GROWTH WITH ALEX
SEO / GEOPublished November 18, 2026·12 min read

The Ultimate Guide to GEO and AEO: Getting Cited by ChatGPT, Gemini, Claude and AI Overviews

What generative engine optimization (GEO) and answer engine optimization (AEO) really are, what the research and Google's own documentation say works, how big AI referral traffic actually is, and a step-by-step plan to measure and improve your brand's visibility in AI answers.

By Alexandre Contador, Fractional CGO & GTM Advisor

Person working on a laptop at a wooden table
Photo by Myriam Jessier on Unsplash

Buyers have started asking assistants for shortlists. "What's the best CRM for a 20-person SaaS team?" now gets a synthesised answer from ChatGPT, Claude, Gemini or Google's AI Overviews — often with a handful of named vendors and no ten blue links. If your brand isn't in that answer, a growing share of the buying journey happens without you.

The good news is that the fundamentals of earning those mentions are knowable, and a surprising amount of them are plain good SEO. This guide separates what's proven from what's hype, shows how large the channel really is, and gives you a repeatable plan to measure and improve your visibility.

SEO vs. AEO vs. GEO: what the terms mean

The vocabulary is messy, so define it once:

TermGoalWhere it applies
SEORank pages in search resultsGoogle, Bing
AEO (answer engine optimization)Be the source of a direct answerFeatured snippets, AI Overviews, voice and chat answers
GEO (generative engine optimization)Increase how often and how well your content appears in generated responsesChatGPT, Claude, Gemini, Perplexity, AI Mode
LLMO / AI search optimizationUmbrella terms used interchangeablyEverything above

The term GEO comes from a research paper by Aggarwal and colleagues, later presented at KDD 2024. They introduced it as "a black-box framework" for helping content creators improve visibility in generative engine responses, along with a benchmark (GEO-bench), and reported that the right strategies can boost visibility by up to 40% — with effectiveness varying by domain. Treat that figure as a best case under experimental conditions, not a promise.

The honest summary: these are overlapping skill sets, and whichever label you use, the work is the same — be easy to find, easy to trust and easy to quote.

How AI search actually works (and why it matters)

Understanding the mechanism tells you where to intervene.

  1. Training knowledge. The model "knows" brands and concepts from its training data. You influence this slowly, through the volume and consistency of how you are described across the web.
  2. Retrieval and grounding. For many queries, the assistant searches the live web (or an index), pulls candidate pages and writes an answer grounded in them. This is where your pages, structured facts and freshness matter.
  3. Query fan-out. Google notes that AI Overviews and AI Mode may use "query fan-out" — running multiple related searches across subtopics and data sources — which can surface a wider and more diverse set of supporting links. One user question becomes many retrieval queries, so topical breadth and depth beat a single optimised page.
  4. Synthesis and selection. The model decides which sources to cite and which brands to name, weighing relevance, apparent authority, consistency across sources and how quotable the passages are.
Person using a laptop in a dark workspace
One buyer question fans out into many retrieval queries — cover the subtopics, not just the head term. · Photo by Myriam Jessier on Unsplash

What Google itself says

Google's AI features documentation is unusually clear:

  • No special requirements. A page needs to be indexed and eligible to appear in Search with a snippet. Indexing and serving are still not guaranteed.
  • Standard best practices apply: allow crawling in robots.txt and CDN/hosting settings, use internal links, provide a good page experience, make important content available as text, support it with quality images and video, and keep structured data consistent with the visible text.
  • You don't need new machine-readable files, AI-specific text files, special markup or special schema.
  • Controls exist. Use nosnippet, data-nosnippet, max-snippet or noindex to limit what's shown, and Google-Extended to limit use for AI training and grounding in other Google systems.
  • Measurement is blended. Traffic from AI features is included in the Search Console Performance report under the "Web" search type.

Google's separate guidance on people-first content lists warning signs that map neatly onto GEO tricks to avoid: content made mainly to attract search traffic, heavy reliance on automation, chasing a "preferred" word count, and changing dates just to seem fresh.

How big is AI search traffic, really?

Hype and dismissal are both wrong. The most useful public dataset we found is Previsible's 2026 State of AI Discovery report: 166 GA4 properties tracked from November 2024 to May 2026, covering 6.77 million LLM-driven sessions (AI Overviews excluded).

