How to Use Claude to Write Blog Content: The Cleo Skill Guide

Updated
Skill · .md

The skill behind this guide: Cleo - Blog Content Writer AI Skill. Holds the brief, the outline and your banned-words list between sessions, so every post goes through the same process instead of being improvised - $19, one payment, yours permanently.

View the Cleo skill →

Anyone can get a model to produce a thousand words about anything. The problem is never that the words are machine-written; it is that they are empty, and emptiness has a recognisable texture. This guide is about the process that removes it: a brief with a position in it, an outline that argues rather than lists, drafting one section at a time, and an edit whose only job is deletion. Everything here works in Claude, ChatGPT or any AI chat.

First, what Google actually says

Worth settling before anything else, because the anxiety around this question wastes a lot of people's time in both directions. Google's spam policies define the thing that gets penalised as scaled content abuse, and the wording is specific: it is "when many pages are generated for the primary purpose of manipulating search rankings and not helping users", focused on "creating large amounts of unoriginal content that provides little to no value to users, no matter how it's created". One of the listed examples is using generative AI tools to produce many pages without adding value.

Read that carefully and it says two things at once. Using a model to write is not the offence. Producing volume with nothing in it is, and the fact that a human typed it does not save it. So the question to ask about a draft is never "does this look AI-written", it is "would someone be worse off if this page did not exist". A post with your data, your customer's actual words and a position you are prepared to defend passes that test whoever or whatever assembled the sentences. A post that restates the top five results in a friendlier voice does not.

The tells, concretely

People say AI writing "sounds like AI" and cannot say why. It is worth being precise, because each of these is fixable and most survive a casual read.

The tell What it looks like The fix
The throat-clearing opening In today's fast-paced digital landscape, email marketing remains more important than ever. Delete the first paragraph. Almost always the post starts better at paragraph two
The restated title The opening tells the reader what the post is about, which they knew, because they clicked the title Open with the claim, the number or the disagreement
Rule of three, everywhere Every list has exactly three items; every sentence has three adjectives Make the lists the length they actually are. Two is fine. Seven is fine
The not-just-X construction It is not just about writing faster, it is about writing better. Ban it. Say the second half only
Hedge stacking This can often be one of the more effective approaches for many teams. State it or cut it. If you cannot state it, you do not know it yet
Uniform rhythm Every paragraph three sentences, every sentence the same length Break one. A three-word sentence resets the reader's ear
The conclusion that recaps In conclusion, blog writing with AI requires a thoughtful approach. End on the last real point, or on what to do on Monday
No proper nouns, no numbers Nothing named, nothing counted, nothing dated. The surest tell of all See the next section, because this is not a style problem

The one thing a model cannot supply

The other tells are cosmetic. This one is structural, and it is the whole reason most AI blog content fails commercially rather than aesthetically.

A model can produce the shape of expertise without any of the content of it, because it has read thousands of posts about your subject and none of your work. It does not know what your customer said on the call last Tuesday, what the number actually was, which supplier let you down, or what you tried first that did not work. Those are the parts a reader remembers and the parts that make a page worth linking to.

There is a test that takes five seconds per paragraph: could this sentence have been written by someone who has never done the thing? If yes for the whole post, no amount of rewriting will save it, because the missing ingredient is not prose quality. It is you. The workflow below is built to make you supply that up front rather than notice its absence at the end.

The process, in order

Prompt 1 - the brief, which is where the post is won or lost

"Write a blog post about X" is a topic. A brief names the reader, the change you want in their head, the position, and what would make the post wrong. That last one is the field almost nobody fills in, and it is the one that stops the draft hedging.

