How to Learn GEO and AEO: A Free 6-Week Curriculum
Every week, 38% of people in six countries use generative AI to look things up, the Reuters Institute found. Yet most courses on getting cited in those answers still teach FAQ schema and llms.txt, two things Google says its AI search doesn't need. This free six-week curriculum teaches GEO and AEO from the engines' own sources instead, with a lab each week.
TL;DR
- Learn it in this order: how answers are built, which crawlers can read you, how to measure, which pages get cited, who else gets cited, then how to prove a change.
- Read the engines' and Google's own pages and independent studies, and skip FAQ schema and llms.txt, which Google says its AI search doesn't use.
- Ask every question several times before you trust it: the same question can name a brand one week and leave it out the next.
- Finish with the capstone: 20 buyer questions on 3 engines, 3 runs each, one fix, then the same again.
You can learn GEO and AEO for free in six weeks, at about four hours a week, from the companies that run the AI engines and a few independent studies. Learn it in this order: how an AI answer is built, which crawlers can read your site, how to measure, which pages get cited, which sources you don't own, and how to prove a change. Practise each week on one real brand, and finish by measuring that brand again.
Generative engine optimization (GEO) and answer engine optimization (AEO) are two names for the same work: getting your brand named, and your pages cited, in the answers that ChatGPT, Google's AI Overviews and AI Mode, Perplexity, Gemini, Claude and Copilot give.
It's worth six weeks. In its 2026 survey of six countries, the Reuters Institute found that 38% of people use generative AI to get information each week, and 57% see AI summaries in search engines. When a summary appears, people click less: Pew Research Center found a result was clicked on 8% of Google visits with an AI summary, against 15% without one.
Most courses on the subject come from companies that sell a tool, and they teach to it. This curriculum assigns only free primary sources: the engines' own documentation, the standards behind crawling, academic studies, and our own published research, which is marked as ours. You don't need a tool to finish it. Where one of our free tools saves time, the week says so.
Why Learn It From the Sources
Advice about AI search goes stale faster than the courses that teach it. Since May 2026:
- Google's FAQ rich result stopped showing on 7 May, and its guide to generative AI features says AI search needs no special schema.org markup.
- On 15 June, Google added a note to the same guide: Google Search ignores llms.txt files.
- On 31 August, the Search generative AI control went live for every site: a setting in Search Console that takes a site out of AI Overviews and AI Mode.
- OpenAI's help page now says ChatGPT rewrites a question and sends it to partner search providers.
- Bing announced page-level ai-grounding and ai-training preferences that replace its old Copilot opt-outs from 1 February 2027.
Each change appeared first on the company's own page. So learn how the system works from the people who built it, keep their links, and check them again every few months.
The other thing most courses skip is measurement. The same question can name a brand one week and leave it out the next, so a single check can't tell you where you stand, and no tactic helps until you can tell a real change from noise. That is why week 3 is the longest.
The Curriculum at a Glance
The six weeks at a glance
About 23 hours in all, reading and labs together. Every reading is free.
| Week | Topic | Time | Lab | You've learned it when |
|---|---|---|---|---|
| 1 | How AI answers are built | 4 h | Ask one buyer question on three engines and log every page each cites | You can explain why two engines cited different pages |
| 2 | Which crawlers can read your site | 3 h | Read your robots.txt line by line and fetch a key page as each AI crawler | You can say which bot block costs citations and which costs nothing |
| 3 | Measuring without fooling yourself | 5 h | Ask 20 buyer questions on three engines, three times each, and record who gets named | You can say whether a 10-point change between two weeks is real |
| 4 | The pages AI cites | 4 h | Compare your page for your top question with the pages AI cited, then rewrite it | You can name the format AI cites for that question, and why your page isn't it |
| 5 | The sources you don't own | 4 h | List every page cited for your top five questions and mark which ones name you | You can name the ten pages that shape your category's answers |
| 6 | Proving a change | 3 h | Ask the same 20 questions again and compare the change with your range | You can show a boss or client a change that isn't noise |
You need one brand to practise on, ideally your own or a client's; a spreadsheet; and ChatGPT, Perplexity and Google's AI Mode, whose free versions are enough. Every reading below is free to open, and the long ones are marked optional.
