Hello there and welcome to the world of AI-powered content creation! In recent years, especially in 2022 and 2023 the field of natural language processing has witnessed remarkable advancements, leading to the development of increasingly sophisticated language models. Among these, ChatGPT3.5 and ChatGPT4 stand out as two powerful tools for generating written content – without doubt they made the process of article writing more efficient and accessible than ever before. But do they? Are we able to write one prompt – true, a bit richer in words – and the ChatGPT 3.5 or 4 spits out optimized article? No editing needed? Are we there already?
In this comparison, we will delve into the capabilities and differences between ChatGPT3.5 and ChatGPT4 (ChatGPT4 plus subscription and not ChatGPT4 API) to help you understand which one might be the better choice for your specific writing needs. Whether you’re a content creator, marketer, or simply someone interested in exploring the cutting-edge technology of AI writing, this article will serve as a valuable guide to aid you in making an informed decision. Let’s embark on this journey to uncover the nuances of these AI writing models and discover which one reigns supreme in the world of article creation.
1. Prompting to Create Table of Content
So, the first step in the process of creating articles in usually creating Table of content – this way you shortly draft out what your article’s focus will be. Let’s do this. As you can see in the picture below my prompt to instruct the ChatGPT3.5 and ChatGPT4 was:
Create a table for content for the article “Best dog food in 2023, according to pet experts.” The table of contents should contain at least 3 main titles and if needed a few subtitles to each of the main titles. Don’t include an introduction and conclusion in the table of contents.
Let’s take a look at results:
Above you can see my prompt is pretty simple, I want GPT to create a Table of contents for an article about dog food, with at least 3 main titles and if needed subtitles. I don’t want GPT to include either introduction or conclusion. So, who is was better? ChatGPT3.5 or ChatGPT4?
chatGPT3.5:
it provided Table of content for Dog Food
it provided 6 titles (the requirement/prompt was at least 3!)
it provided subtitles for each of the main titles
it provided conclusion in the Table of Content, but I asked it not to!
chatGPT4:
it provided Table of content for Dog Food
it provided exactly titles (the prompt was at least 3!)
it provided subtitles for each of the main titles
it didn’t include conclusion in the article, just as I asked!
All in one, in this case scenario when we wanted to generate the Table of Contents – the GPT4 did follow prompt more to the point. It created exactly 3 main titles (while asked to create at least 3), and it did not include introduction and conclusion into the TOC (Table of content) draft, compared to GPT3.5 which ignored this part of the request.
Content-wise, which is a subjective matter, especially if you don’t have a clue about dog food, I do believe GPT3.5 was more down to the point with what I wanted. I wanted a recommendation article about best dog foods and GPT3.5 did create in second chapter “Criteria for Evaluating the Best Dog Food Brands” and then in the third chapter a suggestion for “Top Picks for the Best Dog Food in 2023” and then it did list empty list of Dog Food Brands (it is up to the writer which Dog Food brands will be included). So, by my opinion, this was the heart of the article’s TOC. GPT4 on the other hand covered brand section only in the second chapter, however, it did also focus on age related nutrition as well as special dog food requirements.
2. Prompting to create an article based on suggested Table of Content
OK, so we have Table of Contents, both look decent and they both cover approximately what we want. Next thing we want to do, is make ChatGPT to write an article for us. The prompt we used was:
Based on the Table of Contents you’ve suggested, create an article that contains at least 1500 words. Add h2 tags for main titles in the article and h3 for the subtitles. If you bold words use strong tag.
So, my requirement is to create at least 1500 words, to use h2 and h3 tags and strong tag instead of just bolding the words.
ChatGPT3.5 created approximately 1200 words, however, it did stop at approximately 600 words and asked me if I want it to continue generating. I had to confirm I want it to continue generating the content. At the end it produced 1200 words, and it included all the tags as I requested. As you can see it also includes the
tag, for which I didn’t ask (it happened in the past too, but when I asked it to remove the
tag, it always did.)
What about GPT4? 503 words ladies and gentleman, and not a word more. And this was what the Chatbot said:
Please note that due to length limitations in this platform, the article might be slightly shorter than 1500 words, but it will follow the structure and format you requested.
Slightly? Slightly would be 1300 words instead at least 1500 as requested – not 503 words. In this context the wold “slightly” is definitely an overestimate. 🙂 In addition to this, the content it provided was more like an expansion of initial TOC, and not an article – it basically expanded the TOC into 500 words, but yes, just as chatGPT3.5 it did follow all my instructions regarding html tags.
When I tried to navigate GPT4 to “go on” or “expand content”, it provided about the same amount of words. No major improvement from his side.
Here’s a screenshot of my task (I didn’t screenshot the entire chatGPT3.5 or chatGPT4 result, just the main portion so you get a feeling what I received):
In conclusion – is creating an article based on suggested Table of Content possible with ChatGPT?
Yes, absolutely, but not only with one prompt, because if you’re aiming at 1000+ word article, you probably won’t get there, either with ChatGPT3.5 or ChatGPT4 plus subscription. Even more, one would expect to get more content with paid GPT4 version, but that is not the case. So, what should you do? I would defenitely go step by step and make several prompts:
1. create TOC on the [insert topic]
2. create article based on the TOC you’ve suggested
And you’ll need to “human-review” the article GPT provided and if needed expand sections that are lacking or you think deserve more attention.
You can expand these sections simply by prompting something like that:
4. Suggest 2 Premium Dog Food Brands
… or whatever you think needs more content.
All in one, don’t expect the GPT4 to be better because you’re paying for it, even people who worked through API reported very similar problems (limitations in the word count output). For me, it was disappointment because I expected it to generate words more generously compared to GPT3. Unfortunately, it doesn’t – it has word (token) limitations and is not pushing itself to the limit, which is discouraging because at the end of the day, you’re paying for it and not getting anything more compared to the free version. The good thing is, even if you write with GPT, you’re still the one navigating the articles and not the bot. And you get this “advantage” with both free and paid ChatGPT version.