My Editing Process: How to Edit Blog Posts [+ AI Prompts]

In this guide, I will walk you through the editing process that helps me deliver ready-to-publish drafts. 

You will learn how to:

All steps come with AI prompts, ready to copy/paste. To help you surface issues methodically and save time — whether you edit someone else’s work or self-edit before handing the draft to your editor.

Here’s a video outlining the process.

My editing process starts with the outline. 

Because fixing major structure issues is easier — and faster — at the outline stage than in a polished draft.

So if you’re working with writers, ask them for a detailed outline before they start writing. If you’re a writer, ask for feedback on the outline before you crack on.

Yes, it may seem like an extra round of revisions, but it isn’t.

From experience, when I take time to refine the outline, I need fewer edits later. So I save time.

Here’s how I stress-test an outline before drafting. 

MECE is a consulting term I picked up from Ryan Law, and it stands for “Mutually Exclusive, Collectively Exhaustive.”

The idea is simple: no section should overlap with another (mutually exclusive), and together they should cover the topic in full (collectively exhaustive).

A MECE outline keeps the article clean, avoids redundancy, and ensures you cover every major idea.

Make sure you deliver everything you promise earlier in the article.

If the title, intro, or a header hints at an idea or benefit, the article must cover it.

Promised the reader will learn how to edit their blog post? You must give them an actionable step-by-step guide. Promised AI prompts? You must give them AI prompts. 

Unfulfilled promises are misleading or — in the best-case scenario — distracting. 

I aim for 70-80% of the article to be “how-to” content, with only 20-30% left for “what” and “why.” 

Most readers don’t need a detailed explanation of why their problem exists. Or why it’s important to solve it.

They want to know how to solve it.

But that’s where a lot of articles fail. Because, as writers, we don’t always have practical experience doing things we write about, so we skim over the “how” part and focus on the easier “why” or “what” questions.

When structuring my sections, I start with the takeaway, then back it up.

I BLUF it.

BLUF = Bottom Line Up Front. The main message or takeaway comes first, and the rationale, evidence, or examples follow.

Why it matters?

Readers are busy. If they don’t get the answer right away, they bounce.

And I hear that LLMs are more likely to pull your content into their answers if you BLUF your sections. 

At this step, I rewrite all headers into dense, benefit-focused ones.

For example, “Track in-app behavior for actionable insights on feature usage” is much clearer than “Track user behavior.”

Such high-concept headers tell the reader exactly what they’ll get from a section, even if they only skim the contents table.

Parallelism is when the headers follow the same grammatical structure.

If one H3 starts with a verb, they all should. If one is a noun phrase, keep them all as noun phrases. 

It may seem like a minor tweak, but it reduces the mental effort needed to process information, making the article easier to digest. 

The high-level edit ensures the big picture holds up: the arguments are sound, the facts are accurate, and the ideas flow.    

If I start editing with the draft (no outline), I run the outline checks covered above before I move on to verify facts, stats, assumptions, and claims.

If the outline has already passed them, I jump straight in.

Spotting inaccuracies, unsupported claims, unjustified assumptions, faulty reasoning, or poor arguments in our own work is hard. 

Fortunately, that’s where AI does a good job.

I mean, I still don’t trust the data it comes up with, but it’s pretty good at flushing out fallacies or missing evidence. 

To resonate with your audience, speak at their level.

Writing for beginners? Don’t use specialist jargon. For the senior folks? Don’t oversimplify or overexplain core concepts or you won’t keep them on the page. 

Getting this balance isn’t easy. 

AI can help here, but, ideally, ask an SME to check the articles. At least the first few you write for the target audience. They will know immediately if the article resonates and brings value or not.

One thing I pay extra attention to at this stage:

Statements that sound condescending, patronizing, or insulting. Ones that can make them feel stupid.

Because nothing alienates the reader faster than talking down to them.

And I look for a more neutral and empathetic way to make my point or persuade them.

For example, if I were to convince a team to stop gating all their ebooks behind forms, I’d start by acknowledging why they do it (lead-gen targets, pressure from Sales). Then I’d explain why that approach often hurts reach and brand trust.

And only after that would I suggest alternative ways to capture leads.

Once the structure is solid, reasoning robust, and stats straight, I move to line editing.

This is where I tighten the language and remove fluff.

So the final piece is sharp, engaging, and pleasant to read. 

I start copyediting by removing unnecessary words, like redundant adjectives or adverbs, tautologies, or verbose expressions.

For example, “unanimous” is perfectly fine on its own. No need to precede it with “completely”. “Gifts” are always “free”, so why add the adjective? “Reverse back”? Can you reverse forward?