9.9×

growth in monthly LLM referral sessions, Nov 2024 → May 2026 (65,249 → 644,478)

Source: Previsible

92.4%

of trackable standalone LLM referral traffic came from ChatGPT

Source: Previsible

0.17%–1.71%

AI penetration of total sessions by industry (health lowest, SMB highest)

Source: Previsible

Monthly LLM referral sessions by platform, May 2026
ChatGPT610910 sessions
Gemini18119 sessions
Claude8528 sessions
Perplexity6788 sessions
Copilot339 sessions

Source: Previsible, 2026 State of AI Discovery (166 GA4 properties)

Three conclusions:

  • Volume is small but compounding. Nearly tenfold growth in 18 months, still well under 2% of sessions in any industry the report measured.
  • ChatGPT dominates referrals, but not influence. Other datasets give different splits (one analysis puts ChatGPT near 75%, another near 63% for a B2B panel), so treat any single share as dataset-specific. The trend lines are also volatile: in this report, Claude grew about 64× while Perplexity fell 61% from its peak.
  • Clicks understate impact. Referral sessions only capture people who click through. A buyer who sees you recommended and later types your name into Google shows up as direct or branded traffic. That is why prompt-level monitoring (below) matters.

The GEO/AEO playbook, step by step

Step 1: Fix the foundations (crawlability and indexation)

If a page can't be crawled, indexed or shown with a snippet, nothing else matters. Check:

  • robots.txt and CDN/WAF rules aren't blocking Googlebot, Bingbot or the AI crawlers you want (GPTBot, ClaudeBot, PerplexityBot and similar — decide deliberately, since blocking them reduces your chance of being cited).
  • Key pages are indexable, in your XML sitemap and reachable via internal links.
  • Important content is in HTML text, not hidden in images, tabs that don't render, or heavy client-side JavaScript.
  • Bing indexation is healthy; some assistants can draw on Bing-style indexes for retrieval, so it's worth checking.

Step 2: Build a prompt map of your buyers' questions

List the questions a buyer asks an assistant at each stage:

StageExample prompts
Problem"Why is our trial-to-paid conversion low?"
Category"What is product-led growth and does it work for B2B?"
Shortlist"Best [category] tools for a 50-person SaaS company"
Comparison"[Your brand] vs [competitor]: which is better for…"
Validation"Is [your brand] worth it? Reviews and pricing"

Aim for 30–100 prompts across personas, then cluster them by intent. These become both your content roadmap and your measurement set. Use the free LLM Visibility Audit to run a first 20-prompt baseline across ChatGPT, Perplexity and Gemini.

Step 3: Create "answer-ready" content

Models quote passages, not pages. Write so a passage can be lifted without losing meaning:

  • Lead with the answer. Put a 1–3 sentence direct answer under each question-style heading, then elaborate.
  • Make claims specific and citable. Include numbers, dates, definitions, comparisons and named sources. The GEO research found adding quotations, statistics and citations was among the most effective tactics.
  • Use clear structure. Descriptive H2/H3s, short paragraphs, tables for comparisons, numbered steps for processes.
  • Show who is speaking. Bylines, credentials, first-hand experience and original data — Google's E-E-A-T framing treats trust as the most important element.
  • Cover the fan-out. For each core topic, publish the adjacent subtopics (pricing, alternatives, implementation, mistakes, benchmarks) and interlink them. That's the content cluster model this site uses for GTM, AI and PLG.
  • Keep content current honestly. Update when something changes, and say what changed. Don't fake freshness.

Step 4: Add "own the facts" pages

Assistants repeatedly need the same facts about you. Make them unambiguous and consistent everywhere:

  • What you are in one sentence; who it's for; what it costs; how it compares.
  • A comparison page per major alternative, fair and specific. (If you're a tool vendor, "best X" lists and "X vs Y" pages are frequent citation targets.)
  • A pricing page in plain text.
  • An "about"/entity page with founders, dates, locations and links to authoritative profiles.
  • Structured data where it genuinely mirrors visible content (Organization, Product, FAQ, Article). Google is explicit that no special schema is needed, so use it for clarity, not as a ranking hack.

Step 5: Earn third-party mentions

Models weigh what others say about you. Prioritise:

  1. Review and directory sites relevant to your category (G2, Capterra and niche directories).
  2. Comparison and "best of" lists written by credible third parties.
  3. Communities where buyers ask: Reddit, Slack/Discord groups, LinkedIn discussions, Hacker News for dev tools.
  4. Digital PR and expert quotes: be the named source for data journalists and analysts.
  5. Podcasts, YouTube and webinars, which produce transcripts that models can ingest.
  6. Partner and integration pages from the tools you work with.

This is the slowest lever and often the highest impact. It overlaps with classic link building but optimises for mentions in context, not just links.

Step 6: Make your site friendly to retrieval

  • Fast, stable, mobile-friendly pages; avoid interstitials that block content.
  • Descriptive titles and meta descriptions; one primary topic per page.
  • Canonical tags and clean URL structure so signals consolidate.
  • Internal links with descriptive anchors from related pages (this is how the model discovers the cluster).
  • Optional: an llms.txt or similar summary file. Google says it is unnecessary for its features; other tools may read it, so it's low-cost housekeeping — not a strategy.

Step 7: Monitor and iterate

Run your prompt set on a schedule and log, for each prompt and each assistant: are you mentioned, in what position, with what framing, citing which URLs, and which competitors appear? Look for patterns — consistently missing on comparison prompts, wrongly described, cited only from old pages — and map each gap to a fix.

