How to Use ChatGPT for SEO Content That Google Still Ranks
AI writes the draft. A person makes it worth ranking. Here is how we do it at DMT, with the prompts we use and our own numbers.

ChatGPT for SEO works. Raw ChatGPT output for SEO does not, and the difference between the two is one editing pass. In the largest controlled test we know of, NP Digital published 744 AI-drafted articles across 68 client sites. Half went live exactly as the model wrote them. Half were edited by a person first. The edited half earned 5.44 times the organic traffic, and the gap widened over time. This is how we use ChatGPT and Claude for SEO content at Digital Marketing Tribe: the research step, the two prompts, the editing pass, and what happened when we ran the same checklist on our own blog this week.
What ChatGPT is good at in SEO, and what it cannot do
ChatGPT is a fast, competent drafter. It structures an article, covers the obvious subtopics and writes clean sentences in minutes. It cannot supply firsthand experience, your own data, an accountable author, or a reason for Google to trust your page over the thousands of near-identical drafts it produced for everyone else. Those come from you, after the draft.
Google does not penalise AI-generated content as such. Its guidance on AI-generated content says the question is whether the content is helpful and shows experience and expertise, however it was produced. That is exactly the problem. A raw draft is a statistical average of what already exists, so it contains nothing an evaluator can trust. NP Digital’s market data across 100 companies shows the result: AI now writes 52.1% of new content and earns 4.9% of organic traffic. “Humaniser” tools that swap synonyms do not fix this, because hiding the tool adds no trust.
Our ChatGPT SEO workflow: research first, then two prompts
We do not start in ChatGPT. We start in Google Search Console and a keyword tool, because the keyword is the one thing no model can pick for you. For this article, Search Console showed 95 queries around “ChatGPT for SEO” bringing 3,600 impressions to a page of ours that answered a different question. That gap is the brief.
Then we read the pages that already rank and note what they lack. Only then does the model get involved. At DMT we run this in Claude Code with a written house standard every draft has to meet, and the same brief works pasted into ChatGPT. The two prompts below are that standard, trimmed to fit.
Those two placeholders are the point. The model is told not to fake the things it cannot know, so the editing pass starts with a map of exactly where a person has to step in.
The five-step edit that makes ChatGPT SEO content rank
NP Digital’s winning half got five things in the 20 minutes before publish. In our words:
- Add something you saw happen. A client result, a number from your own dashboard, a job that went wrong. Proof a person with skin in the game was here.
- Add something only your business knows. Your pricing, your process, your data. The test we use: if a paragraph could sit on a competitor’s site unchanged, delete it or replace it.
- Cut the filler. Every “in today’s fast-paced landscape”, every hedge, every sentence that exists to reach a word count.
- Check every fact, date and link, and add sources. Freshness was the top citable trust factor in NP Digital’s survey of six AI platforms, at 91%.
- Put a named person on it. A byline with a real role, a bio and a photo. Not “the team”.
Step two on one of our own posts
Before: “Imagine having a Swiss Army knife in the digital wilderness. That’s your digital marketing agency.”
After: “A digital marketing agency covers the whole funnel: SEO, Google Ads, social, content and web development, run as one plan. At Digital Marketing Tribe that means the same team that builds a client’s website also runs its SEO and Google Ads.”
One of those could sit on any agency’s site. One could only sit on ours.
What we found when we ran it on our own blog
On 11 September 2026 we scored all 31 posts on this blog against those five steps. 23 had no client story or DMT number in them. None of the 31 linked to a single outside source. 21 showed a publish date before 2025, and every post’s schema carried the wrong date after a site migration. Six posts earned nearly all of the blog’s clicks in Search Console; ten had earned none in six months.
So we did the work. Ten dead posts were removed with redirects. The dates were corrected on the 21 that stayed. The nine posts that earn traffic each got a real result from a published case study, one to three sources, and a filler cut. Two posts written from inside specific roles were reattributed to the people in those roles. We will report what Search Console shows for those nine posts in this article in six weeks, whichever way it goes.
For a client example of the same principle, Underworld Laser, a laser tag arena in Menai, saw organic traffic rise 170% and its AI visibility score rise 67% after we fixed the technical gaps and added the structured data and content its pages were missing.
How long it takes and what it costs
The edit takes us 20 minutes on a short post and over an hour on a long one, and we do not treat that as time the model saved. The model saves the blank-page hour. The editing hour is where the ranking comes from.
If you would rather hand that hour to us, our blog copywriting service starts from $280 per post and a four-post monthly retainer from $900, with this process built in. Full detail is on our pricing page. For the other half of the picture, our AI SEO and GEO guide covers being cited inside ChatGPT, and making your website AI-ready covers the technical side.
Frequently asked questions
Does Google penalise content written with ChatGPT?
No. Google judges AI-generated content the way it judges everything else: on whether it is helpful, reliable and made for people. Raw drafts fail that test not because of how they were made but because of what they lack.
Can Google tell that ChatGPT wrote something?
Assume it can, and notice that it does not matter. Detection is not the mechanism; evaluation is. A draft with a real author, real data and real sources ranks regardless of the tool that produced the first version.
How much editing does a ChatGPT draft need?
At least 20 minutes, working through the five steps above. A spelling pass and a fresh intro sentence is not an edit. NP Digital’s data shows lightly touched AI content performs almost the same as raw output.
The short version
Use ChatGPT for SEO content on every post if you like. Pick the keyword from real data, brief the model with a written standard, tell it to mark what it cannot know, and then spend the 20 minutes that turn a draft into a page Google has a reason to rank. Never let the model hit publish. If you want us to run that process on your blog, talk to us about blog copywriting.

Lachlan Best
Lachlan Best is an Account Manager and Copywriter at Digital Marketing Tribe. He specialises in copywriting, Google Ads and client management, and is the day-to-day contact for many DMT clients.
Stay Sharp on Digital Marketing
Monthly SEO, Google Ads and digital marketing updates, straight from the DMT team. No spam, unsubscribe any time.
"*" indicates required fields
Ready to Get Real Results?
Your free, no-obligation strategy session shows you exactly where the biggest opportunities are for your business.


