THE MINI GUIDE — NO FLUFF, JUST STEPS

Managing SEO &

AI Visibility

with Claude

Claude chat · Claude MCP · Claude Code — prompts, use cases, and a step-by-step playbook to get cited inside AI answers.

Get found where buyers now search: ChatGPT · Claude · Perplexity · Gemini · Google AI Overviews

1

SETUP · CONNECT YOUR DATA

Connect your data sources

Open Integrations in Ryze and connect the sources behind your SEO and AI-visibility work. Each one feeds live data into every audit, rewrite, and report.

Analytics

Google Analytics 4 · Search Console · AppsFlyer — traffic, queries, conversions, indexation.

Commerce & CMS

Shopify · Framer · WordPress · GitHub — the pages and content you publish.

One click, OAuth-secured

Sign in, grant read access — no API keys to paste.

2

SETUP · CONNECT EACH SOURCE

Pick your account & property

Connecting a source takes seconds — click Connect, authorize, then choose exactly which property or site Ryze should read.

Google Analytics 4 — choose your property

WordPress — choose the site to publish to

Google Search Console: click Connect, then sign in with Google and grant read-only access (Search performance + indexing). No property setup needed — Ryze pulls every verified site.

3

SETUP · CONNECT CLAUDE

Bring it all into Claude

Add Ryze as a custom connector in Claude.ai — about 90 seconds. Then chat with your ad, analytics, and search data wherever you already work.

Find this on the Ryze Home page — copy the MCP URL, then follow the steps →

claude.ai — add custom connector

1 In Claude.ai, click your avatar (bottom-left) → Settings.

2 Open Connectors, then click Customize.

3 Click the + by Connectors → Add custom connector.

4 Name it “Ryze AI” and paste the MCP server URL.

5 Leave Advanced settings as-is → click Add.

6 Under “Not connected,” open Ryze AI → click Connect.

7 Pick your workspace → Allow Claude to connect.

8 Done — “Connected to Ryze AI.” Claude can now call GA4, Search Console, Google & Meta Ads, and more.

WHY THIS MATTERS NOW

Search moved from ranking to being the answer

AI engines don’t hand back ten blue links. They compose one answer and name a handful of brands. SEO earns rankings; GEO (generative engine optimization) earns the citation inside that answer.

2–4

brands named in a typical AI buying answer — vs. 10 links on a Google page

<20%

of AI-cited sources also rank top-of-page on Google — down from ~70%

~58%

of buyers now start product research inside an AI chatbot

~60%

of Google searches end with no click at all

Sources: industry GEO research, 2026 (Brandlight · Siftly · Omnia · Relato).

THE MENTAL MODEL

How an AI engine actually builds its answer

You’re not optimizing for position #1. You’re optimizing to be an extractable, trusted passage the model pulls and cites.

Prompt

User asks a real question

›

Query fan-out

Splits into 3–5 sub-queries

›

Retrieval (RAG)

Pulls live pages from the web

›

Extract

Lifts self-contained passages

›

Answer

Composes + cites 2–4 sources

Takeaway: win on extractability + trust signals across many sources — not on a single keyword ranking.

YOUR TOOLKIT

Three tools, one workflow

Think → connect → execute. Each Claude surface owns one job in the loop.

Claude (chat)

THINK + WRITE

• Audit how AI answers your buyer prompts
• Rewrite pages to be answer-first & quotable
• Draft FAQ, comparisons, JSON-LD schema

Claude MCP

YOUR REAL DATA

• Connect Search Console, GA4, Drive, Ads
• Pull live query data into the work
• Write briefs back, open tasks & tickets

Claude Code

DO IT AT SCALE

• Audit the whole site / repo at once
• Bulk-add schema, fix crawler access
• Stand up a recurring AI-visibility monitor

1

STEP 1 · CLAUDE CHAT

Audit your AI visibility

Before changing anything, see exactly how AI answers your real buyer questions today — who gets named, and why.

Run it with web search ON

so Claude pulls live answers, not memory.

Capture the citation set

which brands appear and which pages got pulled.

Find the gaps

the prompts where you’re invisible = your hit list.

claude.ai — web search on

Act as a GEO analyst. For each prompt below,

search the web and report:

1. which brands the AI answer names

2. whether [MyBrand] appears + how it's

described

3. which pages are cited, and why

Prompts:

> best [category] tool for [audience]

> [MyBrand] vs [Competitor]

> how to [job your product does]

Return a table + my top 5 visibility gaps.

2

STEP 2 · CLAUDE CHAT

Rewrite for extractability

AI lifts short, self-contained passages. Lead with the answer, turn headers into the questions people actually ask.

Answer-first

first 2–3 sentences must answer the query — no warm-up.

Question headers

rewrite each H2 as the literal user question.

Self-contained

every section must make sense lifted out alone.