Before week 1, if search is new to you
The engines lean on ordinary search, so its basics still matter. If crawling, indexing and ranking are new to you, start with Google's own two guides, then our three short definitions.
- Search engineHow Google Search Works20 min
- Search engineSEO Starter Guide40 min
- Ranqo guideWhat Is Generative Engine Optimization (GEO)?10 min
- Ranqo guideWhat Is Answer Engine Optimization (AEO)?10 min
- Ranqo guideGEO vs AEO vs SEO12 min
Week 1: How AI Answers Are Built
What to learn
An AI answer that cites sources is built in three steps: the engine turns the question into searches, runs them, then writes from what came back and links some of it. Google calls the first step query fan-out: in its AI Mode announcement, the system issues related searches at the same time, across subtopics and data sources. OpenAI says ChatGPT rewrites a question for its search partners, Perplexity searches in real time and numbers its citations, and Claude searches when current information would help.
Two consequences run through the rest of the course. The pages that win were found by searches you never saw, so they often aren't the ones ranking for your exact question. And familiarity counts: in our study of 102 brands, global brands were named in 72.9% of answers to buyer questions that didn't mention them, and niche brands in 11.4%.
Google's guide also draws a line worth learning early: making pages to match fan-out searches in order to manipulate AI answers counts as scaled content abuse.
- Search engineOptimizing your website for generative AI features on Google Search30 min
- Search engineExpanding AI Overviews and introducing AI ModeWhere Google names query fan-out10 min
- AI companySearching the web with ChatGPT10 min
- AI companyHow does Perplexity work?5 min
- AI companyEnable and use web search (Claude)5 min
- TextbookRetrieval Augmented Generation in Azure AI SearchHow a question becomes parallel subqueries25 min
- TextbookSpeech and Language Processing, retrieval and RAG chaptersOptional2 h
- Ranqo guideQuery Fan-Out: What AI Engines Actually Search15 min
- Ranqo guideWhat Google's 2026 AI Search Guide Adds12 min
- Ranqo guideHow to Get Cited by ChatGPT: The Two-ChatGPTs Playbook12 min
- Ranqo toolQuery Fan-out GeneratorOptional10 min
The lab
Pick one question a buyer in your category asks before they know any brand names, like "best project tool for a 10-person agency". Ask it in ChatGPT, Perplexity and Google AI Mode, and paste every cited page into a sheet with the engine beside it. Then put the question into our free Query Fan-out Generator, search two of the related queries it suggests, and see which of those results turned up in the answers.
You've learned it when you can explain why two engines cited different pages for the same question.
Week 2: Which Crawlers Can Read Your Site
What to learn
Each AI company runs more than one crawler, and they do different jobs. OpenAI's GPTBot collects training data, OAI-SearchBot decides whether your pages can appear in ChatGPT search, and ChatGPT-User fetches a page when a person asks for it. Anthropic splits its bots the same way, and Perplexity runs a search bot and a fetcher and says neither trains a model. Refusing a training crawler keeps your pages out of future training and costs nothing in today's answers. Refusing a search crawler takes you out of that engine's answers.
Three more things belong in this week. RFC 9309, the robots.txt standard, says its rules are not access authorization, and a fetcher acting for a person may not follow them at all. Google-Extended isn't a crawler and doesn't decide whether you appear in AI Overviews; the Search generative AI control does. And most AI crawlers don't run JavaScript, so text that appears only after scripts run can be invisible to them.
- StandardRFC 9309: Robots Exclusion Protocol30 min
- AI companyOverview of OpenAI Crawlers10 min
- AI companyDoes Anthropic crawl data from the web?5 min
- AI companyPerplexity Crawlers5 min
- Search engineGoogle's common crawlersWhat Google-Extended does and doesn't control10 min
- Search engineSearch generative AI control5 min
- Search engineBing Webmaster Guidelines20 min
- Search engineClearer preferences for training and grounding10 min
- Ranqo guideEvery AI Crawler User Agent in 2026, Ranked15 min
- Ranqo guideIs Your Site Blocking AI Crawlers?12 min
- Ranqo guideWhat AI Actually Sees When It Crawls Your Site10 min
- Ranqo guideHow to Verify Googlebot, GPTBot and ClaudeBot in Your LogsOptional17 min
- Ranqo toolAI Crawler Inspector5 min
The lab
Open your robots.txt and, beside every line that names a bot, write which job that bot does. Then run our free AI Crawler Inspector on your homepage and on the page that answers your top question: it fetches each one as every AI crawler and as a browser, so you can see who gets the page and who is refused. Last, load that page with JavaScript turned off in your browser and check that the answer is still in the text.
