Updated August 2026. Originally published June 2024.
Three articles on this site make the argument better than I can.
- How to Develop a Go to Market Strategy took me weeks in 2024.
- How to Build a Cybersecurity Content Strategy took me a day, and it is the better article.
- CrowdStrike vs SentinelOne also took a day.
Now go and count the citations.
The one that took weeks links to nothing. The two that took a day link to Gartner, to vendor documentation, to independent evaluation data from MITRE Engenuity. Dozens of sources between them, every one of them checked.
That is the point of this article, and it is the opposite of what most people assume. Working this way did not make the writing thinner. It made it better sourced, because the hours I used to spend hunting for a current, trustworthy reference are no longer spent that way.
The two processes, side by side
Here is what I recommended in 2024, and what I actually do now.
| 2024 (what this article used to say) | 2026 (what I do now) |
|---|---|
| 1. Pick a topic with search volume | 1. Document your own raw material |
| 2. Ask AI for an outline of the topic | 2. Ask for an outline of your material |
| 3. AI generates the first draft | 3. It interrogates you, section by section |
| 4. Edit the draft into your voice | 4. It combines your raw material with your answers |
| 5. Add your own examples | 5. It finds and checks the sources |
| 6. Fact-check | 6. You review every line and every citation |
| 7. Publish | 7. You add the opinions, then approve |
Look at step three in each column. That is the whole difference.
What I got wrong in 2024
Step three used to be: have AI generate the first draft, then edit it.
That is the mistake, and I made it for a year.
If the model writes first, everything after that is you editing something you did not think. You end up steering a draft rather than saying what you know. The output reads like a competent summary of what already exists on the internet, because that is exactly what it is.
I called this the difference between AI-generated content and content generated with AI assistance. I was right about the distinction and had the mechanics backwards.
Back then it could not have worked any other way. You could type your own experience into a chat window, but the output still came back as an AI spit-off. There was nowhere to put your material so the system could work from it.
The order that works
Now I do it in the opposite order, and the inversion is the entire method.
1. I document the raw material myself
Step by step, on a specific topic. Not a draft, and not an outline. The true raw story, including my own decision making at each point:
- What I actually did
- The mistakes
- What I learned
- Tools I tried, tools I chose, and why I rejected the others
- Metrics before and metrics after
This is the part nobody wants to do, and the only part that cannot be delegated.
2. I hand it over and ask for an outline
Not an outline of the topic in general. An outline of my material.
The difference matters. One produces the article everyone else would write. The other produces the shape of what I actually have.
3. It asks me questions, section by section
This is the step that did not exist before. It goes through the outline and interrogates each part:
- What was the number?
- What happened when that failed?
- Who pushed back?
- What would you tell someone starting this today?
The questions surface things I would have left out, because when you have lived something you stop noticing which parts are unusual.
4. It combines the raw material with my answers
By that point everything in the piece originated with me. The structure and the sentences are assisted. The substance is not.
Why this only became possible recently
The reason the old process was AI-first is that there was no alternative.
A general model has no access to your history. It cannot know what you tried in 2021 and abandoned, or what your metrics did afterwards, because none of that has ever been written down anywhere it can reach.
What changed is that I can now point the conversation at my own materials and my own knowledge base. My documents. My metrics. Five years of what worked and what did not.
The model is no longer guessing at my experience from the outside. It is working from the record of it.
That is why the questions in step three are useful rather than generic. They come from something that has read the actual material.
Give AI the admin work
Content production is mostly administration. That is the honest description of it.
Consider what actually fills the hours:
- Finding a current source for a claim you know is true
- Checking whether a statistic you remember from two years ago still holds
- Tracking down whether a vendor page still exists at the same URL
- Confirming a number before you publish it
- Assembling citations so a reader can check your work
All of that is real and necessary. None of it requires your judgment. It requires patience and time.
Hand it over.
