GEO tutorial for beginners

How to Get Started With GEO Without Being an SEO Expert: 8 Steps, 60 Minutes

Generative engine optimization, explained as a to-do list. Every step works by hand. And if you connect Ayzeo to Claude or ChatGPT, your assistant does the analysis for you with the copy-paste prompt in each step.

Founders, marketing managers, agency teams and consultants who need AI visibility results without becoming GEO experts.

  • A website you can edit
  • 60 minutes
  • Optional fast lane: Ayzeo Pro or trial, plus Claude or ChatGPT

Last updated: August 27, 2026

Table of Contents Expand

The basics

GEO in five sentences

Generative engine optimization (GEO) is the work of making your brand show up in the answers of AI assistants like ChatGPT, Claude, Gemini and Perplexity.

When someone asks an assistant for a recommendation, it runs a handful of web searches, reads a few pages, and names a few brands.

Which brands it names depends on what those pages say, and every assistant searches a different index.

GEO means finding out which questions matter for you, seeing which sources the assistants read, and making sure those sources talk about you.

You do not need to be an SEO expert for any of this: the eight steps below are the whole method.

Which index each assistant searches
AssistantSearches with
ChatGPTBing
ClaudeBrave Search
Gemini and Google AI OverviewsGoogle
PerplexityOwn index

SEO stays the base. GEO is the extra work per model.

Step 0 · 5 minutes · optional

Connect your assistant

The fast lane: connect Ayzeo to your AI assistant once, and every step below becomes a copy-paste prompt instead of manual research. Works with ChatGPT (developer mode), claude.ai, Claude Desktop and Claude Code, on any Ayzeo Pro or Enterprise plan, trials count.

  1. Add https://ayzeo.com/mcp/v2 as a connector in your AI client.
  2. Sign in with your Ayzeo account when the browser window opens (OAuth, no tokens to copy).
  3. Done. Ask your assistant to list your Ayzeo projects to confirm the connection.
claude.ai and Claude Desktop
  1. Settings, then Connectors, then Add custom connector.
  2. Paste https://ayzeo.com/mcp/v2 and confirm.
  3. Sign in with your Ayzeo account in the browser window.
ChatGPT (developer mode)
  1. Settings, then Apps and Connectors, then Advanced, enable Developer mode.
  2. Create a new connector with the URL https://ayzeo.com/mcp/v2.
  3. Sign in with your Ayzeo account when prompted.
Claude Code
  1. Run: claude mcp add --transport http ayzeo https://ayzeo.com/mcp/v2
  2. Complete the sign-in when the browser opens.

Project brief

Paste this once at the start of the conversation (or save it as a project or space in your assistant) so every later prompt has context:

Project brief
Context for all GEO work: Company [name], website [domain]. We sell [what] to [whom] in [markets]. Our positioning in one sentence: [...]. Our main competitors: [...]. Topics we want to be known for: [...]. Topics we do not want to be associated with: [...]. Tone for any content you draft: [...]. Always use the Ayzeo tools for data, never guess numbers.

Step 1 · 10 minutes

Homework: make your site machine readable

Technical SEO is the base. AI crawlers give you seconds, not minutes.

The manual way

Run any technical site check, then open Google's "AI features and your website" guidance and fix what it lists: crawlability, indexing, text that is visible without JavaScript, accurate structured data. Explicitly allow AI bots in robots.txt.

Fast lane: let your assistant do it

Copy-paste prompt
Fetch the latest website analysis for my Ayzeo project [domain] or run a new one if it is older than 7 days. Rank the open issues by their impact on AI visibility, not by count. Give me the 3 most important fixes with ready-to-paste snippets (for example the robots.txt block that explicitly allows AI bots, JSON-LD, meta descriptions) and tell me what the score would be after these fixes.

What you get back: A ranked fix list with copy-paste snippets and the expected score change.