You've learned it when you can point at a line in your robots.txt and say whether it costs you citations or nothing.
Week 3: Measuring Without Fooling Yourself
What to learn
Measuring AI visibility means asking the questions buyers ask before they know your name, and counting how often you're named: named in 6 of 20 answers is 30%. Our docs define visibility and share of voice the way our dashboard computes them, and the same definitions work in a spreadsheet.
The hard part is that answers move. In a peer-reviewed study of Google Search, two identical runs shared about two-thirds of the sources AI Overviews cited (an overlap of 0.66), against 0.78 for ordinary results. Asked the same shopping questions again, whether an answer named any brand at all changed on 11% to 13% of questions. Most of the movement comes from a minority: in our paper, most combinations of brand, question and engine didn't move at all.
77.5%
of brand, question and engine combinations were named every time or never, across repeated runs of unbranded questions (Ranqo, 102 brands, arXiv:2606.20065)
The rest flip, and a few flips move a small score a long way: with 20 questions, each answer is 5 points on one engine. In our tracking, 95% of run-to-run moves in an overall score were 14.5 points or smaller with no fix shipped, as our noise study shows. So report a range, not a number: ask each question three times and give the lowest and highest run, or use the interval for a proportion in chapter 6 of OpenIntro Statistics.
- TextbookOpenIntro Statistics, chapters 5 and 6Confidence intervals for a proportion1.5 h
- StudyHow Generative AI Disrupts Search (SIGIR 2026)30 min
- StudyAuditing AI-Generated Product RecommendationsOptional30 min
- Ranqo researchGEO at scale: 102 brands, 5 enginesThe persistence section30 min
- Ranqo docsNoise bandsWith the Visibility and Share of voice pages beside it15 min
- Ranqo guidePrompt Research: 7 Places to Find Buyer Questions12 min
- Ranqo guideOne in Seven Tracked Prompts Can Never Show You a Win10 min
- Ranqo guideWhy You Can't Measure AI Visibility Just Once10 min
- Ranqo guideWhat Is a Good AI Visibility Score?10 min
- Ranqo toolAI Visibility CheckerOptional5 min
The lab
Write 20 questions a buyer asks before knowing any brand names: problems, comparisons, "best X for Y". Our prompt research guide lists seven places to find them. Ask each in ChatGPT, Perplexity and Google AI Mode three times, on different days, and log every answer in the audit sheet: named or not, position, rivals named, pages cited. That's 180 answers, your capstone baseline. For a quick first read before you start, our free AI Visibility Checker asks 10 such questions on ChatGPT and shows who gets named.
You've learned it when you can look at a 10-point change between two weeks and say whether it's real.
Week 4: The Pages AI Cites
What to learn
The pages AI cites have a shape. In our paper, ranked lists took 35.7% of the citations that went to content pages, more than any other format, and for most buyer questions the cited pages are lists, comparisons and reviews written by someone else. Before writing anything, look at what the engines cite for your question: the format those pages share is the one to beat.
Read the paper that named the field with its method in view. Aggarwal et al. found that adding citations, quotations and statistics to a page raised its visibility by up to 40%, and that stuffing it with keywords did not. Their engine was a simulation: the top five Google results fed to gpt-3.5-turbo, scored on the paper's own measures, with gains that varied by subject. Learn the method from it, and don't quote it as a promise.
Two more sources round out the week. Jiang et al. studied ChatGPT's live search and found that, among the pages it retrieved, a newer publication date predicted citation no better than chance, while the page's position in search results did better. Google's rater guidelines explain what it means by experience, expertise, authoritativeness and trust, and for which questions freshness matters.
