I Use ChatGPT and AI to Help Me Write Blogs; Here Are My Best … – Business Insider

Jump to

This as-told-to essay is based on a conversation with Samantha North, a blogger and content writer in Portugal. Her job and income have been verified by Insider. The following has been edited for length and clarity.
I have two websites. I started Digital Émigré in September 2020 to provide resources for people who want to relocate to European countries with the long-term goal of getting second citizenship there.
I got the idea when I decided to leave the UK for Portugal because the UK left the EU.
My second website is my personal website, samanthanorth.com, where I teach people how to make money from blogging, search-engine optimization, and LinkedIn, as well as other kinds of digital marketing.
Last year, I made $115,000 from my blogs.
I have a background in journalism and content writing. Before I started using AI, I would always write my own blog posts.
I would start by reading articles online and doing research about the topic. I’d also draw on my personal experience of moving countries to think about what people might be searching for.
I’d confirm this using Semrush and Ahrefs, SEO-keyword research tools, to see how the keywords performed.
Those tools can tell you how much competition there is from other sites on a keyword topic.
Then, I’d put together an outline of the blog. Once I had that, I’d write the different sections of the piece with information I’d researched.
I’ve always been a bad manual typist and started using AI for voice typing.
Last year, I started experimenting with an AI tool, Jasper, for marketing content. I would input basic prompts, such as “write a blog post about” a subject. It felt like magic at first. But then I got used to its style and tone.
AI is great for interrogating my own ideas. When I’m thinking of a new direction to take a blog in, I can ask AI if it fits well with the existing content, and it will give a very well-rounded and useful answer. But I can’t just put AI content into my blogs because it wouldn’t sound like me. It’s too bland.
I started using ChatGPT this year, as Jasper focuses more on marketing content. I found it more flexible and versatile.
I subscribed to ChatGPT Plus, which is $20 a month.
I’ve found that ChatGPT hasn’t been able to replicate my keyword-research process. It can’t generate the exact keywords I need, with their volumes and their level of competition — at least, I’ve not discovered any of that yet.
But it’s useful for “topic clustering” — generating ideas on or related to a topic. For example, if I wanted to use AI to generate ideas for blogs related to Portugal, I’d create a prompt asking it to generate a list of semantically related keywords.
They might not be the exact keywords people are searching for. But seeing them generated helps open my mind to things I may not have thought of. Then I can start plugging those ideas into the keyword-research tool to find one of the specific keywords that people are looking for.
In the past couple of months, I’ve started using Claude 2.
I find the interface simpler than ChatGPT’s. The text it produces tends to be a lot more like what I would write. It’s very clean and right to the point. I don’t need to intervene as much as I do with ChatGPT.
I believe it’s still really important to have personal experience and opinions in posts. I think Google favors that experience, expertise, authority, and trustworthiness in its algorithm.
I use AI to outline the blog and for subheadings, titles of posts, and meta descriptions.
More recently, I’ve started using ChatGPT to craft introductions for blog posts. I ask, for example: “Craft an introduction to a post about X, Y, or Z with a compelling hook, and make use of the copywriting framework PAS — which refers to problem, action, solution.”
Once I’ve written a post, I’ll input it into ChatGPT and ask it to write a summary of the piece and some actionable steps for the reader in the body of the article.
I trained ChatGPT on my previous blog posts.
If there’s a section of a blog that’s generic, such as a history of Portugal, I would probably get ChatGPT to generate that. Afterward, I’d edit it to make sure the facts were correct and add external links to it.
If it generates a section that is flowery and full of adjectives, I ask it: “Rewrite the text in a less flowery and more down-to-earth way.” That generally produces something in line with what I want.
I train it on my voice by feeding it three posts by me and asking it to describe to me the tone, voice, and style.
Then I ask it to generate the text in that style. It might need a few tweaks, but I’ve found that works well.
While using AI has given me the scope to produce more content, it still takes a lot of time to manually edit, regenerate, and fact-check it.
I don’t let AI do everything for me. I don’t think content entirely generated by AI is as valuable to the reader as AI content edited by a person. It lacks that human expertise and experience. I worry that the internet could become full of generic, AI-generated content. But I use it judiciously because I believe that it brings the best results.
If I’m writing something that requires a lot of personal expertise, or if I’m writing about something detailed or specific, I’ll still write it based on my own research.
It’s about finding the balance between using these tools and interweaving human experiences and data.
Read next
Read next