Example output, demo project hubspot.com

  • AI visibility score: 82/100 (analysis from Jul 31).
  • Fix 1: robots.txt has no explicit AI-bot policy. Paste this block to allow GPTBot, ClaudeBot and PerplexityBot (snippet included).
  • Fix 2: key pages show no visible update date. Add a "Last updated" line to the pricing and product pages.
  • Fix 3: llms.txt and JSON-LD are valid, LCP is 0.7s. Nothing to do here.
  • Expected score after fixes: about 90/100.

You do: Hand the snippets to whoever maintains the site.

Step 2 · 10 minutes

Ask the right questions: your 20 prompts

Everything starts with the questions your buyers type, not with keywords.

The manual way

Write down 20 questions your buyers ask before they choose a provider, in their words, without your brand name. Group them into 4 to 6 topics.

Fast lane: let your assistant do it

Copy-paste prompt
I want to track how AI assistants answer my buyers' questions. My brand: [brand]. What we sell: [one sentence]. Our customers: [persona]. Markets and languages: [countries, languages]. Propose 20 prompts a real buyer would type into ChatGPT, Perplexity or Gemini. Rules: never include my brand name; mix broad category questions, "alternatives to [market leader]" questions, persona situations ("I am a ... and need ..."), and "best ... in 2026" questions; vary length and phrasing; group them into 4 to 6 tags. If Google Search Console is connected in Ayzeo, use the top queries as inspiration. Return a table with prompt and tag that I can paste into Ayzeo.

What you get back: A table of 20 buyer-style prompts, grouped into 4 to 6 tags.

Example output, demo project hubspot.com

  • "What are the best CRM platforms for small businesses?" (tag: category)
  • "What is the best CRM for start-ups and why?" (tag: startups)
  • "How can AI improve customer service efficiency?" (tag: ai-service)
  • "What tools are available for marketing automation?" (tag: automation)
  • Plus 16 more, grouped into 5 tags, ready to paste into Ayzeo.

You do: Paste them into Ayzeo, add the tags, let them run for 14 days.

Step 3 · 10 minutes

Understand how the AI answers

The AI runs several web searches (query fan-out), reads a few pages, then names brands. You can see all of it.

The manual way

Ask each question in ChatGPT, Claude, Gemini and Perplexity. Note which brands are named, in which order, and which pages are linked.

Fast lane: let your assistant do it

Copy-paste prompt
For the Ayzeo prompt "[prompt text]", pull the latest run for each AI model. For each model tell me: which web searches it ran, which pages it opened, which brands it named and in which order, and why the first-named competitor was chosen (which page or source earned it). Then tell me the one change on my site that would make me a candidate for this answer.

What you get back: The full anatomy of one AI answer per model: searches, pages read, brand order, and the reason the winner won.

Example output, demo project hubspot.com

  • ChatGPT ran 4 web searches, including "best CRM software for small businesses 2026 official pricing" and direct lookups like "site:hubspot.com/pricing" and "site:zoho.com/crm pricing".
  • It opened the official pricing pages, then named: HubSpot CRM first, then Zoho CRM, Pipedrive, Freshsales, monday CRM, Bigin.
  • Why HubSpot won this answer: the free-tier pricing page answered the small-business question directly.
  • Claude ran 1 broader search and named 3 brands. One change that would help there: a comparison page with official pricing per vendor.

You do: Nothing yet. Read two or three of these before you change anything.

Step 4 · 5 minutes

Your baseline

One run is noise. Two weeks of runs are a baseline. Come back to this step 14 days after Step 2.

The manual way

Repeat Step 3 weekly for two weeks and count how often you are named and where. That is your baseline.

Fast lane: let your assistant do it

Copy-paste prompt
Give me my AI visibility baseline from Ayzeo for the last 14 days: mention rate, average mention position, citation rate and sentiment, overall and per AI model. Add the top 5 competitors by mention rate. Explain each metric in one sentence for a non-expert and tell me the one metric I should watch first and why.

What you get back: Your before picture: every core metric with a plain-language explanation, plus your top 5 competitors.

Example output, demo project hubspot.com

  • ChatGPT: mentioned in 54% of answers, cited in 93% of web-search answers.
  • Google AI Overviews: 53% mentions, 25% citations.
  • Claude: 49% mentions, 21% citations.
  • Sentiment: positive to neutral, no negative answers in the window.
  • Top competitors by mention rate: Zoho, Salesforce, Pipedrive, monday, Freshsales.
  • Watch first: citation rate on Claude, the biggest gap between being named and being linked.

You do: Write the numbers down. This is the before picture.

Step 5 · 10 minutes

Crack the sources

AI answers are built from a handful of pages. Most of them you can influence.

The manual way

For every answer, list the linked sources. Sort them into your pages, competitor pages and third-party pages. Third-party pages you can get listed on are your quickest wins.

Fast lane: let your assistant do it

Copy-paste prompt
Using Ayzeo's citation sources for my project, list the domains and pages the AI models cite most across my prompts. Split them into three groups: my own pages, competitor pages, and third-party sources (directories, listicles, reports, press). For every third-party source tell me whether and how I can get listed. For every competitor page that gets cited, name the content type I am missing (sector page, quarterly report, comparison, case study) and rank the gaps by how many prompts they affect.

What you get back: The source map behind your AI answers, split into own, competitor and third-party, with your quickest listing wins.

Example output, demo project hubspot.com

  • Your pages: hubspot.com (44 citations), knowledge.hubspot.com (20).
  • Competitor pages: zoho.com (24), salesforce.com (17), monday.com (13).
  • Third-party: reddit.com (66, the single biggest source), youtube.com (19), techradar.com (16), en.wikipedia.org (14).
  • Quickest wins: get into the TechRadar CRM roundup and answer the two Reddit threads the models keep reading.
  • Biggest content gap: a small-business pricing comparison, cited for competitors in 9 prompts.

You do: Pick the content type that shows up most often. That is your first content piece.

Step 6 · 15 minutes

Bing and the other indexes

ChatGPT reads Bing. Almost nobody has optimized for Bing since 2010.

The manual way

Create a Bing Webmaster Tools account, submit your sitemap, enable IndexNow, and check the AI Performance report there. ChatGPT searches Bing.

Fast lane: let your assistant do it

Copy-paste prompt
Walk me through setting up Bing Webmaster Tools and IndexNow for a site built with [CMS], step by step, for someone who has never used them. Then take these URLs [list your 10 most important pages] and tell me in which order to submit them and how to check that each one is indexed in Bing.

What you get back: A click-by-click Bing setup plan for your CMS and a submission order for your 10 most important pages.

Example output, demo project hubspot.com

  • Step 1: open bing.com/webmasters, sign in and import your site from Google Search Console (two clicks, no verification file needed).
  • Step 2: submit sitemap.xml under Sitemaps.
  • Step 3: enable IndexNow with your CMS plugin or the generated API key.
  • Submission order for your 10 URLs: pricing and product pages first, then the comparisons.
  • Check back in 7 days: URL Inspection shows "Indexed" per page.

You do: Submit the pages, check again after a week.

Step 7 · 30 minutes

Build the content the AI is missing

Sector pages, reports and comparisons get cited. Generic blog posts rarely do. Budget 30 minutes plus the writing.

The manual way

Build the content type your competitors get cited for and you lack (sector page, report, comparison), with a direct answer at the top, examples, data and a visible update date.

Fast lane: let your assistant do it

Copy-paste prompt
From my Google Search Console data in Ayzeo, show me the top 10 content opportunities, mark which ones already have a page, and recommend the single page to create first, using the gaps from my source analysis. Draft its outline: title, H2 sections, the questions it must answer, the data or table it should contain, the schema type, and a "last updated" line. Write for a buyer, not for a search engine.

What you get back: A prioritized opportunity list and a complete outline for the one page to build first.