What comes back is not an article. It is research material with citations attached, ready for you to review. And you do have to review it. This is not a step you skip, and it is still a serious amount of work.
The difference is that reviewing sourced research and adding your own opinion to it is much faster than doing the whole thing yourself, over and over, every time you write.
That is how a pillar article goes from weeks to a few days. Not because the writing got faster, but because the administration around the writing stopped consuming the calendar.
The part that is not negotiable
Nobody has a problem with AI being used to generate content. That is not the objection, and it never really was.
What search engines raise, and what readers notice, is authenticity. Whether there is a person behind the text who actually knows the subject.
Experience, expertise, authoritativeness and trustworthiness have to come from a credible human author. Someone who reads it, reviews it, makes additions, and approves it. Someone whose name is on it and who would be embarrassed if it were wrong.
In practice that means:
- I read every line
- I add the opinions, because a model cannot have one about my own work
- I check the research it brings me rather than trusting it
- I cut the parts that are accurate but that I would not say
- I decide when it is finished
That is not a loophole or a compliance exercise. It is the reason the article is worth publishing at all.
The same rule, in a different place
If this sequence feels familiar, it is because I use it on things that have nothing to do with writing.
When we moved a marketing team onto AI, the first thing I did was ask an SDR to document every step of his real workflow, including the steps he thought were too small to mention. When our leadership team wanted an AI strategy, I asked them to interview a small team and blueprint what actually happens before buying anything.
Both are in the first AI project most companies skip.
Document the raw thing first. Then let the system work on it. Then optimise.
Content is the same problem. Start with the model and you get the version of your topic that already exists. Start with your own record of what happened and you get something only you could have written, in a fraction of the time it used to take.
If you take one thing from this
Write down what you did before you ask for anything.
The mistakes, the tools you rejected, the numbers before and after. It will feel like a waste of an afternoon. It is the afternoon that makes everything after it yours.
Pro Tip for Using AI Assistance to Improve Your Content
#1. Converting Your Articles into a Workflow
You no longer have to spend hours planning and creating workflows from scratch to enrich your content. AI can understand and convert your existing article into a workflow and give you specific instructions on how to implement it rapidly. Here is the prompt:
Convert the following article into a detailed workflow diagram. Include the specific names of the shapes for each object as used in Lucidchart (e.g., Terminator, Process, Decision). Also, provide detailed instructions for each step in the workflow and ensure logical improvements where necessary: [insert URL or text]
I experimented with it for this article and here is the result:
#2. Converting your articles into video scripts
Your articles can be video scripts in a matter of seconds. If you want something basic, you can simply use:
Convert this article into a video [insert your article text here]
As you get familiar with the basics, you can instruct AI to convert your articles into specific outlines such as:
Convert the following article into a video script and include:
- Grab audience attention
- Present key points in a logical flow
- Call to action
Take a Different Approach When Deciding the Topics You Want to Write About
The SEO game hasn’t just evolved; it has been fundamentally transformed. Bloggers who have built their content strategy around making complex facts digestible must acknowledge that even basic AI tools now perform this function. In the broader context, blogs are generally categorized into YMYL (Your Money Your Life) and Non-YMYL topics. Google defines YMYL content as topics that could significantly impact the health, financial stability, or safety of people, or the welfare or well-being of society.
If your blog is YMYL, it likely discusses topics based on facts from government and university sites. You produce content that is easy to read, well-structured, and thoroughly referenced. So, what’s the issue with being thorough? Nothing, except that AI can accomplish this in seconds and often more efficiently by delivering the facts from original references with decent readability output.
I believe focusing your content around your own validated experience will have the most rewarding results. When you create content powered by your own knowledge and expertise, you actually provide added value to the worldwide web.
This article was substantially rewritten in August 2026. The original June 2024 version recommended generating a first draft with AI and editing from there. I no longer do that, and the comparison above shows what replaced it. The material on repurposing articles into workflows and video scripts is unchanged from the original.
