Skip to content

Module 08 · GEO · Guide 3 of 4

The GEO Playbook: Tactics to Earn AI Citations

A practical GEO checklist for 2026: AI crawler access, CDN bot settings, quotable content, entity clarity, genuine off-site presence and what llms.txt really does.

By Updated 5 min readAdvanced

What you'll learn

  • Audit whether search and AI answer bots can reach your site, including CDN-level blocking.
  • Decide what llms.txt is worth for your site, given that Google Search ignores it.
  • Retrofit quotable, non-commodity passages into your highest-value pages.
  • Build entity clarity and genuine off-site presence without crossing spam lines.
  • Prioritize GEO work in an order that pays off for a typical content site.

What is GEO? covered how generative engines work, and SEO for AI Mode & AI Overviews covered Google specifically. This guide is the engine-agnostic checklist: what to actually do, organized by layer.

Layer 1: Be retrievable#

AI answers start with retrieval from a search index. If the right bot cannot reach your pages, nothing else matters.

  • Rank in the underlying indexes. Google's AI features retrieve from the normal Search index, so classic SEO is the foundation. Also register with Bing Webmaster Tools: Bing's index feeds Microsoft Copilot, and Bing has been one of the third-party search providers behind ChatGPT search, alongside OpenAI's own crawler.
  • Allow the answer bots, deliberately. Every major vendor now separates training bots from search bots and from user-triggered fetchers. Blocking a training bot does not remove you from that vendor's AI search; blocking its search bot does. A minimal policy that allows AI search and opts out of training looks like this:
robots.txt (excerpt: search yes, training no)
# AI search and answer bots: allow (citations and referral traffic)
User-agent: OAI-SearchBot
User-agent: Claude-SearchBot
User-agent: PerplexityBot
Allow: /
 
# Model training: a business decision. This example opts out.
User-agent: GPTBot
User-agent: ClaudeBot
User-agent: Google-Extended
User-agent: Applebot-Extended
Disallow: /
 
# Google AI Overviews and AI Mode use ordinary Googlebot.
# Google-Extended does NOT remove you from them.

The full bot directory (Meta, Amazon, Apple, Google-Agent and the rest), verification by IP and the trade-offs of each recipe are in AI Crawlers & Bot Access. Paste your own file into the AI crawler checker to see who is allowed.

  • User-triggered fetchers play by different rules. ChatGPT-User, Perplexity-User and Google-Agent fetch a page because a person asked, and their vendors say robots.txt may not apply. Anthropic's Claude-User is the documented exception that honors it.
  • Check your CDN, not just robots.txt. Cloudflare switched new domains to block AI crawlers by default on 2025-07-01, and from 2026-09-15 blocks training and agent bots by default on pages that show ads. Many sites now block AI search without the owner knowing. Check the bot or AI crawler settings in your CDN or WAF and your server logs.
  • Serve content without JavaScript. Many AI crawlers do not execute JavaScript. Confirm with curl that your content is in the initial HTML.
  • Do not opt out of Google by accident. Search Console now has a Search generative AI setting. If someone set it to Exclude, the site will not appear in AI Overviews or AI Mode however good the content is.

Where llms.txt fits#

llms.txt is a proposal from llmstxt.org for a markdown file at /llms.txt that gives a language model a curated map of your most important pages. The format is plain markdown: an H1 with the site name, a > summary, then sections of annotated links. Rankyl publishes one, and it looks like this:

llms.txt
# Rankyl
 
> A free, in-depth curriculum covering SEO, AEO and GEO for web developers.
 
## Docs
- [What is SEO?](https://rankyl.haikalakif.com/docs/foundations/what-is-seo): search optimization from zero
- [What is GEO?](https://rankyl.haikalakif.com/docs/geo/generative-engine-optimization): how AI engines pick and cite sources
- [The GEO Playbook](https://rankyl.haikalakif.com/docs/geo/geo-playbook): concrete tactics to earn AI citations
 
## Optional
- [Full curriculum index](https://rankyl.haikalakif.com/docs): every guide, grouped by module

Be honest about what it does:

  • Google Search does not use it. Google's May 2026 guide lists llms.txt under things you can ignore, and a 2026-06-15 docs update says it is not needed and will "neither harm nor help" visibility in Google Search.
  • No major answer engine has publicly committed to using it for ranking or citation selection as of October 2026.
  • Chrome Lighthouse checks for it in its "Agentic Browsing" category (added May 2026). That is a tooling audit for agent readiness, not a ranking signal.