- StudyGEO: Generative Engine Optimization (KDD 2024)Read the method and the limits, not just the abstract45 min
- StudyUnderstanding ChatGPT Citations30 min
- Search engineSearch Quality Rater GuidelinesThe sections on E-E-A-T and on freshness1 h
- Ranqo guideThe 5 Page Types AI Cites Most10 min
- Ranqo guideHow to Publish a 'Best X' List AI Will Cite12 min
- Ranqo guideThe AI Page Refresh Checklist12 min
- Ranqo guideThe Page Checks Most Sites Fail for AI SearchOptional12 min
- Ranqo toolAI Content Grader5 min
- Ranqo toolAI Readiness Score5 min
The lab
Take the page on your site that should answer your top question. Put it beside the pages the engines cited in week 1 and list what they have that it lacks: a ranked list, a comparison table, named sources, a direct answer near the top. Run it through our free AI Content Grader and AI Readiness Score for a second opinion, then rewrite and publish it. This is the fix your capstone measures.
You've learned it when you can name the format AI cites for your top question, and say why your page isn't it yet.
Week 5: The Sources You Don't Own
What to learn
Most citations go to other people's pages. In our paper, the brand's own website earned a small share of the 149,912 citations we measured; the rest went to other companies' pages, media, video, forums, Wikipedia and review sites.
2.9%
of 149,912 AI citations went to the brand's own website (Ranqo, 102 brands, arXiv:2606.20065)
Off-site work starts with finding the pages that shape your category's answers, then earning a place on them: a correction to a list, a data point for its editor, a genuine answer in a thread. Our citation gap guide shows how to find the pages that cite your rivals but not you.
Learn the rules of each place before you work in it. Wikipedia's guideline says conflict-of-interest editing is strongly discouraged, and paid editors must disclose. Each subreddit sets its own rules. Breaking them costs more than any citation is worth.
- Ranqo researchYour Website Is Only 2.9% of AI Citations12 min
- Ranqo guideOff-Site GEO12 min
- Ranqo guideCitation Gap Analysis15 min
- Ranqo guideHow to Earn a Spot on the 'Best X' Lists AI Cites12 min
- Ranqo guideReddit Marketing for AI Visibility15 min
- GuidelineWikipedia: Conflict of interest15 min
The lab
For your five most important questions, list every page cited in your week 3 answers. Mark the ones that name you and the ones that name a rival but not you; those are your gaps. Pick three you can genuinely improve, and pitch each one: a correction, a source, a product that belongs on the list.
You've learned it when you can name the ten pages that shape your category's answers.
Week 6: Proving a Change
What to learn
Proving a change means comparing like with like: the same questions, on the same engines, asked the same number of times. Count the answers you won and lost rather than the move in a score, as our guide to measuring GEO results explains. A change inside your week 3 range is noise until it repeats. Our docs set out when we count a win as confirmed and how long a measurement window runs.
Traffic is the other half. AI assistants send visitors, but many arrive unlabelled; our GA4 guide sets up the channel and shows what it misses. And Bing Webmaster Tools now has an AI Performance report that counts how often Copilot and Bing's AI summaries cite your pages: a citation count straight from an engine.
- Ranqo guideHow to Tell If a GEO Change Worked12 min
- Ranqo docsConfirmed winsWith the Measurement windows page beside it10 min
- Ranqo guideHow to Track AI Traffic in GA415 min
- Ranqo guideDark AI TrafficOptional10 min
- Search engineIntroducing AI Performance in Bing Webmaster Tools10 min
The lab
Ask your 20 questions again, three runs on each engine, and fill in a second sheet. Compare each engine's new range with week 3's. Then set up the AI assistant channel in GA4 and add your site to Bing Webmaster Tools, so next month's check has traffic and Copilot citations beside it.
You've learned it when you can show a boss or a client a change that isn't noise.
The Capstone: One Brand, Six Weeks
The labs add up to one project: measure a brand, change something, measure it again. It's the loop all GEO work runs on, done by hand once so you know what every number in a dashboard means.
The capstone, one brand from week 3 to week 6
The same questions, engines and number of runs before and after the fix.
| When | What you do | Answers read |
|---|---|---|
| Week 3 | Baseline: 20 buyer questions on 3 engines, 3 runs each | 180 |
| Weeks 4 and 5 | One page rewritten, three cited pages pitched | None |
| Week 6 | The same questions, engines and runs again | 180 |
Working out a rate and its range
Each run is 60 answers: 20 questions on 3 engines. Say the brand is named in 15, 18 and 21 of them. The three runs are 25%, 30% and 35%, and across all 180 answers the brand was named 54 times, or 30%. Report 30%, with a range of 25% to 35%. In a spreadsheet with named in column E, the overall rate is =COUNTIF(E2:E181,"yes")/180.