Keep your voice

use a Claude Project + Style to stay on-brand.

claude.ai — paste your page

Rewrite this page to be AI-citable:

- Open with a 2-3 sentence direct answer

to: "[target question]" (no preamble)

- Turn each H2 into the exact question a

user would ask

- Make every section self-contained and

quotable out of context

- Add one verifiable stat or example

per section

Keep our voice. Return the new draft

+ a list of what changed and why.

[paste page]

3

STEP 3 · CLAUDE CHAT

Add schema + entity signals

Make the structure machine-readable, then go earn mentions off your own domain — that’s what builds entity trust.

Generate JSON-LD

FAQPage, Article (author + dates), Organization.

Mirror the page

schema should describe what’s already visible on-page.

Build entity recognition

consistent mentions across third-party sources.

Target where AI looks

roundups, forums, directories, profiles.

claude.ai — schema + off-page

From this page, generate valid JSON-LD:

- FAQPage for the Q&A sections

- Article w/ author, datePublished,

dateModified

- Organization w/ sameAs profile links

Then list 5 third-party places (roundups,

forums, directories) where earning a

mention would strengthen entity

recognition for "[brand / topic]" —

with a one-line pitch for each.

4

STEP 4 · CLAUDE MCP

Plug Claude into your real data

MCP connectors let Claude read live numbers and write work back — so audits and rewrites run on truth, not guesses.

Search Console

Real question queries + pages stuck at #5–15 to rescue

Drive / CMS

Read brand voice; write content briefs & drafts back

GA4 / Ads data

Demand + related questions to seed prompt coverage

Tickets / repo

Open one task or PR per page that needs the rewrite

claude.ai — connectors on

> Using Search Console, list pages ranking #5-15 for question

queries. For each: draft an answer-first intro + the JSON-LD,

then save the brief to Drive and open a ticket.

WHY MCP WINS

One prompt reads your data, decides the fix, and files the work — no copy-paste between tools.

5

STEP 5 · CLAUDE CODE

Fix the whole site at once

Run Claude inside your repo to do the repetitive, site-wide work that no one wants to do by hand.

Unblock the crawlers

first cause of invisibility — verify robots.txt.

Bulk schema

add Article + FAQ JSON-LD across every page.

Restructure at scale

rewrite H2s into questions where it fits.

Ship safely

one PR per change; never alter meaning.

terminal — your site repo

# install once

npm i -g @anthropic-ai/claude-code

# from inside your repo

claude

> Audit this site for AI visibility:

1. robots.txt — confirm GPTBot,

ClaudeBot, PerplexityBot,

Google-Extended are NOT blocked

2. add Article + FAQ JSON-LD where

missing

3. rewrite H2s as questions w/o

changing meaning

4. open one PR per change

6

STEP 6 · CLAUDE CODE

Build a visibility monitor

GEO is a system, not a one-off. Track citations on a schedule so you catch drops before they cost you demand.

Mention rate

how often AI names you on your prompt set.

Share of voice

you vs. competitors, per engine.

Beat the citation cliff

freshness fades fast — refresh on a cadence.

Alert on drops

get pinged the moment share of voice slips.

terminal — recurring job

claude

> Build a script that, weekly:

- sends our prompt set to the

OpenAI, Claude & Perplexity APIs

- logs which brands + URLs each

answer cites

- computes our mention rate and

share of voice per engine

- writes /reports and diffs vs

last week

Then schedule it and alert me when

share of voice drops.

DON’T WASTE EFFORT

What works vs. the myths

Google’s own 2026 guidance debunked several popular “hacks.” Spend your time on signals that actually drive citations.

DO — this earns citations

• Lead with the answer; headers = real questions
• Original data, stats & named expert authors (E-E-A-T)
• Earn third-party mentions; keep entities consistent
• Keep AI crawlers unblocked + refresh content often
• Track citations across every engine, not just one

DON’T — low-ROI / debunked

• Stuff keywords — there’s no “prompt density”
• Treat llms.txt or secret schema as a magic unlock
• Chunk pages into micro-fragments for the bot
• Publish on your domain with zero outside mentions
• Optimize for a single model and call it done

PUT IT TOGETHER

Your 7-day visibility sprint

1

Day 1

Audit your top 20 buyer prompts in Claude (Step 1)

2

Day 2–3

Rewrite 5 highest-intent pages answer-first + schema (2·3)

3

Day 4

Connect Search Console + Drive via MCP; pull real questions (4)

4

Day 5

Run the Claude Code site audit; fix crawlers + bulk schema (5)

5

Day 6

Earn 3 third-party mentions — roundups, profiles, outreach

6

Day 7

Stand up the weekly monitor, then repeat monthly (Step 6)

The one rule

Be the most extractable, best-sourced answer — and earn citations everywhere AI looks.

Built with Claude · chat · MCP · Code