So treat it as an optional navigation aid for agents and tools that choose to read it. We publish ours because it costs minutes to maintain, not because it moves citations.

Layer 2: Be quotable#

Write so a model can lift a passage confidently, and so a human finds it worth reading:

  • Lead each section with the claim. Answer-first structure (the AEO pattern) is what selection favors. You do not need to cut pages into fragments: Google says chunking is unnecessary.
  • Put numbers in sentences. "Moving to static generation cut our median LCP from 4.1 s to 1.8 s" is quotable; "performance improved significantly" is filler.
  • Attribute claims ("according to CrUX data, May 2026") with links. Passages that look verifiable are easier to trust.
  • Define terms in single, clean sentences. Definitions are among the most quoted shapes on the web.
  • Use real lists and tables. Structure survives extraction; styled div soup does not.
  • Publish original data and first-hand experience. Google's May 2026 guide calls non-commodity content (first-hand reviews, original analysis, expert takes) the thing most likely to matter in the long run. In our view, original data is the strongest single GEO asset, because engines must cite you as the primary source.

Layer 3: Be a clear entity#

Models assemble what they know about you from many sources. Remove ambiguity:

  • Consistent naming of your brand, products and authors across your site and the web
  • Organization and Person JSON-LD with sameAs links to your official profiles (Structured Data, Entity SEO). This helps identification; it is not an AI citation lever, and Google needs no special schema for AI features
  • An about page that answers who, what, since when and for whom in plain language. Models often retrieve about pages for grounding facts
  • Visible author expertise: bios, credentials, publication history (E-E-A-T)
  • Accurate business data: Google Business Profile for local and Merchant Center feeds for products, both of which feed Google's AI answers

Layer 4: Be genuinely present off-site#

Generative engines triangulate. Brands that independent sources discuss get recommended. Ahrefs (May 2025, 75,000 brands) found brand mentions correlated with AI Overview visibility at 0.664 against 0.218 for backlinks: a correlation, not proof, but a strong hint.

  • Communities. Reddit, forums and YouTube are heavily cited by AI search. Genuine participation (helpful answers, not promotion) shows up in answers. See Reddit, Forums & Communities and YouTube SEO.
  • Comparison and review coverage. Independent "best X" articles, review platforms and industry roundups are what engines retrieve for "what's the best..." prompts. Earn inclusion on merit.
  • Digital PR and data stories (Digital PR & Brand Signals) earn citations twice: in classic rankings and in model knowledge.
  • Preferred sources. Since 2026-05-27, sources people star as preferred in Google also surface in AI Overviews and AI Mode. Ask your audience to add you.
  • Wikipedia and Wikidata, only where you genuinely meet notability rules. Self-written promotional entries get deleted.

The compounding loop#

GEO tactics reinforce each other:

original data -> cited by publications -> retrieved by engines
      ^                                          |
      |                                          v
 brand searches  <-  users see citations  <-  AI answers name you

Next: Measuring AI Visibility, proving any of this works.

Check your understanding

0 of 4 answered

  1. 1.Your robots.txt allows every bot, yet ChatGPT and Claude never cite you and your logs show no OAI-SearchBot or Claude-SearchBot hits. The site sits behind Cloudflare. What should you check first?
  2. 2.What is an honest description of llms.txt in October 2026?
  3. 3.You want to appear in ChatGPT search answers but do not want your content used to train OpenAI models. Which robots.txt setup fits?
  4. 4.Which off-site tactic is safest and most durable for AI visibility?

Sources

Primary documentation this guide was checked against, last reviewed 5 October 2026.

  1. Optimizing your website for generative AI features on Google Search · Google Search Central, 2026-05-15
  2. Latest Google Search documentation updates (llms.txt clarification 2026-06-15; spam policies 2026-05-15) · Google Search Central
  3. Overview of OpenAI crawlers · OpenAI
  4. Does Anthropic crawl data from the web, and how can site owners block the crawler? · Anthropic
  5. Content Independence Day: no AI crawl without compensation · Cloudflare, 2025-07-01
  6. The /llms.txt file proposal · llmstxt.org
  7. An analysis of AI Overview brand visibility factors (75K brands studied) · Ahrefs, 2025-05-26

Spotted something wrong or outdated? Report itHow we review

Finished reading?

Mark it complete to track your progress. Saved in this browser, no account needed.