Example output (illustrative)

  • #1 "crm for small business", ACOS 87: position 11, 12,400 impressions, no dedicated page yet. Create this one first.
  • #2 "marketing automation tools", ACOS 81: position 8, existing page could be enhanced.
  • Outline for #1: title, 6 H2 sections, the questions to answer, a pricing table, Product schema, and a visible "Last updated" line.

You do: Write or commission the page, publish it with a visible date, submit it to Bing and Google.

Step 8 · 5 minutes

The routine: your 14-day monitor

GEO changes weekly. You do not need to watch it daily, your assistant does. Five minutes every second Monday.

The manual way

Put a 30-minute check in your calendar every second week: rerun the questions, note what changed, fix one thing.

Fast lane: let your assistant do it

Copy-paste prompt
You are my GEO monitor. Using Ayzeo, compare the last 14 days with the 14 days before: mention rate, position, citations, sentiment, per AI model. Report only what changed materially. Name the competitor that gained the most and explain why (which page or source). Then give me the three highest-leverage actions ranked by impact versus effort. Keep it under 300 words.

What you get back: A 300-word delta report: what moved, who gained, and your three next actions.

Example output (illustrative)

  • vs the previous 14 days: ChatGPT mention rate 54% to 57%, citations up 9, sentiment unchanged.
  • Biggest competitor gain: Zoho, driven by a new pricing comparison the models started citing.
  • Action 1: answer the Reddit thread on CRM pricing that 3 prompts keep reading.
  • Action 2: refresh the update date on the comparison page and resubmit it via IndexNow.
  • Action 3: pitch the TechRadar roundup for inclusion.

You do: Save this prompt together with the project brief as a project or space in your assistant. Run it every second Monday, act on one of the three.

Copy blocks

The prompt library: every prompt in this tutorial

The project brief plus all 8 step prompts, in order. Copy them one by one, or download the whole set.

Download all prompts (.txt)

Project brief
Context for all GEO work: Company [name], website [domain]. We sell [what] to [whom] in [markets]. Our positioning in one sentence: [...]. Our main competitors: [...]. Topics we want to be known for: [...]. Topics we do not want to be associated with: [...]. Tone for any content you draft: [...]. Always use the Ayzeo tools for data, never guess numbers.
Step 1. Homework: make your site machine readable
Fetch the latest website analysis for my Ayzeo project [domain] or run a new one if it is older than 7 days. Rank the open issues by their impact on AI visibility, not by count. Give me the 3 most important fixes with ready-to-paste snippets (for example the robots.txt block that explicitly allows AI bots, JSON-LD, meta descriptions) and tell me what the score would be after these fixes.
Step 2. Ask the right questions: your 20 prompts
I want to track how AI assistants answer my buyers' questions. My brand: [brand]. What we sell: [one sentence]. Our customers: [persona]. Markets and languages: [countries, languages]. Propose 20 prompts a real buyer would type into ChatGPT, Perplexity or Gemini. Rules: never include my brand name; mix broad category questions, "alternatives to [market leader]" questions, persona situations ("I am a ... and need ..."), and "best ... in 2026" questions; vary length and phrasing; group them into 4 to 6 tags. If Google Search Console is connected in Ayzeo, use the top queries as inspiration. Return a table with prompt and tag that I can paste into Ayzeo.
Step 3. Understand how the AI answers
For the Ayzeo prompt "[prompt text]", pull the latest run for each AI model. For each model tell me: which web searches it ran, which pages it opened, which brands it named and in which order, and why the first-named competitor was chosen (which page or source earned it). Then tell me the one change on my site that would make me a candidate for this answer.
Step 4. Your baseline
Give me my AI visibility baseline from Ayzeo for the last 14 days: mention rate, average mention position, citation rate and sentiment, overall and per AI model. Add the top 5 competitors by mention rate. Explain each metric in one sentence for a non-expert and tell me the one metric I should watch first and why.
Step 5. Crack the sources
Using Ayzeo's citation sources for my project, list the domains and pages the AI models cite most across my prompts. Split them into three groups: my own pages, competitor pages, and third-party sources (directories, listicles, reports, press). For every third-party source tell me whether and how I can get listed. For every competitor page that gets cited, name the content type I am missing (sector page, quarterly report, comparison, case study) and rank the gaps by how many prompts they affect.
Step 6. Bing and the other indexes
Walk me through setting up Bing Webmaster Tools and IndexNow for a site built with [CMS], step by step, for someone who has never used them. Then take these URLs [list your 10 most important pages] and tell me in which order to submit them and how to check that each one is indexed in Bing.
Step 7. Build the content the AI is missing
From my Google Search Console data in Ayzeo, show me the top 10 content opportunities, mark which ones already have a page, and recommend the single page to create first, using the gaps from my source analysis. Draft its outline: title, H2 sections, the questions it must answer, the data or table it should contain, the schema type, and a "last updated" line. Write for a buyer, not for a search engine.
Step 8. The routine: your 14-day monitor
You are my GEO monitor. Using Ayzeo, compare the last 14 days with the 14 days before: mention rate, position, citations, sentiment, per AI model. Report only what changed materially. Name the competitor that gained the most and explain why (which page or source). Then give me the three highest-leverage actions ranked by impact versus effort. Keep it under 300 words.