In week 6, an engine whose rate falls outside its old range deserves a second look, and one that stays outside it for another run has changed. For a sense of scale over many runs: our product treats a move of 15 points or less in an overall score as noise.
The audit sheet
One row per answer. Copy the header and the three example rows into a new sheet, delete the examples, and fill in a row for every answer.
date,engine,run,question,named,position,rivals_named,cited_pages
2026-11-14,ChatGPT,1,"best project tool for a 10-person agency",no,,"Rival A; Rival B","example.com/best-tools; example.org/review"
2026-11-14,Perplexity,1,"best project tool for a 10-person agency",yes,3,"Rival A","example.com/best-tools"
2026-11-14,Google AI Mode,1,"best project tool for a 10-person agency",no,,"Rival B; Rival C","example.net/compare"Six Things Courses Teach That the Sources Don't Support
You'll meet these in courses and in client requests. Each has a primary source you can point to, and the date that source says so.
6 claims courses still teach
What each claim's own source says, and since when.
| Courses teach | What the source says | Since |
|---|---|---|
| Add FAQ schema so AI can lift your answers | Google says its AI features need no special schema.org markup, and its FAQ rich result stopped showing on 7 May 2026. (Google's guide, FAQ removal) | May 2026 |
| Publish an llms.txt for AI engines | Google Search ignores the file, no other engine documents reading it, and 2.13% of desktop pages had a valid one in July 2025. (Google's guide, Web Almanac) | 15 June 2026 |
| Block Google-Extended to stay out of AI Overviews | Google-Extended doesn't affect Search. The Search generative AI control in Search Console removes a site from AI Overviews and AI Mode. (Google's crawlers, The control) | 31 August 2026 |
| Update the date and AI will cite you | Among the pages ChatGPT retrieved, publication recency predicted citation no better than chance; search-result position did better. (Jiang et al.) | October 2026 |
| Adding statistics and quotations lifts visibility 40% | That was a relative gain on the paper's own measures, in an engine it simulated with gpt-3.5-turbo, and results varied by subject. (Aggarwal et al.) | November 2023 |
| Check once to see where you stand | Asked again, whether an answer named any brand at all changed on 11% to 13% of questions, and an overall score moved up to 14.5 points with nothing changed. (Uberti-Bona Marin et al., our noise study) | 2026 |
For twelve more, with the evidence for each, see our GEO myths post.
How to Keep Up After Week 6
Watch the pages that change first
- Google Search Central's documentation updates, where the FAQ and llms.txt changes were posted.
- The OpenAI, Anthropic and Perplexity crawler pages from week 2, which list each bot and its addresses.
- The Bing blog, for Copilot and Bing Webmaster Tools.
- The IETF's AI Preferences working group, which is writing a shared vocabulary for how AI may use your content.
Read a new study with three questions
Who ran it, and what do they sell? Which engine, and when? How many times was each question asked? A study run once, on one engine, by a company selling the fix tells you less than its headline.
Take the syllabus with you
The whole curriculum as plain text, with every link:
FREE GEO AND AEO CURRICULUM: 6 WEEKS, ABOUT 23 HOURS
BEFORE WEEK 1 (optional, new to search)
- How Google Search Works: developers.google.com/search/docs/fundamentals/how-search-works
- SEO Starter Guide: developers.google.com/search/docs/fundamentals/seo-starter-guide
- What Is Generative Engine Optimization (GEO)?: ranqo.ai/blog/what-is-generative-engine-optimization-geo-guide
- What Is Answer Engine Optimization (AEO)?: ranqo.ai/blog/what-is-answer-engine-optimization-aeo
- GEO vs AEO vs SEO: ranqo.ai/blog/geo-vs-aeo-vs-seo
WEEK 1: HOW AI ANSWERS ARE BUILT (about 4 h)
Read:
- Optimizing your website for generative AI features on Google Search: developers.google.com/search/docs/fundamentals/ai-optimization-guide - Expanding AI Overviews and introducing AI Mode: blog.google/products-and-platforms/products/search/ai-mode-search/
- Searching the web with ChatGPT: help.openai.com/en/articles/9237897-searching-the-web-with-chatgpt
- How does Perplexity work?: www.perplexity.ai/help-center/en/articles/10352895-how-does-perplexity-work
- Enable and use web search (Claude): support.claude.com/en/articles/10684626-enable-and-use-web-search
- Retrieval Augmented Generation in Azure AI Search: learn.microsoft.com/en-us/azure/search/retrieval-augmented-generation-overview
- Speech and Language Processing, retrieval and RAG chapters (optional): web.stanford.edu/~jurafsky/slp3/
- Query Fan-Out: What AI Engines Actually Search: ranqo.ai/blog/query-fan-out
- What Google's 2026 AI Search Guide Adds: ranqo.ai/blog/google-ai-optimization-guide-2026
- How to Get Cited by ChatGPT: The Two-ChatGPTs Playbook: ranqo.ai/blog/how-to-get-cited-by-chatgpt
- Query Fan-out Generator (optional): ranqo.ai/free-tools/query-fanout-generator
Do: Ask one buyer question on three engines and log every page each cites.
