← All articles GUIDE

How to use Claude for SEO: the complete workflow

Most people use Claude to write a blog post and stop there. The real leverage is connecting it to your live search data so it can do the research, the technical audit, the publishing, and the measurement. This is the full workflow, step by step.

To use Claude for SEO, connect it to your Google Search Console and your site's files, then run the full loop: data-driven keyword research, content that answers the query completely, technical and indexing audits, internal linking, and AI Overview optimization. Connected to live data, Claude does the research and publishing, not just the writing. The judgement of what deserves to rank stays yours.

Four ways to run SEO: the honest comparison

Before the workflow, it helps to see where this approach actually fits. Here is the honest version against the alternatives most small businesses weigh up.

Approach Monthly cost Speed to publish Reads your live data? You own the workflow?
Hire an SEO agency $1,000–5,000/mo Weeks per deliverable Yes (they do) No. Leave and it stops.
SEO SaaS suite $100–300/mo You still do the work Yes No. Rented dashboards.
Do it manually $0–30/mo (tools) Slow. It is your time. If you check by hand Yes, but it eats your week
Claude + your data (this guide) $0–30/mo (APIs) Same day, end to end Yes, directly Yes. It runs in your repo.

An agency is still the right call if you want zero involvement and have the budget. A SaaS suite is fine if you enjoy the dashboards and do the work yourself. But if you want the research, the writing, and the publishing to happen in one place, on your own data, at near-zero monthly cost, Claude connected to your stack is now a real option. This guide walks the whole loop.

What "Claude for SEO" actually means

The phrase gets used two ways. The shallow version is "ask Claude to write a blog post," which is useful but is maybe a fifth of the job. The version that moves rankings is Claude with tools connected: it reads your Search Console, knows which queries you already show up for, sees which pages Google has indexed, pulls real keyword volumes, writes the page, publishes it, and checks the result. That is a workflow, not a chat.

The honest framing: Claude does not replace SEO judgement. It removes the manual labour around it. You still decide which topics are worth owning and whether a page is genuinely the best answer. Claude does the data pulls, the drafting, the structuring, and the publishing that used to take a full afternoon each.

The five jobs Claude can actually do for SEO

SEO is not one task. It is five, and Claude handles each one differently. The loop below is the same one we run on this site every week.

01

Keyword research, from data

Connected to a volume source, Claude pulls real search volumes and competition for a topic, expands it with the questions people actually type, and ranks the list by opportunity. No more guessing which phrasing has demand. The output is a prioritised queue, not a brainstorm.

02

Content that answers the query

Claude turns a target keyword into a full draft with the structure search engines reward: a direct answer up top, clear headings, a comparison table, and an FAQ. The skill matters here. A generic prompt produces generic text. A structured one produces a page that is genuinely the best result.

03

Technical and indexing audits

With Search Console connected, Claude checks which pages Google has actually indexed, finds the ones stuck on page two, flags crawl waste, and tells you which sitemap URLs were never discovered. This is the unglamorous half of SEO that usually never gets done.

04

Internal linking and clusters

Claude reads your whole site, groups related pages into topic clusters, and generates the cross-links between them. Internal linking is how Google discovers your long tail and how authority flows to your money pages. Doing it by hand across a growing site is the kind of work nobody keeps up with.

05

AI Overview optimization (AEO)

A growing share of searches end with an AI summary at the top. Claude structures pages in the exact pattern those summaries cite: answer-first opening, comparison table, FAQ schema, stated methodology. Getting quoted there is the new position one.

The rest of this guide takes each of those five and shows the actual setup and the prompt patterns that work.

Step 1: Connect Claude to your real search data

This is the step that separates real SEO from blind writing, and it is a one-time setup. Claude talks to outside tools through MCP servers, small connectors that give it a specific capability. For SEO, the one that matters most is a Google Search Console connection.

Once it is wired in, you stop describing your site to Claude and start asking it to read your site. The difference in output is night and day. Instead of "write me an article about CRMs," you can say "look at my Search Console, find every query where I rank between position five and fifteen, and tell me which pages are one good edit away from page one." That is a real instruction against real data.

Step 2: Run keyword research from data, not vibes

Most keyword research dies as a brainstorm in a doc. The fix is to put real numbers behind it. Connected to a volume API, Claude takes a seed topic, pulls the search volume and competition for every variation, expands the list with the long-tail questions people actually search, and hands you a ranked queue. You pick from evidence instead of intuition.

A pattern that works: give Claude one seed phrase and ask it to harvest the autocomplete expansions first (the real phrasings people type), then pull volumes for the promising ones, then rank by the opportunity formula of demand against competition. The result is a work queue you can publish against for months.

Weak prompt "Give me keywords about CRMs."
Strong prompt "Harvest autocomplete expansions for 'free crm' and 'crm for', pull search volume and competition for each, drop anything under 100 searches a month, and rank what's left by volume divided by competition. Give me the top 20 as a table I can publish against."

Step 3: Write pages that are genuinely the best answer

This is where most "AI SEO" goes wrong. A lazy prompt produces a page that reads like every other AI page and ranks like none of them. The pages that rank share a structure: they answer the query in the first paragraph, they include specifics and proof only the author would know, they use a comparison table where a comparison is what the searcher wants, and they close with a real FAQ.

Claude is good at producing that structure on demand, but you have to ask for it. The skill below encodes the structure so you are not re-explaining it every time. The judgement of whether the finished page deserves to rank, whether it actually helps the reader more than the current top result, stays with you. That review is the part that keeps your site out of the thin-content filter.

