Can you leverage GPT-3 and artificial intelligence to improve your blog? And which software is the best for this?
What is GPT-3 and What Makes It Different?
GPT-3 (generative pre-trained transformer 3) is the most advanced AI for generating text available (via an API with company approval) in 2021.
The hype surrounding GPT-3 is a mix of star power (Elon Musk is the founder) and that “Holy crap” feeling you get when thinking about what AI can do now. That said, it’s not an entirely novel technology. GPT-3 was trained via the Internet (Using CommonCrawl and sites like Wikipedia). That dataset is an order of magnitude larger than competing AI’s and it’s what makes GPT-3 special at this time.
While it shows a lot of promise, it is still a little bit more hype than substance, in my opinion.
My Thoughts On Using Artificial Intelligence To Write Blog Posts
GPT-3 is NOT at a point where you can click a button and “Poof” you magically have a great blog post. In fact, if you try to do that, you’re very likely to experience the following problems.
- Content that doesn’t make any sense whatsoever.
- Content with dozens of extremely false statements. (The AI doesn’t seem to care very much about truth; it simply makes things up randomly).
- Content that isn’t enjoyable to read even if it feels like something a person might have written (because it’s using a lot of fluffy language with no meaning behind it).
Despite these problems, the AI is actually capable of doing a handful of very useful things.
- Generating ideas: Blog topic ideas, subheading ideas, etc.
- Generating Copy: It can take your headline and give you a better headline (improving your click-through rate in Google). It can also take your meta-description and write a better meta description (improving your click-through rate in Google). Or it can rewrite a paragraph you’ve already written to sound better.
- Generating Images: Holy Crap! I actually pretty badly want to test this out as it’s a time bottleneck on this blog. GPT-3 can generate an image based off its description! However, I’ve yet to find any commercially available software that lets me test this feature.
- Creates original text: The results it spits out will usually pass by plagiarism detectors like CopyScape.
While the AI isn’t capable of generating an entire blog post on its own, it can be quite helpful in assisting in certain areas. I feel like the AI is ready to act as a form of Intellisense to help writers make better blog posts.
With that said, let’s go over the commercially available software out there right now.
Best AI Software for Bloggers
OpenAI’s GPT-3 is what all this software is using to generate results. Thus, these options are just window dressing on the same API (although depending on how the various tools use the API they could generate dramatically different results).
Also of note, OpenAI currently has a waitlist to use its API. Meaning they haven’t given access to a large number of companies (software choices are limited). I wanted to code an app using OpenAI’s GPT-3, and have yet to be approved for access.
Here is OpenAi’s pricing model (See table below). The higher-priced options are the more advanced models, and a token equates to about 750 words of text generation.
|Model||Price per 1K tokens|
I bring these prices up because it helps when understanding the cost of the following services. To blow through $109 of tokens for instance (with the Davinci model), you’d have to generate roughly 1,362,500 words. That said, with the ADA model being almost 100x cheaper, it makes you wonder if these software tools are always using the best model.
1) Jarvis by Conversion.Ai ($109/m)
Jarvis’s $29 plan is likely good enough for your needs. The long-form content editor (where it generates an entire blog post for you) appears to be the major add-on when going up in price. I know that many people who are buying think of this as the killer feature, but I personally hated it. I thought what the AI produced was pure garbage most of the time.
However, the software did have a few features I quite enjoyed. The long form editor had a feature where it would generate title ideas for your blog post. I thought this was great.
Another feature I loved was its content improver. You hand the AI a paragraph and it will rewrite it for you in multiple ways. I found it very helpful for brainstorming purposes.
Jarvis also has a feature to create blog post outlines. And while I didn’t find the outlines themselves to be very helpful, sometimes one of the bulletpoints was something I hadn’t considered. Something that definitely should have been in my post.
It also had the ability to generate “Title and Meta Descriptions.” However, when I tried it it was only generating unique meta descriptions and let my title unchanged. Kind-of a bummer.
Also, when I told the AI to generate blog post ideas it wasn’t very encouraging. It did come up with topic ideas related to my niche, but it was always the most obvious ones (that would be impossible to rank for). “50 ways to make money online” is not exactly an original idea.
I managed to find several other AI softwares out there that promised to do the same thing as Jarvis. Unfortunately, I’m not made of money and couldn’t try out every single AI product as in-depth as I tried out Jarvis.
However, since they all use GPT-3 on the back-end, I would expect similar results. Well, unless one uses ADA and another uses Davinci to get its data. Or one just has a vastly superior UI.
3) WriteSonic ($10/m)
4) Copy.AI ($35/m)
5) CopyShark.ai – ($89 per month)
6) Market Muse ($79/m to $1,000/m) – This one is insanely expensive.