- Oct 10, 2009
- 1,326
- 583
If you're prompting your favorite AI like this: "Give me a 1,000 word article on X" you're going to get back crap like this:
"In today's fast-paced digital landscape, content is more important than ever…"
Editing that fluff out takes longer than writing from scratch.
The problem isn't the AI.
It's that you're handing it a vague topic and letting it decide on where to go from there...
So what does it do??? Pads sentences with nonsense, talks in circles, basically, fills the article w/ fluff until it finally hammers out that 1200 words you asked it for.
The fix is simple: stop giving it a topic + word count, give it a mapped outline!
The difference is like giving someone a map of exactly where to go and a general direction of where to walk.
Here's the whole method. Takes about 5 minutes.
Step 1: Get the entities, not just keywords.
Entities will upgrade your content because search engines (and the LLMs behind AI Overviews) think in entities..... the specific people, concepts, products, and attributes tied to a subject.
So before you prompt anything, get the entity list for your topic.
An entity map can help prevent obvious topical omissions and make an article more comprehensive.. while having a map like this itself isn't a ranking formula... it does help! You will still need usefulness, clarity, and some original information to really make sure your article will do well... but you can think of an entity map as a starting point to hedge against fluff.
There's a free tool called entityexplorer.com that will help you in this process... basically, you type in your topic and it builds a mind map of the related entities/subtopics pulled from Google.
Build out and prune your map to the entities that actually matter for your article...
Once you're done, export it to JSON format. This is very important: JSON is great for feeding to AI because it's highly efficient machine-readable format.
Step 2: Feed the map to the AI as the outline.
Paste the JSON into ChatGPT/Claude with a prompt like this:
Add your original info, citations to other sources, basically anything to differentiate the article to make it a bit better.
Because the AI is working from a fixed list of entities instead of guessing, you get dense, on-topic copy with almost no filler. You're now editing facts, not deleting fluff.
Why it works: the fluff comes from ambiguity. Remove the ambiguity.... give the model the exact entity coverage up front....and it has nothing to pad with.
Bonus: you end up covering the same entity neighborhood the search engines reward, so it's better for rankings and reads cleaner.
Try it on one article and compare it to your usual prompt. The difference is stupid.
Here's an example of an entity map generated for BHW... These take about a minute to gen and make a big difference to your content.
"In today's fast-paced digital landscape, content is more important than ever…"
Editing that fluff out takes longer than writing from scratch.
The problem isn't the AI.
It's that you're handing it a vague topic and letting it decide on where to go from there...
So what does it do??? Pads sentences with nonsense, talks in circles, basically, fills the article w/ fluff until it finally hammers out that 1200 words you asked it for.
The fix is simple: stop giving it a topic + word count, give it a mapped outline!
The difference is like giving someone a map of exactly where to go and a general direction of where to walk.
Here's the whole method. Takes about 5 minutes.
Step 1: Get the entities, not just keywords.
Entities will upgrade your content because search engines (and the LLMs behind AI Overviews) think in entities..... the specific people, concepts, products, and attributes tied to a subject.
So before you prompt anything, get the entity list for your topic.
An entity map can help prevent obvious topical omissions and make an article more comprehensive.. while having a map like this itself isn't a ranking formula... it does help! You will still need usefulness, clarity, and some original information to really make sure your article will do well... but you can think of an entity map as a starting point to hedge against fluff.
There's a free tool called entityexplorer.com that will help you in this process... basically, you type in your topic and it builds a mind map of the related entities/subtopics pulled from Google.
Build out and prune your map to the entities that actually matter for your article...
Once you're done, export it to JSON format. This is very important: JSON is great for feeding to AI because it's highly efficient machine-readable format.
Step 2: Feed the map to the AI as the outline.
Paste the JSON into ChatGPT/Claude with a prompt like this:
Step 3: Assemble and edit.
Add your original info, citations to other sources, basically anything to differentiate the article to make it a bit better.
Because the AI is working from a fixed list of entities instead of guessing, you get dense, on-topic copy with almost no filler. You're now editing facts, not deleting fluff.
Why it works: the fluff comes from ambiguity. Remove the ambiguity.... give the model the exact entity coverage up front....and it has nothing to pad with.
Bonus: you end up covering the same entity neighborhood the search engines reward, so it's better for rankings and reads cleaner.
Try it on one article and compare it to your usual prompt. The difference is stupid.
Here's an example of an entity map generated for BHW... These take about a minute to gen and make a big difference to your content.