of answering complex questions, remembering what you had previously communicated with it and is built on the improved GPT-3.5 (aka Davinci 003). It has made progress on a couple of the major problems with AI-generated content…
When I was playing with it I asked it not a question but instead to generate an article and it answered my question with a great blog post.
TLDR:
Instead of asking ChatGPT a question you can ask it to create an article on any topic.
The result is typically a ~500-word article that is as good as or better than paid AI writing tools.
You can currently use ChatGPT for free!
However, like other AI writing tools, the AI that was built and trained at Originality.AI can successfully detect if it was AI-generated.
Let’s say you are looking to cover a topic about “does AI-generated content rank in Google”.
Side note Google is saying that they can detect AI.
Your prompt could look like this.
A unique thing about ChatGPT is that since it remembers what you said to it and what it replied to you… you can simply ask it to expand on the topic further.
In total it provided ~1000 words of well-written content with 2 simple prompts 100% for free.
ChatGPT output is able to be identified by our AI trained to detect an AI writer easily.
Comparing the ChatGPT content to that of paid AI content writers like Jasper.AI, Article Forge, Copy.ai, Rytre.ai, Shortly.ai, WriteSonic, and the ~1000 others shows that ChatGPT is as good (maybe better) at producing articles or blog posts.
Let’s look at a few examples from several tools with the same title input and then compare the results and the ability for the content to be identified as AI generated using Originality.AI.
Title: What are the top 5 risks to humanity from AI-generated content?
Here are several tips for getting the most out of ChatGPT.
Say you ask ChatGPT “write a long article about the 3 best gifts for my wife” and it suggests jewelry as one of the suggestions you can follow up with a question asking it to exclude jewelry.
Can ChatGPT be asked to make its content sound more human and therefore less able to be detected as generated by AI?
Not yet ChatGPT but nice try!
Comparing the 3 results I have to say that the ChatGPT output is superior to that of Jasper or WriteSonic. In fairness I did not fully optimize either of the tools but I also didn’t optimize my questions to ChatGPT.
Overall I am incredibly impressed with ChatGPT and its ability to not just answer a question but to be used as a free AI content writer.
Most importantly ChatGPT like all AI has serious limitations still and caution must be used to fact-check and add value to anything ChatGPT produces: