How to Write a Blog Post With AI That Still Ranks (and Doesn't Read Like AI)

Yes, a blog post written with AI can rank in Google. Google does not care whether a model typed the words. It cares whether the page is the most useful answer to the search. The problem with most AI posts is not that they are AI. It is that they are generic, and generic loses to anything specific.

So the goal is not to hide the AI. It is to use it for what it is good at (structure, options, first drafts) and keep for yourself what it cannot do: choose the right search, supply real experience, check facts and make the final calls. Below is the eight-step workflow we follow with Claude and ChatGPT, plus what Google's guidance actually says, checked September 2026.

What Google actually says about AI content

Three Google Search Central pages cover this. Here is the short, accurate version.

AI content is allowed. Google's February 2023 post on AI-generated content says its ranking systems reward original, high-quality content that demonstrates E-E-A-T (experience, expertise, authoritativeness, trustworthiness), and that the focus is "the quality of content, rather than how content is produced." Its FAQ says appropriate use of AI or automation is not against its guidelines.

Using AI to manipulate rankings is spam. The same post says using automation, including AI, to generate content "with the primary purpose of manipulating ranking" violates Google's spam policies. The spam policies name it scaled content abuse, and the first example is using generative AI tools to generate many pages without adding value for users. The rule applies no matter how the content is made.

AI gives no bonus either. Google's FAQ is blunt: using AI "doesn't give content any special gains." If the post is useful, helpful and original, it might do well. If not, it might not.

Google's page on using generative AI content adds two practical points. Focus on accuracy, quality and relevance, including metadata like the title element, meta description, structured data and image alt text. And consider telling readers how content was made where they would reasonably wonder, which its helpful content guide frames as the "Who, How and Why" of a page. Google also says giving AI an author byline is probably not the best way to be clear about it.

That is the whole policy. Everything below is about making a post that passes it.

Step 1: Pick one keyword and one intent

Do this yourself, with data. Ask an AI "what should I write about" and you get what everyone else got.

Start in Google Search Console with queries your site already gets impressions for, then search your chosen query in a private window and read the top five results. Decide:

• The exact query, in the searcher's words.
• The intent: are they trying to learn, compare or buy? One post, one intent.
• Your angle: the qualifier a big publisher cannot own. "For a small site", "on a free plan", "without an agency".

If you write for a query you cannot say in one sentence, the draft will wander, and AI drafts wander more.

Step 2: Write a brief before you open the chat

A brief is the difference between a draft you edit and a draft you rewrite. Keep it to one page:

1. The query and the intent.
2. Who is searching and what they already know.
3. The one-sentence answer the post will give.
4. What the top results cover, and the gap none of them fill.
5. Your own material: the experience, numbers, screenshots or opinions only you have.
6. Facts that must be checked, with the source for each.
7. The pages on your site to link to.

The brief is also your guard against scaled content abuse. A post with a real brief is a post someone thought about.

Step 3: Let the AI propose the outline, then fix it

Paste the brief and ask for an outline:

"Here is my brief. Propose an outline of H2 headings for a post that answers the query in the first paragraph, then covers the rest in the order a reader would need it. Mark which sections depend on facts I must supply. Do not write the post."

Then correct it. Delete the sections every competitor already has if they add nothing. Move the answer up. Add the section for your gap. Your H2s should read like a table of contents someone could navigate. The model is good at giving you options here. Choosing is still your job, which is the principle behind the ten prompts we actually reuse.

Step 4: Add your own experience and data

This is the step that decides whether the post ranks, and the one people skip.

Google's helpful content questions ask whether a page shows first-hand expertise, for example from having actually used a product, and whether it provides original information or analysis. A model trained on the web can only give the web's average back to you. What it cannot give:

• What happened when you tried it, including what went wrong.
• Numbers from your own account, analytics or client work.
• Screenshots of your setup.
• An opinion you would defend, with the reason.
• The steps in the order you actually do them.

Write these in rough notes under each outline heading before any drafting. Bullet points are fine. If you have nothing to add under a heading, question whether the heading belongs in the post.

Step 5: Draft section by section, in your voice

Do not ask for the whole post in one go. One long prompt produces one long average. Draft per section, with your notes and your voice attached.

Before the first section, give the model a voice file: ten paragraphs you wrote that sound like you, and a short list of words you never use. Then:

"Write the section '[H2]' using only these notes and facts [paste]. Short sentences, plain English, second person. Do not add facts, numbers or examples that are not in my notes. If something is missing, write [CHECK] instead of guessing."

The [CHECK] instruction matters. It turns the model's urge to fill gaps into a visible to-do list instead of a confident invention.

Step 6: Fact-check everything that can be wrong

Read the draft once looking only for claims: prices, plan limits, feature names, statistics, dates, quotes, "studies show". For each:

• Check it on the source's own page, not a roundup or the model's memory.
• Add when you checked it: "(checked October 2026)".
• If you cannot verify it, delete it.

Assistants with web search help you find sources faster, but still open the source and read the line yourself. A single wrong price is the kind of error that makes a careful reader leave and never trust the page again.

Step 7: Edit answer-first, and strip the tells

Now edit like a reader.

Answer first. The first sentence of the post answers the query. Most AI drafts open with a warm-up about how fast things are changing. Delete it and see if the post gets better. It almost always does.

Strip the tells. The same habits that give away AI video scripts give away AI blog posts:

• Lists of exactly three, everywhere.
• Hedging ("it's important to note", "can potentially").
• Hype words: game-changer, unlock, delve, seamless, revolutionary.
• Generic claims with no names, numbers or examples.
• Every sentence the same length.
• A summary paragraph that repeats the post.

Read the draft out loud once. Anywhere you would not say it that way, rewrite it.

Step 8: Run the on-page basics

Google's AI content guidance specifically mentions metadata, so treat these as part of writing, not an afterthought:

1. Title with the query's phrasing, under about 60 characters.
2. Meta description: one sentence promising the answer, under 155 characters.
3. H2s that work as a table of contents.
4. Internal links: from this post to two or more of your key pages, and from two or three older posts to this one.
5. An FAQ of four to six real questions, each answered in its first sentence.
6. Alt text that describes each image.
7. A byline and a date, and a line on your process if a reader might ask how the post was made.

Structured, dated, answer-first pages are also what ChatGPT and Perplexity tend to cite, so this step pays twice.

The 15-minute version you can do today

You will not write a full post in fifteen minutes. You can write the brief that makes the post good.

1. Minutes 0 to 3. Pick one query your site already gets impressions for in Search Console. Write it down in the searcher's words, plus the intent: learn, compare or buy.
2. Minutes 3 to 7. Search it in a private window. List what the top three results cover and one thing none of them do.
3. Minutes 7 to 11. Write the one-sentence answer your post will give, then three bullet points of experience or data only you have.
4. Minutes 11 to 13. List every fact the post will need, each with the source you will check.
5. Minutes 13 to 15. Paste the brief into Claude or ChatGPT with the outline prompt from Step 3. Save the outline. Stop.

Tomorrow you draft from a brief instead of from a blank prompt, and that alone changes what comes out.

What not to do

• Do not publish at volume. Ten thin AI posts a week is the pattern Google's scaled content abuse policy describes. One strong post a week beats it.
• Do not paste the model's facts. Every number goes through Step 6.
• Do not skip Step 4. Without your experience, the post is the average of the pages already ranking, and there is no reason for Google to prefer it.
• Do not credit the AI as author. Google advises against it. Put your name on it and own it.

Where tools fit

This workflow runs on an assistant and Search Console. Two kinds of tools can make it faster once you publish every week.

A prompt collection saves you rewriting the same instructions. Our free Solo Marketer Prompt Library has the prompts that pair with this workflow, such as the intent clusterer for Step 1 and the title and meta pass and FAQ generator for Step 8, with copy buttons.

A content optimizer scores your draft against what currently ranks, which helps on crowded queries. Our Surfer vs Semrush comparison for small sites covers which kind of tool to buy first, and the AI SEO tools hub lists the rest, including tools for getting cited in AI answers.

None of them do Step 4 for you. That part is why the post ranks.