Step 4: Fix indexing, the half of SEO nobody does

You can publish the best page in your niche and earn nothing if Google never crawls it. On a young or low-authority site, that happens constantly: pages sit undiscovered, the crawl budget gets wasted on test pages and utility URLs, and the sitemap goes stale. This is invisible unless you look, and almost nobody looks.

With Search Console connected, Claude looks for you. It checks which URLs are actually indexed, finds the ones marked "discovered, not indexed," spots the junk pages eating crawl budget, and tells you exactly which money pages to request indexing for first. We ran this on our own site recently and found that most of it had never been crawled at all. The fix was a sitemap cleanup and a round of indexing requests, both of which Claude scoped in minutes.

Then there is the striking-distance work: the pages already ranking on page two that one edit could push to page one. Claude pulls them straight from your query data and tells you which are worth the effort. That is the highest-return hour in SEO, and it is sitting in data most people never open.

Step 5: Get cited in AI Overviews

The search results page changed. For a large and growing share of queries, Google now puts an AI-generated summary at the top, and the links it cites get the attention. Being one of those citations is the new top position, and the rules for earning it are different from classic ranking.

The content that gets quoted is unusually consistent: it opens with a direct, self-contained answer of 40 to 60 words, it puts a comparison table above the fold, it uses FAQPage schema, and it states its methodology. Claude can produce content in exactly that shape, and it can scan which of your existing pages are close. The softer secret is that AI Overviews lean heavily on community sources like Reddit and on YouTube, so the strongest plays pair an on-site page with a genuine answer in the right community and a short video.

The honest limits

A few things Claude will not do for you. It will not invent authority: a brand-new domain still has to earn trust over time, and no workflow shortcuts that. It will not make a thin page rank by force, and trying to mass-produce filler is the fastest way to get filtered. And it will not replace the judgement of knowing your customer better than any model does. The specifics that make a page the best answer, the real numbers, the lived experience, the opinion, have to come from you.

What it does remove is the labour. The research that took an afternoon, the indexing audit nobody ever runs, the internal linking that never gets kept up, the structuring of every page for citations. Hand those to Claude and the part that is left is the part that was always the actual work: deciding what deserves to exist and making sure it is genuinely good.

SoloStack runs its entire SEO operation this way, in one repo, on the same database as the CRM and the rest of the business. We build that setup with you, live, in the workshop.

Common questions

It does both, but the content writing is the smaller half. Connected to the right tools, Claude can pull your Search Console data and find pages stuck on page two, run keyword research through a real volume API, audit which of your pages Google has actually indexed, generate internal links between related articles, and check whether you appear in AI Overviews. The writing is one job in a much longer workflow. The leverage comes from Claude doing the research, the technical audit, and the publishing, not just producing a draft.
Three connections cover most of it. A Google Search Console connection (via an MCP server) lets Claude read your real query and indexing data instead of guessing. A keyword volume source like DataForSEO gives it search volumes and competition so topic selection is data-driven, not vibes. And access to your site's files lets it write and publish pages directly. Without those, Claude is a smart writer working blind. With them, it is running the actual loop: research, write, publish, measure, repeat.
Google's guidance is about quality and helpfulness, not whether a human or AI typed the words. Thin, templated, near-duplicate pages get filtered whether a person or a model wrote them. The way to stay safe is the same as it has always been: each page answers a real question fully, includes specifics only you know, and is genuinely the best result for that query. Claude is good at structure and speed. The judgement of whether a page deserves to rank stays with you.
The difference is the tooling around it, not the model. ChatGPT in a browser tab writes you a draft from its training data. Claude running in your project folder, with Search Console and your files connected, reads your actual traffic, edits the real page, and commits it live. One produces text you paste somewhere. The other operates your site. For SEO specifically, that connection to live data and the ability to publish is what turns it from a writing aid into a workflow.
No, but you do need to do a one-time setup: connecting Claude to your Search Console and your site. That part is copy-paste instructions, not programming. After it is wired up, the work is conversational. You say 'find my pages on page two of Google and tell me which are worth optimising' and Claude pulls the data and answers. The setup helper skill walks through the connection step by step if you get stuck.
AEO is Answer Engine Optimization: getting cited inside Google's AI Overviews and in answers from tools like ChatGPT and Perplexity. It matters because a growing share of searches now end with an AI summary at the top, and being the source it quotes is the new position one. The content patterns that win citations are specific and learnable: answer-first openings, comparison tables, FAQ schema, clear methodology. Claude is well suited to producing content in exactly that structure.
The work happens faster, but Google's clock is unchanged. A page Claude helps you publish today still needs to be crawled, indexed, and earn position over weeks. What changes is throughput and feedback speed: you can publish more, audit indexing the same day, and spot striking-distance pages the moment the data shows them. The compounding is the same as any SEO. You just get to the volume and the fixes a lot quicker.

Skills referenced in this article

Every skill linked above, in one table. Each is a free, downloadable Claude Code prompt.

Skill What it does
Keyword Research with DataForSEO Pull real search volumes and competition so Claude picks topics from data, not guesswork
SEO Blog Writing Turn a target keyword into a structured, on-brand article ready to publish
Answer Engine Optimization (AEO) Structure pages to get cited in Google AI Overviews and LLM answers
Programmatic SEO Generate hundreds of templated landing pages from one data source
Scrape Google Search Pull live SERP and People-Also-Ask data to map a topic before you write

Run your whole SEO operation in one repo

The SoloStack workshop wires Claude into your Search Console, your keyword research, and your site, so research, writing, and publishing all happen in one place, on your own data. We build it live with you in one session.

See the workshop →

or explore the full SoloStack stack →