Help me write a brief for a blog post. Interview me one question at a time, and do not move on until my answer is specific. Get these out of me: 1. Who exactly is reading this, and what they already believe about the subject. 2. What they should believe or do differently by the end. 3. The position. Not the topic: the thing I am claiming that a reasonable person could disagree with. 4. What I know that someone writing this from research alone could not know: a number, a name, a case, something that went wrong. 5. What would make this post wrong. Under what circumstances my position does not hold. 6. The one search phrase this should earn, and the question behind it. After the interview, write the brief back to me in under 200 words. Then tell me plainly which of the six is weakest and what it would take to fix it. If item 4 is empty, say that the post is not worth writing yet and tell me what to go and find.

Prompt 2 - an outline that argues

Most AI outlines are a list of headings in a plausible order, which is why the draft reads like a list of headings with prose between them. Ask for the argument instead, and fixing a weak structure costs a minute rather than an afternoon.

Using the brief below, outline the post as an ARGUMENT, not a list of topics. For each section give me: - the heading - the single claim that section makes, in one sentence - what the reader must accept before this section works, and which earlier section earns it - the specific evidence it needs from me, marked [I SUPPLY] - the objection a sceptical reader has at this point Rules: - No section may exist only to provide background. If a section makes no claim, cut it and say so. - Order the sections so each one is possible only because of the one before it. Tell me if my material does not support that order. - End with the three sections most posts on this topic include that we should deliberately leave out, and why. BRIEF: [PASTE]

Prompt 3 - draft one section, never the whole post

A model asked for a whole post optimises for finishing it. Asked for one section with the evidence in hand, it optimises for that section. The difference in output quality is larger than any prompt wording trick.

Draft ONE section only: [HEADING], making the claim [CLAIM]. Material to work from, and the only material you may use: [PASTE YOUR NOTES, NUMBERS, QUOTES, EXAMPLES] Hard rules: - Every factual statement must trace to my material. If a sentence needs a fact I have not given you, write [NEED: what] rather than supplying it. - No introduction to the section and no summary of it. - Do not write "it is not just X, it is Y". Do not open with "in today's". Do not use the word "landscape". - Vary sentence length deliberately. At least one sentence under six words. - Name things. Prefer the specific number over "many", the named example over "for instance, a company might". - Target [N] words. Going under is fine; going over is not. Then, separately, list every sentence you are least confident belongs in the final post, and say why.
Skill spotlight
Cleo - Blog Content Writer AI Skill
Cleo - Blog Content Writer AI Skill
$19 · one payment

Holds the whole sequence between sessions - the brief format, your banned-words list, your house rules on sentence length and hedging - so the tenth post follows the same discipline as the first without you re-explaining what good means each time. It asks for the brief before it writes, and refuses to draft a section whose evidence is missing.

View Cleo - Blog Content Writer →

The edit is the job

This is the step everyone skips, and it is where a draft becomes something a person finishes reading. Two passes, and they must be separate conversations, because a model asked to both improve and cut will improve, since improving is easier and looks like more work.

Prompt 4 - the deletion pass

Edit the draft below by DELETION ONLY. You may not add a word, rewrite a sentence, or improve anything. You may only cut, and then show me what is left. Cut, in this order: 1. The first paragraph, unless it makes a claim. 2. Every sentence that tells the reader what the post will do, what it just did, or what is coming next. 3. Every hedge that does not change the meaning: "can often", "may sometimes", "in many cases", "generally speaking". 4. Every adjective that survives its own removal. 5. Every example that repeats a point already made by another example. 6. The concluding recap. Target: at least 25% shorter. Output the cut version, then a list of anything you removed that you think I will miss, so I can put it back.

Prompt 5 - the tell audit

Run this last, in a fresh conversation, on the cut version. Asking the conversation that wrote the draft to detect its own habits gets you a defence.

Audit the text below for the specific habits that make writing read as machine-produced. Do not rewrite anything. Flag, quoting the line each time: - openings that restate the title instead of making a claim - "not just X, it is Y" and its variants - lists that are three items long for no reason - stacked hedges - three or more consecutive paragraphs of similar length and shape - any paragraph containing no proper noun, number, date or concrete example - transitional sentences that carry no information - any sentence that could have been written by someone who had never done the thing being described Then give me one number: the percentage of paragraphs that contain at least one specific, checkable fact. Show your working so I can disagree with it.
Complementary prompt pack
Creative & Publishing Prompt Library
Creative & Publishing Prompt Library
$9 · one payment

The writing jobs that sit around the blog itself - headlines, scripts, long-form structure, publishing workflow - as ready prompts rather than a persona you load. Useful when the next piece is not a blog post.

View Creative & Publishing Prompt Library →

One post becomes ten, but only if the post was good

A finished piece is raw material for the social posts, the newsletter section and the snippets that send people back to it. The method is in repurposing one post into a month of content. The order matters though: repurposing a weak post produces ten weak things and a lot of activity. If the post did not pass the specificity test, it is not raw material, it is filler with a distribution plan.

If the piece is meant to rank, pair this with the on-page work in the SEO specialist guide, and with the line-level craft in the copywriter guide.

Where it fails

  • Invented specifics. The moment you ask for concrete detail without supplying it, you get plausible statistics, studies and quotes that do not exist. This is the single most dangerous failure on a published page, and the instruction that prevents it is telling the model to write [NEED: what] rather than fill the gap.
  • Length as a proxy for value. Ask for 2,000 words and you will get 2,000 words, most of which are the same 900 words with connective tissue. Set a ceiling, never a floor.
  • Agreeableness. Ask whether the draft is good and you will be told yes. Ask what a sceptical reader stops reading at.
  • Your voice, from a description. "Write in a friendly, professional tone" produces the house style of the internet. Paste three things you actually wrote and ask it to name the patterns, then keep that list.
  • The same shape every time. Left alone, a model converges on one structure. This is why the outline stage has to be a separate decision rather than something you accept.
  • Volume. Producing many pages fast is exactly the behaviour Google's scaled content abuse policy describes. The tool makes it easy; that is a reason for restraint, not a licence.

How do you install the Cleo skill?

The download is a ZIP with SKILL.md at the root of the archive, not inside a nested folder - that folder structure is the usual reason an upload fails. In the Claude desktop app, open Customize → Skills, upload the ZIP and toggle it on. Claude loads it whenever your request matches what it covers.

Skills need code execution enabled, under Settings → Capabilities. Anthropic's help centre currently lists Skills on Free, Pro, Max, Team and Enterprise, while its Academy tutorial lists Pro, Max, Team and Enterprise - so if you are on the free plan, check Settings → Capabilities for your own account rather than taking either page's word for it. On Team and Enterprise that switch sits with an owner, which is why the menu is sometimes greyed out on a work account. The full walkthrough is in the skill installation guide.

In ChatGPT or Gemini there is no upload step: open SKILL.md, copy the contents, and paste them into custom instructions. You lose automatic triggering and keep the method.

Who is this for?

Founders and marketers writing their own blog, content teams trying to keep quality steady across a lot of posts, and freelance writers using a model as a drafting tool rather than a replacement. It works in Claude, ChatGPT or any AI chat.

The catalogue splits the neighbouring jobs rather than bundling them into one vague writing assistant:

If you want the whole writing stack rather than one file, the Content Writer Tool Bundle is five AI skills and two tool skills for $69 instead of $133 separately. The wider sets are the content writer skills and the creative, content & design skills.

In summary:

Google penalises volume without value, not machine-written sentences, so the question about a draft is whether anyone would be worse off without the page. Bring the specifics a model cannot have, brief it with a position and a way of being wrong, outline the argument rather than the headings, draft one section at a time, and run a deletion-only edit in its own conversation. For the process held between sessions, Cleo - Blog Content Writer AI Skill ($19); for the writing jobs around the blog, the Creative & Publishing Prompt Library ($9). Both work in Claude, ChatGPT and any AI chat, with a 30-day money-back guarantee.

Skill · .md · Works with Claude & ChatGPT

Cleo - Blog Content Writer AI Skill

One file that asks for the brief before it writes, outlines the argument rather than the headings, drafts a section at a time, and marks a missing fact instead of inventing one. Your banned words and house rules, kept between sessions. No subscription. Yours permanently.