Your first 90 days

A realistic 30-day GEO roadmap for beginners

  1. Week 1Questions, health check, start the baseline (connect the assistant if you take the fast lane).
  2. Week 2Technical fixes, Bing Webmaster Tools, IndexNow.
  3. Week 3First content piece from the source analysis.
  4. Week 4First before/after comparison, set up the two-week routine.

Days 31 to 90

  • Second content piece and one third-party listing per month.
  • Monthly before/after comparison against the Step 4 baseline.
  • Keep the two-week monitor routine, act on one action per run.

Questions

Frequently asked questions

How long until I see results?

Technical fixes and Bing indexing show up within days. Mentions and citations usually start moving after 4 to 8 weeks, once new content is crawled and the models' search indexes refresh. That is why the Step 4 baseline matters: without it you cannot see the movement.

Do I still need SEO?

Yes. Every AI assistant except Perplexity leans on a classic search index (Bing, Brave or Google), so crawlability, structured data and good content remain the foundation. GEO is the extra, per-model work on top, not a replacement.

Which AI model should I optimize for first?

Start where your buyers are. For most teams that is ChatGPT, which means Bing (Step 6). Your Step 4 baseline shows where you are weakest relative to demand; fix the biggest gap first.

What if my brand is never mentioned?

Normal at the start. The fastest path is Step 5: find the third-party pages the assistants already cite (directories, comparison posts, review sites) and get listed there, then build the content type your competitors get cited for.

Do I need a developer?

Only for Step 1, and only if you cannot edit robots.txt, meta tags or structured data yourself. Everything else is writing, publishing and using free webmaster tools. The fast-lane prompts hand you ready-to-paste snippets you can forward to whoever maintains the site.

Can I start without a sales call or a long setup?

Yes. The manual way needs nothing but this page. The fast lane is a 14-day trial plus one URL in your assistant: add https://ayzeo.com/mcp/v2, sign in, done.

Which AI model and setting should I use for the prompts?

A strong model with high reasoning, never the free tier: this work needs a lot of context. In Claude pick the largest available model; in ChatGPT use the strongest reasoning model your plan offers.

Changelog

Changelog: what changed in the models

GEO shifts with every model update. We log the changes that affect this playbook, newest first.

  • ChatGPT runs more site: searches and relies less on Reddit since the early-August model update. Step 6 (Bing) matters more, Reddit seeding less.

Ready for the fast lane?

Everything on this page works by hand. Ayzeo makes it a copy-paste job: 14-day trial, no sales call needed.