Done when: You can explain why two engines cited different pages.
WEEK 2: WHICH CRAWLERS CAN READ YOUR SITE (about 3 h)
Read:
- RFC 9309: Robots Exclusion Protocol: www.rfc-editor.org/rfc/rfc9309.html
- Overview of OpenAI Crawlers: developers.openai.com/api/docs/bots
- Does Anthropic crawl data from the web?: support.claude.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler
- Perplexity Crawlers: docs.perplexity.ai/docs/resources/perplexity-crawlers
- Google's common crawlers: developers.google.com/crawling/docs/crawlers-fetchers/google-common-crawlers
- Search generative AI control: support.google.com/webmasters/answer/16908024
- Bing Webmaster Guidelines: www.bing.com/webmasters/help/webmaster-guidelines-30fba23a
- Clearer preferences for training and grounding: blogs.bing.com/search/2026/10/your-content-your-choice-clearer-preferences-for-training-and-grounding/
- Every AI Crawler User Agent in 2026, Ranked: ranqo.ai/blog/ai-crawler-user-agents
- Is Your Site Blocking AI Crawlers?: ranqo.ai/blog/is-your-site-blocking-ai-crawlers
- What AI Actually Sees When It Crawls Your Site: ranqo.ai/blog/what-ai-sees-when-it-crawls-your-site
- How to Verify Googlebot, GPTBot and ClaudeBot in Your Logs (optional): ranqo.ai/blog/verify-googlebot-and-ai-crawlers
- AI Crawler Inspector: ranqo.ai/free-tools/ai-crawler-inspector
Do: Read your robots.txt line by line and fetch a key page as each AI crawler.
Done when: You can say which bot block costs citations and which costs nothing.
WEEK 3: MEASURING WITHOUT FOOLING YOURSELF (about 5 h)
Read:
- OpenIntro Statistics, chapters 5 and 6: www.openintro.org/book/os/
- How Generative AI Disrupts Search (SIGIR 2026): arxiv.org/abs/2604.27790
- Auditing AI-Generated Product Recommendations (optional): arxiv.org/abs/2609.18729
- GEO at scale: 102 brands, 5 engines: ranqo.ai/research/geo-at-scale
- Noise bands: ranqo.ai/docs/methodology/noise-bands
- Prompt Research: 7 Places to Find Buyer Questions: ranqo.ai/blog/prompt-research
- One in Seven Tracked Prompts Can Never Show You a Win: ranqo.ai/blog/choosing-prompts-to-track
- Why You Can't Measure AI Visibility Just Once: ranqo.ai/blog/why-you-cant-measure-ai-visibility-once
- What Is a Good AI Visibility Score?: ranqo.ai/blog/what-is-a-good-ai-visibility-score
- AI Visibility Checker (optional): ranqo.ai/free-tools/ai-visibility-checker
Do: Ask 20 buyer questions on three engines, three times each, and record who gets named.
Done when: You can say whether a 10-point change between two weeks is real.
WEEK 4: THE PAGES AI CITES (about 4 h)
Read:
- GEO: Generative Engine Optimization (KDD 2024): arxiv.org/abs/2311.09735
- Understanding ChatGPT Citations: arxiv.org/abs/2610.15382
- Search Quality Rater Guidelines: static.googleusercontent.com/media/guidelines.raterhub.com/en//searchqualityevaluatorguidelines.pdf
- The 5 Page Types AI Cites Most: ranqo.ai/blog/page-types-ai-cites-most
- How to Publish a 'Best X' List AI Will Cite: ranqo.ai/blog/best-x-list-checks
- The AI Page Refresh Checklist: ranqo.ai/blog/ai-content-refresh-checklist
- The Page Checks Most Sites Fail for AI Search (optional): ranqo.ai/blog/page-checks-sites-fail-ai-search
- AI Content Grader: ranqo.ai/free-tools/ai-content-grader
- AI Readiness Score: ranqo.ai/free-tools/ai-readiness-score
Do: Compare your page for your top question with the pages AI cited, then rewrite it.
Done when: You can name the format AI cites for that question, and why your page isn't it.
WEEK 5: THE SOURCES YOU DON'T OWN (about 4 h)
Read:
- Your Website Is Only 2.9% of AI Citations: ranqo.ai/blog/where-ai-citations-come-from
- Off-Site GEO: ranqo.ai/blog/off-site-geo
- Citation Gap Analysis: ranqo.ai/blog/citation-gap-analysis
- How to Earn a Spot on the 'Best X' Lists AI Cites: ranqo.ai/blog/how-to-get-on-best-x-ai-lists
- Reddit Marketing for AI Visibility: ranqo.ai/blog/reddit-outreach-for-ai-visibility
- Wikipedia: Conflict of interest: en.wikipedia.org/wiki/Wikipedia:Conflict_of_interest
Do: List every page cited for your top five questions and mark which ones name you.
Done when: You can name the ten pages that shape your category's answers.
WEEK 6: PROVING A CHANGE (about 3 h)
Read:
- How to Tell If a GEO Change Worked: ranqo.ai/blog/how-to-measure-geo-results
- Confirmed wins: ranqo.ai/docs/methodology/confirmed-wins
- How to Track AI Traffic in GA4: ranqo.ai/blog/track-ai-traffic-ga4
- Dark AI Traffic (optional): ranqo.ai/blog/dark-ai-traffic
- Introducing AI Performance in Bing Webmaster Tools: blogs.bing.com/webmaster/2026/2/Introducing-AI-Performance-in-Bing-Webmaster-Tools-Public-Preview/
Do: Ask the same 20 questions again and compare the change with your range.
Done when: You can show a boss or client a change that isn't noise.
CAPSTONE
Week 3: 20 buyer questions x 3 engines x 3 runs = 180 answers.
Weeks 4-5: rewrite one page, pitch three cited pages.
Week 6: the same 180 answers again. Count the change only if it is bigger than your range.Practise on your own brand
Start with the free AI Visibility Checker for a first read on ChatGPT. When you want week 3's questions asked on several engines every week, with the noise accounted for, Ranqo runs that loop. For why one check is never enough, see why you can't measure AI visibility once.
Start a 7-day trialWritten by
Nisha Kumari
Nisha Kumari is Co-Founder at Ranqo, where she leads growth strategy and client acquisition. With a background in digital marketing and financial management, she specializes in SEO, Generative Engine Optimization, and helping brands build visibility across AI platforms.
Share this article
Related articles
What is Generative Engine Optimization (GEO)? The Complete 2026 Guide
Generative Engine Optimization (GEO) is the practice of optimizing content to be cited by AI platforms like ChatGPT, Perplexity, Gemini, Claude, and Grok. This complete guide covers what GEO is, how it works, why it matters in 2026, and how to implement it, with data from every major study published in the last 12 months.
Why You Can't Measure AI Visibility Just Once
Run the same AI query twice and you may get a different answer. So is AI visibility just noise? Mostly not: in our 102-brand study, 77.5% of brand-prompt-engine cells were deterministic, always cited or never. But sentiment flips 6.7x more often than whether you're mentioned. The fix is boring and it works: measure the same prompts repeatedly.
12 GEO Myths That Cost Brands Citations
FAQ schema unlocks citations. High DR transfers to AI. Reddit posting drives Perplexity citations. Once cited, always cited. None of them is supported by the primary research. Twelve of the most expensive GEO misconceptions, with the operator-level fix for each.