Generative engine optimization (GEO)

Generative engine optimization, GEO, means being named and cited in answers written by AI assistants. Think ChatGPT with search, Perplexity, Google AI Overviews and AI Mode, Claude, Gemini, and Microsoft Copilot. You optimise to be quoted, not just ranked.

By , founder of Porteur · Updated 13 September 2026 · Markdown

What it means for a small site

GEO is the practice of getting your page cited as a source in AI answers. There is no submission form. Assistants fetch from web indexes and pick pages that answer the question directly.

For a small site, GEO is a route to discovery without owning the whole results page. A single citation can put your brand in front of buyers who never click.

It differs from SEO. SEO aims to rank and win the click. GEO aims to be named inside the answer. You still need SEO for traffic and revenue tracking.

How to read and measure GEO

Measure GEO by asking assistants the questions your buyers ask, then reading who is named. Repeat for ten to twenty core queries, weekly or monthly.

  • List your buyer queries: “best payroll for startups”, “how to migrate from X to Y”, “pricing page examples”.
  • Ask ChatGPT with search, Perplexity, Copilot, Gemini, and Google AI Overviews, as of 2026 those show sources.
  • Record which pages are cited, where you appear, and which rivals recur.
  • Note the formats assistants choose: lists, comparisons, step-by-steps, figures, and consistent product and person names.

A good result is your brand named with a link. A better result is your comparison page quoted with a figure the answer repeats.

What to do to be cited

  • Write pages that answer the exact question in plain terms: lists, pros and cons, steps, specs, and prices if you publish them.
  • Keep names consistent: your product, plan names, people, and company. assistants look for exact matches.
  • Use clear structure: h2 sections, bullet lists, comparison tables, and short summaries.
  • Quote figures assistants can lift: limits, counts, integrations, version numbers, dates, and example queries in quotes.
  • Publish comparison and alternatives pages that include you and competitors you actually replace.
  • Get onto existing roundups already cited. Look at the cited sources, ask for inclusion, or publish a better, cleaner version that those pages may cite.
  • Let answering crawlers fetch your pages. Check robots.txt and server rules on both www and bare domains.

Example: your /alternatives/yourproduct page adds a table of 8 tools, pricing, and who each is for. Perplexity starts citing it on “yourproduct alternatives”. Before, you were not named. After, your brand appears in the first screen with a link.

Crawlers to allow and why

Vendor or botPurpose as of 2026Blocking effect
GooglebotGoogle Search and pages cited by AI OverviewsRemoves you from Search and AI Overviews sourcing
BingbotBing Search and Microsoft CopilotRemoves you from Bing and Copilot sourcing
OAI-SearchBotIndexing for ChatGPT search resultsHides you from ChatGPT sourced answers
ChatGPT-UserFetches a page on user requestMay break on-demand fetching of your page
Claude-SearchBotIndexing for Claude search answersHides you from Claude sourced answers
Claude-UserFetches on user requestMay break on-demand fetching of your page
PerplexityBotPerplexity indexHides you from Perplexity citations
Perplexity-UserOn-demand fetchMay block real-time fetching
ApplebotSiri and SpotlightLimits Apple surfaces
Applebot-ExtendedControl for Apple Intelligence trainingTraining only, not answer visibility
Google-ExtendedControl for Gemini training and groundingTraining only, not Search ranking or Overviews
GPTBotModel training crawlTraining only
ClaudeBotTraining and model improvementTraining only
CCBotCommon Crawl corpusTraining only
meta-externalagent, meta-externalfetcherMeta training and fetcherTraining only
Amazonbot, cohere-ai, Bytespider, DuckAssistBotVendor-specific usesUsually training or assistant features

Robots.txt is voluntary. Blocking training crawlers does not remove pages already collected. If you must block training, keep OAI-SearchBot, Claude-SearchBot, PerplexityBot, Bingbot, and Googlebot allowed for GEO visibility.

Traps and limits

  • Chasing clicks only. GEO value is brand mention and assisted demand, which analytics may not attribute.
  • Thin programmatic pages. Assistants prefer clear, specific pages, not spun templates.
  • Inconsistent naming. One page says Starter, another says Basic. Assistants split your mentions.
  • No source hygiene. If your page misstates a figure, you lose trust and future citations.
  • Blocking by mistake. A stray Disallow on OAI-SearchBot or PerplexityBot makes you vanish from answers.
  • Expecting a submission form. There is none. Assistants choose from indexes and existing roundups.

Questions

Sources

Check my site, free

Drop your URL to see which assistant surfaces cite rivals for your queries, and what your site needs to be cited, in about thirty seconds, free.

  • Free check, no card
  • Read-only, your own accounts
  • Readable by your agent

Read next