$19
Get the Cleo skill →

KissMySkills is a marketplace of 1,900+ AI skills, 60+ prompt packs, 90+ agents & free tools for Claude, ChatGPT & any AI chat.

Related Skill Guides

Put this into practice

Or get every skill with Unlimited See plans →

Frequently asked questions

Does Google penalise AI-written blog content?+

Not for being AI-written. Google's spam policies target scaled content abuse, defined as when many pages are generated for the primary purpose of manipulating search rankings and not helping users, and the policy says this applies to unoriginal, low-value content no matter how it is created. Using a model to write is not the offence; producing volume with nothing in it is, and a human typing it does not save it. The useful question about a draft is whether anyone would be worse off if the page did not exist.

Why does AI blog content sound like AI?+

Because of a small set of recognisable habits rather than anything mysterious. The throat-clearing opening about today's fast-paced landscape. An intro that restates the title the reader already clicked. Lists that are always three items long. The not-just-X-it-is-Y construction. Stacked hedges such as can often be one of the more effective approaches. Uniform paragraph length and rhythm. A conclusion that recaps. And the surest tell of all, paragraphs containing no proper noun, number or date.

What can an AI not do when writing a blog post?+

It can produce the shape of expertise without any of the content of it. It has read thousands of posts about your subject and none of your work, so it does not know what your customer said on the call, what the number actually was, which supplier let you down, or what you tried first that failed. Those are the parts readers remember and the parts that earn links. The five-second test per paragraph: could this sentence have been written by someone who has never done the thing?

How do I brief an AI to write a blog post properly?+

Write a brief rather than naming a topic. It should contain who is reading and what they already believe, what should change in their head, the position you are claiming that a reasonable person could disagree with, what you know that research alone could not supply, what would make the post wrong, and the search phrase it should earn. That fifth field, the conditions under which your position does not hold, is the one almost nobody fills in and the one that stops the draft hedging.

Should I ask AI to write the whole post at once?+

No. A model asked for a whole post optimises for finishing it; asked for one section with the evidence in hand it optimises for that section, and the quality difference is larger than any prompt wording trick. Approve an outline first, drafted as an argument with a claim per section rather than a list of headings, then draft section by section with the source material pasted in and an instruction to mark a missing fact rather than invent one.

How do I edit AI writing so it reads human?+

Run a deletion-only pass, in its own conversation, with no permission to rewrite or improve anything. Cut the first paragraph unless it makes a claim, every sentence that says what the post will do or just did, every hedge that does not change the meaning, every adjective that survives its own removal, and the concluding recap. Target at least 25 percent shorter. A model asked to both improve and cut will improve, because improving is easier and looks like more work.

How do I stop AI inventing statistics in a blog post?+

Give it the material and forbid it from supplying anything else. The instruction that works is to have it write a placeholder such as NEED: what whenever a sentence requires a fact you did not provide, rather than filling the gap. Asked for concrete detail without source material, a model produces plausible statistics, studies and quotes that do not exist, which is the most dangerous failure on a published page.

How do I install the Cleo Blog Content Writer skill?+

The download is a ZIP with SKILL.md at the root of the archive, not inside a nested folder. In the Claude desktop app open Customize then Skills, upload the ZIP and toggle it on. Skills need code execution enabled, under Settings then Capabilities, or by an owner in organisation settings on Team and Enterprise. Anthropic's help centre currently lists Skills on Free, Pro, Max, Team and Enterprise, while its Academy tutorial lists Pro, Max, Team and Enterprise, so if you are on the free plan check Settings then Capabilities for your own account rather than trusting either page. In ChatGPT or Gemini, paste the contents of SKILL.md into custom instructions instead.

~/get-started

Skills that work. No fluff.

Browse every skill, prompt pack, and agent in the store.

Browse all skills →Or try the free tools