Generative engine optimization (GEO): how to be named by AI answers

You want your product named when ChatGPT, Perplexity or Google shows an answer. This guide shows what changes with AI answers and what to do next. It is written for founders who ship their own site and can edit pages today.

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

What generative engine optimization means now

Generative engine optimization means being named and cited by AI systems. That includes ChatGPT with search, Perplexity, Google AI Overviews and AI Mode, Claude, Gemini and Microsoft Copilot. The term was popularised by a 2023 paper by Aggarwal and colleagues.

The change is simple. A model writes the answer, then cites sources. It picks pages that state facts directly with clear structure. Lists, comparisons, definitions and figures make it easy to quote you.

There is no submit form. Assistants pull from a web index, either Bing’s, Google’s or their own crawl. Your job is to be in those indices, be on the lists answers already cite, and state things so a model can lift them safely.

What changes when a model writes the answer

  • Source selection: assistants favour pages that map cleanly to a question. They look for explicit headings, lists and stable names for products and people.
  • Entity clarity: consistent names matter. Use the same product name, company name and person names across pages, titles and alt text.
  • Citation-friendly structure: short sections, bullet lists for features and pros and cons, and comparison tables help assistants quote you.
  • Coverage over intent buckets: assistants try to answer end to end. A single clear page that covers the task may beat several thin pages.
  • Authority by inclusion: being included in roundups and comparisons that are already cited is a shortcut. Assistants often re-cite what earlier answers used.

Work with this. Write pages that say one thing plainly. Put the quote a model needs near the top. Name the entity, the action and the outcome.

Crawlers you must allow, and what each one does

If the answering and search crawlers cannot fetch you, you are invisible to assistants. robots.txt is honoured voluntarily by vendors. Blocking a training crawler does not remove pages already collected. Blocking the answering and search crawlers is what removes you from answers.

TokenVendorPurpose as of 2026
GooglebotGoogleSearch crawl, AI Overviews citations come from here
Google-ExtendedGoogleControl for Gemini training and grounding, not Search or AI Overviews
BingbotMicrosoftBing Search crawl, also used by Microsoft Copilot
OAI-SearchBotOpenAISearch index for ChatGPT results
ChatGPT-UserOpenAIFetches a page a user asks about on demand
GPTBotOpenAICrawling for model training
Claude-SearchBotAnthropicSearch index for Claude
Claude-UserAnthropicOn demand fetch when a user asks Claude about a page
ClaudeBotAnthropicCrawling for training and model improvement
PerplexityBotPerplexityIndex for Perplexity answers
Perplexity-UserPerplexityOn demand fetch when a user asks Perplexity
ApplebotAppleSiri and Spotlight crawl
Applebot-ExtendedAppleControl for Apple Intelligence training
CCBotCommon CrawlOpen crawl used to train many models
BytespiderByteDanceCrawl for ByteDance properties and training
meta-externalagent / meta-externalfetcherMetaTraining and fetching
AmazonbotAmazonCrawl and training
cohere-aiCohereTraining and fetching
DuckAssistBotDuckDuckGoAssistant fetching

Let the search and answering bots in: Googlebot, Bingbot, OAI-SearchBot, Claude-SearchBot, PerplexityBot. If you want to opt out of model training, control GPTBot, ClaudeBot, Google-Extended and the others. Keep ChatGPT-User, Claude-User and Perplexity-User open so users can fetch your page during a chat.

Add llms.txt at the root if you want a public policy in plain text for AI crawlers. It does not replace robots.txt. It states your stance for model training and assistant access in one place.

The concrete work: pages and patterns that get cited

  • Home and product pages with a one line value statement up top: what it is, for whom, and the key figure a buyer cares about.
  • Feature lists as bullet points near the top. Keep labels consistent across site and docs.
  • Comparison pages that name rivals in the H1 and table headers, and state differences plainly.
  • Alternatives pages that list peers in your space with one line summaries and links to each peer.
  • How-to guides that answer a scoped task end to end on one URL, with a steps list near the top.
  • Pricing that is clear and dated. If you cannot show prices, show how pricing works in bullets.
  • About and team pages that tie names, roles and photos. Models use them for entity grounding.

A fixed page looks like this: yourproduct.com/alternatives names five real alternatives in a table, each with a short line on who it suits, and one paragraph on how you differ. Assistants can cite lines and the table cleanly.

Use structured data where a schema exists, for example Product, Organization, Review, HowTo and FAQ. It clarifies entities and parts of a page. Keep it true to the visible text. Do not stuff claims in JSON-LD you do not show to users.

