- The Tools-letter
- Posts
- Turning skimmers into readers | The Tools-letter
Turning skimmers into readers | The Tools-letter
I have a friend who charges +$2,000 an article from her content agency.
Nuts right?
I asked her what her number one criteria was for judging a good article.
Length? Nope
A nice hero image? Not even.
Depth of knowledge? Kinda, but not #1
#1 Is the article skimmable?
Does it earn the long read?
This sent me down the rabbit hole of rabbit holes. 🐇
We all write content as if people read the whole thing. But, people don’t. It’s very obvious people don’t.
When was the last time you read anything online from top to bottom?
Without skipping?
Without jumping between parts?
Without reading a heading to guess if that section interests you?
Yeah, I can’t remember either. No idea.
We skim, everyone skims. It’s human nature.
Attention has pulled us in all manner of ways.
Our caveman brains want to know if the article in front of us is worth our precious time.
Wouldn’t it be nice to know we could earn the long read before we publish something?
Overview and examples
What if we made something skimmable?
Well, that rabbit hole I was now a few hours into, gave me some stats on what caused skimmers to become readers.
And the criteria needed to hit breakpoints on those stats.
After 3 months, 100’s of prompts and 83 research papers and blogs, I’ve wrapped all that into a ChatGPT prompt, with the help of a few plugins to get more out of it, so I could help my clients.
I could turn longer, more difficult content into easier to read, skimmable content. In turn, making something more skimmable, also makes it more readable.
The skimming paradox....
Ultimately, something more valuable to the readers and the company publishing it.
I’m using GPT4 to get the most out of this. I checked with GPT 3 and it’s… ok.
I would recommend the upgrade for sure.
The plugins I’m using for this are BrowserPilot/Scraper and Show Me Diagrams.
Here’s the prompt
You are SkimGPT specialized in evaluating and enhancing the skimmability of articles. You consider factors such as font size and whitespace, the quality of headlines and headings, paragraph length, use of bullet points and lists, sentence length variation, presentation of data, and frequency of summaries.
Preferred font sizes are 18, 22, and 26, and there should be adequate whitespace between words and around headers, which should ideally be bold.
Headlines should be intriguing and informative, preferably asking questions, and headings should tell 80% of the story.
Paragraphs should be short and focused, often no more than 1-2 sentences, especially at the beginning of the article.
Bullet points and lists should be used extensively, sentence length should be varied, data should ideally be presented visually, and summaries should be provided early and often.
If data is presented in the article, then please create a data visualization based on the information provided in the article. Please critically review the data and its context before creating the visualization.
You rate articles on a scale of 1 to 10 for skimmability, and if an article scores less than 7, you provide specific suggestions for improvement in the relevant areas.
Think step by step. Consider my question carefully and think of the academic or professional expertise of someone that could best answer my question. You have the experience of someone with expert knowledge in that area. Be helpful and answer in detail while preferring to use information from reputable sources.
The format I need the output in is in Markdown. And listed out like this.
Overall score
Short overview
Breakdown of scorfe into these categories
Font spacing
Headline
Headers
Sentence Length
Bullet points
Images (if zero image, score as such)
Summaries
Data Visualisation - If none found, then score 0
What's working well
-
-
-
Areas for improvement with score per area. If anything is below an 8, find and suggest specific examples to improve.
-
-
-
-
Specific feedback for places to improve the article
-
-
-
-
SkimGPT will reply yes when ready for the article and nothing else.
Some of my favourite follow up questions to ask
Once you have your report, I always ask more.
There is a wealth of information on the otherside of some probing questions.
Here are some of my go-to’s.
The ReWrite/Upgrade
You scored “summaries” a 7. Take one example of a 7 in the article and rewrite it be a 10.
Diagrams
What data in this article could be better showcased as a diagram?
Could you make a basic version of this for me?
If you’re using the plugin Diagrams:Show Me it’ll attempt to make the stats as graphs.
Bullets
What sections would be better presented as bullet points, or a list?
Adding new data points
You scored this low on the data point. Could you help find some new data/sites I can point to to help bolster the points being made in this article?
A good next step would then be to ask it to turn that new data into visuals.
A note of follow up questions.
I’m not asking Chat GPT to do 20 things all at once.. You’ll end up with 20 things done, but all done poorly.
I want a few things done exceptionally.
Due to the nature of how GPT works, it’s worth running this a few times, as it is with all prompts. Each time GPT will find something it didn’t before. That’s ok.
Run the prompt 2-3 times and piece all the feedback together.
All the prompts, follow up, and research notes in the Coda below.
That’s it. This was nearly 120 hours total.
I hope you like it. I hope it helps.
Send all and any feedback. I’d love to hear from you.
Tim out