You can do this manually in a spreadsheet, or use a monitoring tool. (Full disclosure: Alex co-founded Rankahead, which tracks prompts, citations and competitors across assistants; the process above works without it.) The free GEO Playbook, AI Search checklist and 11 SEO/AEO workflows are practical starting points.

A 30-60-90 day plan

WindowFocusOutput
Days 1–30Foundations and baselineCrawl/indexation fixes; prompt map; baseline mention rate per assistant; list of citation sources
Days 31–60Content and facts5–10 answer-ready pages; comparison and pricing pages; schema aligned to visible content; internal linking
Days 61–90Off-site and iteration10+ third-party placements pursued; monthly prompt re-test; update losing pages; report mention rate trend

Measurement: what to track

MetricHow to get itWhy it matters
Mention rate% of tracked prompts where your brand appearsCore visibility KPI
Citation rate% of answers that link to your domainShows which pages models trust
Share of voiceYour mentions vs competitors across the prompt setCompetitive context
Sentiment/accuracyManual review of how you're describedCatches wrong or outdated claims
AI referral sessionsAnalytics segment by source (chatgpt.com, perplexity.ai, claude.ai, gemini.google.com…)Direct traffic impact
Branded search liftSearch Console branded queries over timeCaptures the click-later effect
Assisted conversionsCRM "how did you hear about us" + analyticsTies visibility to pipeline

Add a self-reported "how did you hear about us" field to your forms and demo flow. In many B2B funnels, "ChatGPT" and "AI" show up there long before they show up in referral data.

Common mistakes

  • Treating GEO as a separate silo. Fix SEO fundamentals first. Most "AI visibility problems" are crawlability, thin content or weak brand signals.
  • Blocking AI crawlers by accident, or leaving key facts only in PDFs and images.
  • Publishing mass-produced AI content. It's low-trust, easy to ignore, and contradicts Google's people-first guidance.
  • Keyword stuffing and fake freshness. Google's guidance warns against both, and early GEO experiments reportedly found stuffing didn't help.
  • Ignoring off-site signals. If every third-party source describes a competitor and not you, your own site can't overcome it.
  • Measuring once. Answers change run to run and model to model. Use a fixed prompt set, repeated runs and trends, not a single screenshot.
  • Chasing traffic instead of qualified demand. The goal is being the recommended option for the right buyers, not maximum sessions.

How GEO fits your wider GTM strategy

AI visibility is a distribution layer on top of positioning. If your category language is vague or your ICP is fuzzy, models will describe you vaguely too. The best GEO work usually starts upstream: a sharp positioning statement, a clear ICP and consistent messaging everywhere buyers might encounter you. From there, connect your measurement stack with agents — see the guides on MCP for marketers and agentic GTM workflows for how to automate prompt monitoring and reporting safely.

Next steps

  1. Run the free LLM Visibility Audit to get your baseline.
  2. Work through the AI Search checklist.
  3. If you run WordPress, the WordPress AEO autopilot playbook shows how to run it from your dashboard with Claude.
  4. For a done-with-you version, see the SEO & GEO service or the GTM readiness score.

Frequently asked questions

SEO aims to rank pages in a list of links. AEO (answer engine optimization) aims to be the source an assistant uses when it gives a direct answer. GEO (generative engine optimization) is the research term for improving how your content shows up in generated responses. In practice they overlap heavily: Google's own guidance says no special optimization is required for its AI features beyond standard SEO fundamentals.

Google's documentation states you do not need new machine-readable files, AI-specific text files or special schema to appear in AI Overviews or AI Mode. Other AI products may read such files, but there is no evidence it is a requirement. Treat it as optional housekeeping, never as a substitute for good content and crawlability.

Traffic from Google's AI features is included in the Search Console "Web" performance data, with no separate filter. For assistants such as ChatGPT, Claude, Gemini and Perplexity, segment referral sessions in your analytics by source, then supplement with prompt-based monitoring, because many AI answers influence buyers without a click.

Model answers change as retrieval indexes refresh and as models are updated, so there is no fixed timeline. Practically, expect weeks for content and technical fixes to be re-crawled and months to build the third-party mentions that influence what models say about you. Track a stable prompt set weekly so you can see movement.

Organic inclusion in AI answers is not something you can buy directly. Several platforms are testing or running ad formats around AI answers, but those are separate from citations. Focus on being the most credible, easiest-to-quote source for the questions your buyers ask.

Want a second opinion on your GTM?

30 minutes, no pitch — just a direct conversation.

Book a Strategy Call

Next in AI & Agents

Agentic GTM Workflows: 25 Claude and Claude Code Automations for Marketing, Sales and Ops

A practical guide to building agentic go-to-market workflows with Claude and Claude Code: the difference between workflows and agents, five design patterns, 25 concrete automations across marketing, sales and ops, and a maturity model for rolling them out safely.