They don’t add meaning, so slash them. 

Weasel words, like “experts believe,” or “business outcomes,” may sound impressive, but they’re just filler. They weaken authority. 

Replacing them with specific, data-backed details makes the piece more persuasive.

For example, “increased conversions by 12.3%” sounds more convincing than “made significant improvements.”

*The AI rewrite suggestions are often thin or inaccurate, but they’re a good starting point. 

I look for repetitions both within a paragraph and across the article. And fix them by using synonyms or other examples. Or by cutting.

Repetition makes the draft feel lazy — and well, repetitive.

Hedges and qualifiers, like “probably” or “may,” are useful to signal nuance — and expected in certain contexts, for example, in academic writing.

But when overused, they weaken arguments and make the tone feel uncertain.

That’s why I prompt the GPT to find hedging examples. And cut them to make the writing more confident.

I mean, usually.

“I vary the sentence length, and I create music. Music.” citing the classic, Gary Provost. 

Mixing up sentences of different lengths and structures makes a more engaging read.

On the other hand, if every sentence sounds the same, the text gets monotonous. Boring.

Especially a string of long sentences can exhaust readers and put them to sleep.

Only short ones? No better. They feel robotic. And disjointed.  

Active voice is more concise and engaging than passive voice. It sounds more dynamic.

Passive voice, on the other hand, makes it less personal and cautious.

It has its uses, but in marketing content, we want to be as direct and clear.

Phrases like “not all backlinks are created equal” or “a picture is worth a thousand words” may sound witty, but your readers have seen them before. Hundreds of times.

So, I look for fresher ways to convey the meaning.

This step ensures the draft’s tone of voice aligns with the audience’s expectations and the brand’s style.

Different clients want different tones — professional, casual, authoritative. Some like rhetorical questions, others discourage them. The same with bucket brigades. 

When I start working with a new client, I create a style tone-and-style prompt based on their guidelines and samples of previously published content.

I feed them to ChatGPT and ask it to reverse-engineer the prompt.

I export the prompt into markdown and upload it as a project file. 

When it’s time to edit, I use the prompt below to find sentences that miss the mark. 

Checking the draft against past client feedback helps me avoid the same mistakes.

So even if I don’t get it right with the first article, future articles always require fewer edits.

For every client I work with, I create a Notion page where I record all their feedback.

In the past, I used to check my drafts against it manually. Now, I outsource it to AI.

After each round, I analyze it for patterns.

If there’s a lot of it, you can ask ChatGPT to do it for you.

I export the feedback analysis results to markdown and upload them as a project file.

And use this prompt to check every new draft before submission.

Time to switch hats from a developmental editor to a proofreader and fix all the pesky typos and grammar mistakes that sneak in. 

I use Grammarly or ProWritingAid, so it catches most issues as I’m writing, so there’s little left to do.

But, to be extra certain the draft is flawless, I run this prompt.

So, that’s my complete (self-)editing workflow. Step by step, with all the prompts that I use to write content that requires minimal revisions and makes my editors happy.

Want to work with a writer who edits their drafts meticulously to save your precious time? Or an editor that will guide your writers with constructive feedback and turn their drafts into publish-ready gems?

Drop me a line!

To finish, let me answer a few questions that may help you get more value out of the article and improve your editing skills.

Yes, you can execute all the prompts for each individual stage (outline, high-level, line, and accuracy) together. Or upload them as project files, one for each stage, to create a custom editing GPT. Ryan Law has recently explained how he does it for Ahrefs blogs.

However, I haven’t had much luck doing it.

My experience has been that the more you ask LLMs to do at once, the less detailed the responses become. Something to do with their limited context windows and tokens. So running each prompt individually gives better results.

Prompting the LLM to implement the edits automatically will speed up your editing process.

However, I don’t do it because I don’t always agree with the suggestions. Many are too harsh and make the draft sound robotic. So, I always amend the draft manually. 

I’ve used ChatGPT and Gemini for editing extensively. At the moment — in September 2025 — I find Gemini more thorough and its suggestions better aligned with the overall tone of voice.

ChatGPT, on the other hand, can be lazy, often skipping sections of the text or picking just one sentence per section. It also hallucinates mistakes I didn’t make myself.

Some common examples of fallacies you find in B2B content include:

Editing makes B2B content clear, credible, and persuasive.

It simplifies complex ideas without losing accuracy, removes errors that damage trust, and sharpens structure to match search intent and buyer needs.

In short, editing transforms a draft into a publish-ready content asset that builds authority and drives conversions.

The article was originally published on tatarek.co.uk on 10 September 2025.