There's this lack of nuance in AI generated articles that gives em out. How do you fix that?
Anyone here use AI writing softwares for money sites?
- Thread starter SilverKnight
- Start date
Personally, with Sensit.AI, I haven’t had that problem that much. You feed the generator sets of keywords, a good title and description, and you’re good.
For extra assurance, you can write in your own paragraph outlines, but I always just have the AI generate them for me.
How do you mean?
I will go with Jasper any day.
- Joined
- Feb 21, 2012
- Messages
- 453
- Reaction score
- 349
GPT-3 vs a simple spinner (the AI part in WordAI is just for marketing purposes).
Can Jasper spin as well?
- Joined
- Feb 21, 2012
- Messages
- 453
- Reaction score
- 349
Afaik...no, it can't
Yes. It actually makes better content out of a given paragraph.
I do use Jarvis for creating content for my money site, I create around 20 articles a day, already at around 300 written, lengthen at 2500 words ok average (cup why not? It doesn't take long with jarvis to get there).
The content isn't solely generated by jarvis, but is fed from multiple sources and PAA's, and I can attest it is incredible.
The potential is quite high. I will report myself for the results (probably in a few months), when I finish writing for all of my keywords, and finish publishing everything and it'll start to rank.
I have no doubt it'll rank, I will do it with no backlinks at first to see how long can you take it.
Right now I have around 300 pieces written with jarvis and they're good.
Only jarvis and my fingers. (Google docs for keeping the articles because God knows if jarvis will ban the account or smth)
reaaski
Power Member
- Joined
- Mar 11, 2013
- Messages
- 635
- Reaction score
- 950
No. Better hire someone if you want to do it.
- Joined
- Mar 2, 2016
- Messages
- 112
- Reaction score
- 30
Definitely , but AI makes mistakes, you need to correct them
- Joined
- Jul 30, 2020
- Messages
- 105
- Reaction score
- 74
This is part of my plan at the moment, so I appreciate you sharing this.
I have a few questions if you don't mind?
1. Are you training OpenAi with your own "modules" (can't remember what they're called)?
It's quite cheap to train DaVinci-002 with only like 40% of the Token cost compared to generating content. I know it's quite a bit of work to create those hundreds of example, but I'm thinking it could really make a big difference in a niche?
2. When you built your tool, did you also include things like "improve readability" and similar features that tools like Jarvis has? Is that easily available from the API, or are those custom features that Jarvis and others have build themselves?
3. Same question in regards to input before generating content, how have you set this up? The more specific you are with your original input the better the content generated, so I'm curious how you optimise this for both speed and quality?
4. Did you code this tool yourself or did you find someone to do it for you, and does it have a user interface etc? I know it may be a lot to ask, and understand if you don't want to answer, but if you have someone you'd recommend for coding this I'd greatly appreciate it.
5. Does your tool automatically post to Wordpress, or do you still manually have to import the articles? This was a feature I want to build in my tool, that each blog post follows a certain recipe/formula and then pulls in highly relevant pictures from a database of pictures you've previously downloaded. So basically my idea is that you download say 200 pictures and select for them to be used within one "batch of 100 articles", so 2 pics per article. Then the tool posts it all directly to the blog.
I think I need to get going with this project. Originally I was planning to build this tool for myself, but I agree with you that there's a lot of money to be made to be made from selling these tools at the moment.
The key would be to differentiate yourself a bit from the others and build a bit of a moat so it's easier to keep your customers for longer. One way I was thinking of doing this was to go all the way from AI script to output in the form of a ready made video, but it increases the complexity and there are others working on this.
Anyway, thanks again for sharing and would appreciate any answers you'd like to give.
reaaski
Power Member
- Joined
- Mar 11, 2013
- Messages
- 635
- Reaction score
- 950
I built only one module. So I would say this is prealpha stage. I focus on mass content generation. You can either tell it to rewrite something or make it write from scratch. Of course, the first way will be more accurate and less random, but requires way more input and sometimes produces borderline duplicate content.
If I used the first way (giving it content), I'd do it manually (not scraping it, for example by copying it from top, user oriented content). The second option (writing by simple prompts) let's generate much more (turning one sentence into 250 words of text). The trick is to clone those sentences (prompts) and write as many similar articles as possible. So from 10 prompts replicated 100 times, you'd have 100 articles.
Use Deepl with node.js automation to get free translation (their paid api sucks). You can also tell openAI to translate it.
You don't need to build any features because prompt design which is built into it lets you use any kind of command. You can of course train it to serve for that specific case better.
Yeah. That's what it is about. User interface is useful to enter data and control it. It also depends on how you want to utilize it. If you want to build a product based on it, it'd a thing that let's the user enter hundreds of prompts and replicate them using replace features, so they can generate let's say 100 texts for seo purposes in 10 minutes instead of writing each text from scratch like it is in products available on the market.
if you go with scraping and stop controlling the input, it may work but most of the time output will be missed. Also, there are advanced topics it just can't deal with and there it's best to consider paraphrasing.
I always test my set of input prompts before I proceed to massive generation, so I know it will take as little work as possible. Then I clone them using a button in the interface and do it 30 - 50 times unless an array of similar topics is covered.
If you want to create recipes, that's a specific use case. I don't know how you'd plan sending prompts. This is mostly about figuring that out, so your text looks solid and not like random pieces.
Making it suck less is an engineering task. Most of these companies using it have some idea about it but nothing they did so far couldn't be replicated at home. In fact, this forum users provided already better solutions than those companies. They just had an early access to it. Now it's open to everyone. Aka money is the power. Power is money.
I use node.js for this because there are automation tasks within the project (mostly translation). You wouldn't want to use PHP for this.
OpenAi is very generalistic in nature. You shouldn't think of detailed use cases like sales pages or ad copy when using it. And I believe it's a bad cook too. That's one of the lessons I learned.
Last edited:
- Joined
- Jul 30, 2020
- Messages
- 105
- Reaction score
- 74
This is truly helpful to me, thanks so much for your detailed reply!
I guess I'll be on the hunt for a node.js dev soon then, but first I think I need to find a project manager.
Thanks again and best of luck with your project.
reaaski
Power Member
- Joined
- Mar 11, 2013
- Messages
- 635
- Reaction score
- 950
So good luck then. I'm bootstrapping things at home. It was great to realize how much stuff you can reuse and nobody will try to sue you for that reason.
- Joined
- Jul 30, 2020
- Messages
- 105
- Reaction score
- 74
Let me know and send a DM if you're at all interested in some sort of JV.
I have some resources (not related to Ai but generally, have been doing online marketing for 20+ years) and I have a little bit of funding set aside for this project.
I also have a lot of ideas of what type of features I think could be useful for this type of tool. Ideally start with an MVP version and then improve over time.
I posted in the JV section looking for someone, but for some reason my post was deleted, so I guess I must have broken some rule.
Anyway, I really appreciate you sharing your experience in this.
moneyxlovelyxx246
Newbie
- Joined
- Feb 7, 2014
- Messages
- 16
- Reaction score
- 1
Peppertype AI is pretty good.
