Polski
Content & AI search

ChatGPT and Claude in SEO: a workflow, not a toy

How to put ChatGPT and Claude to real use in SEO: keyword research and clustering, briefs, entities, internal linking and structured data. Plus the guardrails: hallucinations, E-E-A-T and helpful content.

Sebastian Baryś
Sebastian Baryś
25 June 2026 14 min read

Introduction

English version of an article first published in Polish on 25 June 2026. Read the Polish original.

ChatGPT and Claude will not do your SEO for you, but steered well they can take half of the tedious work off your hands. The difference between a “toy” and a professional’s tool is not in the model. It is in what you use it for and how you verify the result.

I have seen both extremes. Some people treat AI as an oracle and publish hallucinations as facts. Others reject it entirely and lose hours on work a model would do in a minute. The truth, as usual, lies in the middle, and it takes the form of a concrete workflow.

Below I show how I use ChatGPT and Claude in day-to-day SEO work: where they give a real advantage, where they can do harm, and which guardrails keep it all in check.

What you will learn:

  • Where AI really speeds up SEO work and where it only pretends to help
  • How to use ChatGPT and Claude for keyword research and keyword clustering
  • How to build briefs and content structures without starting from scratch
  • How to work on entities and internal linking with a model’s help
  • Which guardrails protect you from hallucinations and from a penalty for weak content
  • What my practical AI workflow in SEO looks like, step by step

The principle: AI as assistant, not author

Every sensible AI workflow in SEO rests on one principle: the model is an assistant, not an author. It speeds up research, briefs, analysis and drafts, but responsibility for quality, truth and decisions stays with you.

AI generates quickly and convincingly, nonsense included. Your job is not to click “generate” but to know what to keep and what to throw away.

In practice this means a simple split: AI makes the first move (research, cluster, brief, draft), the human makes the last (verification, facts, context, the decision to publish). The closer to publication, the bigger the human’s role.

Keyword research and clustering

AI will not replace a keyword research tool, but it is very good at organising what you pull out of one. The hard data (search volumes, competitiveness, trends) comes from the tool (such as Senuto, a Polish SEO tool), and the model helps you group it, understand intent and arrange it into a structure.

It works best like this:

  • you export keywords from the tool (with search volumes and intent)
  • the model groups them into topic clusters and by intent (informational, commercial, transactional)
  • on that basis you plan the structure: what is the pillar and what is a supporting page
Never ask a model for “search volumes”. It will make them up with complete conviction. Data always comes from the tool, AI only organises it.

Briefs and content structure

This is where AI saves the most time per piece. Given a keyword cluster and some context, the model proposes an article structure in a minute: headings phrased as questions, sections, the angle, FAQ questions. You get a solid skeleton that only needs correcting, instead of building one from scratch.

A good brief from AI contains headings aimed at real keywords, a proposal for answer-first sections (the answer at the start), a list of People Also Ask questions to cover and suggestions for what not to leave out. Your role: add the expert angle and real examples, and throw out whatever is generic.

The same goes for variants: titles, meta descriptions, leads. The model produces ten proposals, you pick the best one and refine it. Generating variants is classic work for AI: fast, cheap and easy to judge.

Entities and internal linking

Internal linking is an area where AI really helps on larger sites. Give the model a list of URLs and their topics and you get suggested connections: which articles should link to which, with what anchor text and why.

On the entity side, AI helps you keep things consistent: whether you describe who you are and what you specialise in the same way everywhere, whether the content around a given topic forms a coherent cluster, whether there are gaps in coverage. This also matters for AI Overviews, where an unambiguous understanding of entities counts.

The model sees patterns in the structure that are easy to miss with the naked eye across 300 pages. But you decide which connections make sense.

Structured data and technical support

AI handles code and structured data very well. Ask it to generate valid JSON-LD for an article, an FAQ or a product and you get a ready skeleton in a few seconds. That is a huge time saving when you roll out schema across many page types.

Where else AI helps on the technical side:

  • writing and explaining rules (regex, robots.txt, redirect rules)
  • explaining messages from Search Console and technical tools
  • drafting scripts that automate reports

The principle stays the same: you always test generated code (at the very least in a structured data validator) before it goes to production. AI writes fast, but you take responsibility for what you deploy.

Guardrails

Using AI professionally in SEO is above all a matter of knowing the risks. I stick to four rules:

  • Verify facts: check every number, date and quote at the source, because a model can be confident and wrong at once
  • Put value first: Google does not penalise AI, only content with no value, and helpful content and E-E-A-T are the filter
  • Add experience: real examples and conclusions from practice are something the model does not have
  • Protect data: do not paste a client’s confidential information into public models
Mass-produced, generic AI content is not a shortcut to success. It is a straight road into the filter for unoriginal content. Quality always holds up, quantity without quality does not.

ChatGPT or Claude: which one for what

Both models are very good, and in practice it is worth having access to both. The differences are smaller than the similarities. From my experience, I split them like this:

Claude

  • long, complex texts and keeping a consistent style
  • working with large documents and analysis

ChatGPT

  • quick tasks and working with data
  • integrations, plugins, the tool ecosystem

This split is a rough guide, not dogma, because both platforms change quickly. How you steer the tool matters more than which one you choose: a good prompt, your own data and a clear goal give a better result than the choice of model alone.

