Free — no email required

Before you pay an agency, run the check yourself.

Three steps, about twenty minutes, nothing to buy. A full technical audit of your site, an AI visibility check across ChatGPT, Perplexity and Gemini, and one prompt that merges both into a fix list ranked by impact. Then you'll know what you're actually buying.

You'll need Claude Code for step 1. Steps 2 and 3 run in a browser.

What you walk away with

Three deliverables an agency would charge you for

This is the discovery phase of a paid engagement. It used to take a senior consultant a day. It's now a skill, a tool, and a prompt.

A technical audit

Every page crawled, scored, and explained: what's blocking indexation, which Core Web Vitals are failing, what schema is missing or deprecated, and where your internal links dead-end.

An AI visibility check

Engine by engine — ChatGPT, Perplexity, Gemini — whether you're named when a buyer asks, and which competitor got recommended in your place when you're not.

A fix list, in priority order

Both reports merged and ranked by impact over effort, with the top five written as specific first actions on your actual URLs — plus the findings you can safely ignore.

The full guide

Do these three things, in this order

Step 1 is a one-time setup and the longest part. Steps 2 and 3 take about five minutes between them.

1About 15 minutes, once

Install the free Claude SEO skill and point it at your site

Claude SEO is an open-source skill for Claude Code — MIT licensed, free, and not ours. It runs a fleet of specialist agents across your site in parallel and hands back a scored, prioritised audit. Install it from inside Claude Code:

claude-code
/plugin marketplace add AgriciDaniel/claude-seo

Adds the skill's marketplace to Claude Code. One time, ever.

/plugin install claude-seo@agricidaniel-claude-seo

Installs the skill itself.

/seo setup

Builds its isolated Python environment and browser. Takes a minute or two.

Then connect Google Search Console and run the crawl. The Search Console connection is what turns a generic crawl into an audit of your actual queries and impressions — do not skip it.

claude-code
/seo google setup

Credential wizard. Connect Google Search Console here — it also unlocks PageSpeed and Core Web Vitals field data.

/seo audit https://yoursite.com

The full crawl. Swap in your own domain and let it run.

Prefer to read the script first? The repo also documents a manual install — clone it, read install.sh, then run it.

git clone --depth 1 https://github.com/AgriciDaniel/claude-seo.git bash claude-seo/install.sh

What comes back:

Crawl and indexation errors

Pages Google can't reach, canonical chains pointing at the wrong URL, robots and noindex rules blocking things you meant to rank.

Core Web Vitals

LCP, INP and CLS on mobile and desktop separately, using real Chrome field data where it exists rather than a lab score.

Schema markup

What's there, what's invalid, and what's been retired — Google dropped FAQ rich results in May 2026 and HowTo before that.

Internal linking

Which pages have authority and pass none of it on, and which money pages sit three clicks deep with nothing pointing at them.

Save the output. You'll paste it into step 3.

View the skill on GitHub
2About 60 seconds

Run the free AI Mention Check

The audit in step 1 tells you why Google can't rank you. It says nothing about whether ChatGPT will name you — and that's increasingly where the question gets asked first.

Locafy's AI Mention Check runs the buyer-intent questions your customers actually ask — “best plumber in Alton”, “who should I call today” — live against ChatGPT, Perplexity and Gemini. It shows, engine by engine and question by question, whether you come up. And when you don't, it names the competitor that got recommended instead.

18 engine-checks, no email gate on the first look

Six questions, three engines. Tools that do this start around $200/month. Run it, then save or screenshot the result — it's the second half of step 3.

Run the AI Mention Check
3About 3 minutes

Drop both reports into Claude and rank every fix by impact

Two reports on their own are two piles of problems. What you need is one list, in order, with the things that don't matter removed.

Start a new Claude conversation, paste or attach both reports, then paste the prompt below. Replace the [BRACKETED] fields with your business details first — the effort scores are meaningless if Claude doesn't know whether “you” is one person or a team of six.

The prioritisation prompt

472 words · paste straight into Claude

You are my SEO strategist. I'm giving you two reports about the same website. Read both before you answer anything.

REPORT A — a technical SEO audit from the Claude SEO skill: crawlability, indexation, Core Web Vitals, schema, internal linking, and content quality.

