Do I need a separate rubric for every type of content?

Yes. The audience and the format both change what good looks like, so an article, a landing page and a LinkedIn post each need their own. I keep specialised rubrics for all three.

What does "humanise the AI" actually mean?

You can usually feel the default AI voice within a sentence or two. The rhythm is too even. The sentences come out the same length, the paragraphs all weigh the same, everything lined up like fence posts. Humanising means handing the AI rules to break that pattern. Vary the sentence length, cut the filler, commit to a point instead of hedging, and give it the list of tells to avoid, such as em dashes everywhere, the "not just this, but that" trick, and three adjectives doing one adjective's job.

Why aim for 80% human rather than 100%?

Because 100% does not exist, and chasing it only eats time. Nobody can tell 80 from 95, whereas anyone can tell 40 from 80. I once spent a whole afternoon on the final 15% and realised the piece felt worse for it. Get to about 80 and stop.

Is this about tricking people into thinking a human wrote it?

No. Machine-sounding prose is simply harder to get through, and people give up on it faster. The 80% target is there to keep readers with you, not to fool anyone. Nobody likes AI slop.

How do I find the gaps in a finished piece?

It takes one prompt. Ask the AI what a reader would still be wondering after the last line, where you raised something and walked off without dealing with it, what example would turn a vague bit concrete, and what a sceptic would want to point out. Half the time the answer is staring at you the second you see it, and you would have missed it on your own.

Do I need special software, or will ChatGPT or Claude do?

ChatGPT or Claude will do the job. It is a bit more manual that way, but still far faster than writing and reviewing by hand. I have built my own rubric generator, knowledge base, Q&A generator and review analyser because I am a developer, but none of that is required to get started.

How long does this actually take?

This article ran to over 800 words and took 20 minutes using the process. The time drops further once your knowledge base is populated, because you can reuse it and let the AI find what applies in areas you might have missed.

How to Create Great Content in 2026 with AI and Process

Q: What about SEO and keywords in 2026?

Leave the keyword stuffing alone. It stopped working years back. What ranks now, and what AI assistants pull from when someone asks them a question, is whatever is most useful. The page that answers the question best wins, so write that page and the rest follows.

Q: Do I need to be a good writer for this to work?

No, and that is the whole point. In most cases a good writer will still do a better job, but competence is no longer the edge because AI gives everyone that. What separates good content from the forgettable kind in 2026 is a process you can repeat and turn into a habit. If you are not producing content now, it is because you have not built the process, not because you cannot write.

Don't get me wrong. In most cases a good writer will beat this process hands down. But I'm average at best, so that point is moot, and it leaves me one option: embrace what AI can do to help. I'm not alone there. Most people are in the same boat.

So competence isn't the edge any more. AI means everyone has it now. The edge is whether you've got a process that turns out good work every time, and whether you can build it into a habit. That's what I'm doing. I run a business, and I've never found time to create content. I enjoy it, but it always slides down the priority list. My fix was to solve the problem with a three-step process, then repeat that process for every piece I write until it becomes a habit.

The three steps:

Content rubrics
Structure and Q&A
Make it useful

I've put time into this because several of my projects have a real need for it. My customers have to produce hundreds of pieces of content a month. By hand, the cost would be absurd, and they don't have the hours anyway. So they follow my process, and it works for them.

This article is the proof. It took 20 minutes to write this way, it runs to over 800 words, and it should hand you something you can use with your own AI today. I've left out some of the nuts and bolts, but there's enough here to go and experiment. If you want more, ask me.

1. Content rubrics

This is the step most people don't know they're missing. You get handed a title and a summary, and off you go. But something's absent. You've no way to measure whether the article hits its goal and connects with the right people.

Decide what good looks like before you write a word. Skip that and the whole thing is hit and hope. You judge the draft against whatever the AI hands back, which is always plausible, and plausible reads fine. It also tends to say nothing. You won't catch it, because you've nothing to hold it against. So you shuffle a sentence, cut a line, swap a word, and never know whether you're improving the piece or just poking at it.

Write the standard down first. Who it's for. What it has to do for them. What makes it credible. What it must never do. For a piece like this one, the rubric says it has to make sense to someone who's never written a word with AI, it can't drift into jargon, and every claim has to be something I'd defend out loud. The more specific you get, the more use it is, to you and to the AI you hand it to.

People skip this because it feels like admin before the real work starts. It's the opposite. The rubric kills off the three rounds of editing where you fiddle with words and can't say why. I've shipped pieces without one. They read fine and landed flat.

One thing worth saying: you need a rubric for every audience and every format. I keep separate ones for articles, landing pages and LinkedIn posts.

2. Structure and Q&A

Build the outline you want the article to follow. Fill it with the questions and answers the piece should be built on. Use that to write the draft, then measure the draft against your rubric. Add the things AI can't know: your testimony, your examples, your outcomes. I run a knowledge base to speed up the personal bits, because once it's populated I can reuse it, and AI is good at surfacing relevant material I'd have skipped.

Then get AI to write it with a strong humanising prompt. You can feel the default AI voice inside a sentence or two. The rhythm's too even. The sentences come out the same length, the paragraphs weigh the same, everything lined up like fence posts. Most readers can't name what's wrong, but they sense it, and they trust you a little less.

Fixing it comes down to rules. Vary the sentence length. Cut the filler. Make it commit to a point instead of hedging round it. Hand it the list of tells to drop: em dashes scattered everywhere, the "not just this, but that" tic, the leverage-and-seamless vocabulary, three adjectives doing one adjective's job. Once it knows what to avoid, it mostly stops.

Then use AI to score how human it reads. Aim for about 80%. There's no such thing as 100%, and chasing it only burns time, because nobody can tell 80 from 95. Anyone can tell 40 from 80. I've spent a whole afternoon on the final 15% and realised the piece felt worse for it.

This isn't about tricking anyone into thinking a person wrote it. Machine prose is harder going, people give up on it faster, and that's the real reason to reach 80%. Nobody likes AI slop.

3. Make it useful, not more SEO-ish

By now the piece is good. Don't start cramming keywords in. What ranks in 2026, and what AI assistants pull from when someone asks them a question, is whatever's most useful. The page that answers the question best wins. Keyword stuffing died years ago, and the search engines and the AI tools now want the same thing you should want anyway: the most helpful answer going.

So find the gaps. It takes one prompt. Ask the AI what a reader would still be wondering after the last line. Where you raised something and walked off without dealing with it. What example would turn a vague point concrete. Where a sceptic would push back. Half the time the answer's obvious the second you see it, and you'd have missed it on your own. AI reviews well when you give it clear rules to review against.

None of this asks you to be a gifted writer. It never did, and now it matters even less. What separates good content from the forgettable kind in 2026 is a process, and a process is something you control. So if you're not creating content regularly, it's not your writing holding you back. It's that you haven't built a process you can repeat, and turned it into a habit.

My results

I use this for everything written on this site. You'd struggle to find the seam between what came out of my head and what came from AI, and I'm happy with that. I'm a developer, so of course I've built my own rubric generator, knowledge base, Q&A generator and review analyser. I've even thought about turning it into a small hosted tool people could use. You don't need that, though. ChatGPT or Claude will do the job. It's more manual that way, and still far quicker than writing and reviewing every word by hand.