The workflow step by step

This is my typical process for creating SEO content with AI, from idea to publication:

  1. 01 Data from the tool. Keyword research in Senuto: keywords, search volumes, intent, questions.
  2. 02 Cluster and plan with AI. The model groups the keywords and proposes a structure: a pillar plus supporting pages.
  3. 03 Brief. Question headings, answer-first sections, FAQ questions, what not to leave out.
  4. 04 Draft and editing. AI writes the first version, I add experience, examples and tone.
  5. 05 Verification and publication. I check the facts, the structured data and the linking, and only then publish.

The human at the start (data, goal) and at the end (verification, decision), AI in the middle, where speed counts. The same pattern works for an article, a service description and a large content cluster.

FAQ: common questions

Can ChatGPT do SEO?

ChatGPT (and Claude) can support SEO, but it will not do it on its own. It is very good at language and analysis work: keyword clustering, briefs, content drafts, data summaries and internal linking suggestions. It has no access to current, reliable data on search volumes and competition, though, it does not know the context of your business, and it can make things up. It is a tool in a specialist’s hands, not a replacement for the specialist.

Does Google penalise content written by AI?

No. Google does not penalise the use of AI, only low-quality content produced at scale purely to rank. What counts is value for the user, not how the text was made. Content written with AI’s help that is substantive, original and backed by experience is fine. Content generated in bulk, with no value and no verification, falls under the spam policies, and that is a real risk.

ChatGPT or Claude for SEO?

Both are very good, and in practice it is worth having access to both. In my experience Claude often handles long, complex texts and keeping a consistent style better, while ChatGPT tends to be more convenient for quick tasks and for working with plugins or data. The differences are smaller than the similarities. How you use the tool and how well you steer it matters more than which one you pick.

What is AI best used for in SEO?

The biggest return comes from repetitive and language-based work: keyword research and clustering, briefs, article structures, internal linking suggestions, summaries of audit data, and generating variants of titles and meta descriptions. It does worse where a reliable data source and business context matter. Those tasks stay with the human and the SEO tools.

Will AI replace the SEO specialist?

No, but a specialist who uses AI will replace one who does not. The model takes over the tedious part of the work, so the expert has more time for strategy, analysis and decisions. The specialist’s value shifts from “churning out content” to understanding context, verifying and setting direction. AI raises the ceiling of what is possible, but someone has to steer it.

How do I avoid AI hallucinations in SEO work?

Stick to three rules: verify every fact and every number at the source, give the model your own data instead of asking it to invent some, and never publish content without human review. Hallucinations are most dangerous with data (search volumes, statistics, quotes), because that is where a model can be confident and wrong at the same time. Use AI to process data you supply yourself, not to make it up.

Worth remembering

  • AI is an assistant, not an author. It speeds up research, briefs and analysis, but a human answers for quality and truth.
  • You gain the most on keyword research, clustering, briefs and structure, and less on writing “ready to publish” copy.
  • Google does not penalise the use of AI. It penalises weak, unoriginal content produced at scale with no value.
  • Verify every number, fact and keyword from the model in your tools. An LLM can hallucinate data and search volumes.
  • E-E-A-T and helpful content are the filter: content has to bring real value and experience, however it was produced.
  • Do not paste confidential data into public models. Information security needs a deliberate decision.

Summary

AI in SEO is a multiplier, not a replacement. It multiplies what you can already do: you research faster, build a brief faster, write a draft faster. But if you do not know what you are doing, it multiplies your mistakes too, and with complete conviction.

A specialist with AI does not win because they have AI. They win because they know what to ask of it and what to reject.

Start with one stage, for example keyword clustering or briefs, and refine it before you add the next. A good workflow is built step by step, not all at once.

Want your team to write with AI wisely?

In my SEO & AI training I show a workflow that speeds up the work without the risk of penalties and hallucinations.

Ask about SEO & AI training

How this text was written

I will not pretend it was written entirely by hand, by candlelight. For the research, the proofreading and the code of this page I used a language model. It is a good tool for digging through documentation and catching the typos you no longer see on the third read of your own text. The English version was also prepared with a model, from my Polish original.

What the model did not do: it did not choose the topic, form the arguments or decide what this data means for your site. That stays with me, as does the last word on what goes into the text and what comes out of it.

Where the model helped

  • digging through sources and organising data
  • proofreading and the English version of the text
  • code, page layout and infographics

What stays with me

  • the choice of topic and what is worth saying about it
  • the arguments, the reading of the data and the conclusions
  • fact-checking and the decision on what goes into the text
Sebastian Baryś – freelance SEO specialist

Written by

Sebastian Baryś

Freelance SEO specialist · Częstochowa, Poland, and remote

I started when nobody had dreamt of Core Web Vitals and rankings were checked by hand. Today I work on making companies visible not only in Google, but wherever people ask questions. I write only about things I check myself in client projects.

SEOGoogle AdsTechnical SEOSEO in the AI era

AI in SEO, done sensibly?

I will help you bring an AI workflow into your content and analysis: faster, without the risk of penalties and hallucinations.

Get in touch