REPORT B — a Locafy AI Mention Check: whether ChatGPT, Perplexity and Gemini name my business when buyers ask, engine by engine, and which competitors got recommended instead.

BUSINESS: [BUSINESS NAME]
WEBSITE: [WEBSITE URL]
WHAT WE SELL: [SERVICE OR PRODUCT]
WHERE WE SELL IT: [CITY OR SERVICE AREA]
WHO DOES THE WORK: [just me / a team of N / an agency]

Work through this in order.

1. INVENTORY
List every distinct problem across both reports, one line each. Deduplicate hard: if the two reports describe the same underlying fix in different language, that is one item, not two. A thin service page shows up in Report A as "low word count" and in Report B as "no passage for the engine to quote" — same fix.

2. SCORE
Give every problem three scores from 1 to 5:
- IMPACT: how much traffic, ranking, or AI visibility fixing it plausibly returns.
- CONFIDENCE: how sure we are it is a real problem, based on evidence in these two reports — not on best practice in general.
- EFFORT: 1 = under an hour, 3 = a solid day, 5 = weeks of work or needs a developer.
Rank by (IMPACT x CONFIDENCE) / EFFORT. Show the arithmetic.

3. RANKED FIX LIST
One table, highest score first:
Rank | Fix | Which report flagged it | Impact | Confidence | Effort | Score | What "done" looks like

4. THIS WEEK
Take the top 5 and write the specific first action for MY site. Not "improve internal linking" — "add a link from /about to /emergency-plumbing using the anchor 'emergency plumbing'". Use the actual URLs from the reports.

5. THE HONEST PART
Tell me which findings to ignore. Every audit tool produces items that are technically true and commercially irrelevant. Name them and say why they don't matter for a business my size.

6. THE AI VISIBILITY GAP
Look at the competitors the engines named instead of me in Report B. For each, say what they appear to have that Report A says I don't — and whether that is fixable on my own site or needs off-site work (citations, reviews, mentions on sources the engines actually quote).

RULES
- Name the report every finding came from. If it isn't in either report, don't claim it.
- Never invent a metric. If a number isn't in the reports, write "not measured".
- Order by what recovers traffic soonest, not by how broken something looks.
- If the two reports contradict each other, say so and tell me which to trust.

Scored, not listed

Impact x confidence, divided by effort. The arithmetic is shown so you can argue with it.

Top five, spelled out

Specific first actions on your real URLs — not "improve internal linking".

What to ignore

Every audit tool flags things that are true and irrelevant. This names them.

Questions

The obvious ones, answered

Is this actually free, or is there a catch?

It's free. The SEO skill is MIT-licensed open source on GitHub — not ours, not gated, no account. The AI Mention Check is Locafy's free tool. Step 3 is a prompt you can copy off this page. The only cost is a Claude subscription to run the skill in Claude Code, which you may already have.

Do I need to be technical to run this?

You need to be able to install Claude Code and paste commands into it. That's the whole bar. The skill does the crawling and the analysis; you read the output. If you can follow a recipe you can finish this in under half an hour.

Why do I need two reports instead of one?

They measure different failures. The technical audit tells you why Google can't rank your pages. The AI check tells you why ChatGPT won't name you — which depends far more on citations, reviews and third-party mentions than on anything on your own site. Fixing one does not fix the other, and most audits only cover the first.

Why would an SEO agency give this away?

Because most of what an agency charges for in month one is this exact diagnosis, and it's now automatable. What isn't automatable is the execution: building the citation and review layer, publishing at pace, and getting mentioned on the sources AI engines quote. That's the part we sell.

What if the fix list is longer than I can handle?

That's the expected outcome, and it's why step 3 ranks by impact over effort instead of listing everything as urgent. Do the top five. Re-run the audit in two weeks and compare. Most sites see the majority of the available gain from a handful of fixes.

Now you know the problem. Want it done?

Diagnosis is the part that automates. Execution isn't — the citations, the reviews, the schema, and the third-party mentions AI engines actually quote take months of steady work. That's what Locafy does, done for you.

Want more of these? The free tools hub has 20 more — map rank heatmaps, suspension risk checks, competitor steal sheets, and the rest.