Be on the lists assistants already cite

There is no submit button. The most reliable route is to appear on pages that AI answers already cite. If a GEO answer cites three roundups and a comparison, be on those four pages next month.

  1. Ask the assistants the buyer’s questions

    Search “best X for Y”, “X alternatives”, “X vs Y”, and task questions like “how to do X in Y” in ChatGPT with search, Perplexity, Google AI Overviews and Claude.

  2. List the cited sources

    Open every citation. Note the domains, the specific URLs and the page types: roundup, comparison, docs, blog.

  3. Qualify where you can be added

    Pick the pages that list vendors or show tables you can slot into. Avoid dead or spammy pages.

  4. Pitch with the snippet they need

    Email the editor with a one line fit, your short description and a table row ready to paste. Match their headings and labels.

  5. Close the loop

    When added, check the assistants again. You should see that page cited, and sometimes your site now cited on related questions.

Example: you sell a meeting notes tool. Perplexity cites three “best meeting notes apps” posts. Get added to those lists with a short line: “YourProduct, meeting notes with AI summaries for teams.” Then update your /alternatives and /comparison pages so assistants have a clean source on your site too.

Write for quoteability: headings, lists and figures

  • One question per page. Use the question in the H1 and a short answer in the first paragraph.
  • Front load facts. Give the definition, the key number or the step list before the story.
  • Use consistent terms. If you call it “Workflows” on /pricing, do not call it “Pipelines” on /docs.
  • Add a comparison table when you mention rivals: headers like Feature, YourProduct, Rival A, Rival B.
  • State dates and versions. “As of September 2026, free plan includes X” helps assistants judge freshness.
  • Quote yourself. A one sentence summary in quotes is often lifted verbatim by models.

Example before: “We offer flexible options for teams of all sizes.” Example after: “Pricing: Free, Pro, Business. As of September 2026, Pro is for 10 seats and includes SSO.” The second line is safer to cite and clearer for buyers.

Technical setup: access, llms.txt and structure

  1. Check robots.txt for AI crawlers

    Allow Googlebot, Bingbot, OAI-SearchBot, Claude-SearchBot and PerplexityBot. Keep ChatGPT-User, Claude-User and Perplexity-User allowed.

  2. Publish llms.txt

    At yourproduct.com/llms.txt state your policy for training crawlers and assistant fetchers. Link to your robots.txt and privacy page.

  3. Fix canonicals and indexability

    Set one canonical per page. No accidental noindex on core pages. Avoid duplicate landing pages for the same intent.

  4. Add structured data

    Use Product, Organization, HowTo, FAQ and Review where they match visible content. Validate them.

  5. Tighten internal links

    Link to /pricing, /alternatives, /comparison and key guides from your header or footer. Use the exact page names in anchors.

  6. Speed and stability

    Aim for 2.5 seconds for the largest contentful paint, 0.1 for cumulative layout shift and 200 ms for interaction to next paint.

A fixed site will have robots.txt that allows answering crawlers, an llms.txt that explains your stance, clean canonicals, and structured data that matches what a user sees on the page.

Measure GEO by asking the assistants

You cannot read a dashboard for GEO yet. You measure it by asking assistants buyer questions and reading who gets named. Track queries and citations over time.

  1. Pick 10 high intent questions

    For a SaaS: “best [category] for [role]”, “[category] alternatives”, “[rival] vs YourProduct”, and “[task] in YourProduct”.

  2. Ask across engines

    Use ChatGPT with search, Perplexity, Google AI Overviews and Claude. Save screenshots and links for each.

  3. Record sources and positions

    Log whether your brand is named, which of your pages is cited, and which third party pages appear.

  4. Close gaps

    If a third party page lists you, but your site is absent, build or improve the equivalent page on your site.

  5. Recheck monthly

    Repeat every 28 days. Assistants change their sources as indices refresh and pages update.

Example: search “YourProduct vs RivalA” in Perplexity. If only RivalA’s page is cited, create /comparison/yourproduct-vs-rivala with a table and a short neutral summary. Ask again next week. You should see it cited when crawled and indexed.

Worked examples on core pages

/pricing: add a three line summary up top. “Plans: Free, Pro, Business. All plans include SSO. As of September 2026, Pro adds 10 seats and API access.” Assistants can lift those lines. They also pair your brand name with “SSO” and “API access” as entities.

/alternatives: list five named alternatives with one line each, then a table with columns Feature, YourProduct, Alt A, Alt B, Alt C, Alt D. Include links to each rival’s homepage. This is cited by roundups and assistants alike.

/guides/getting-started: have a six step list with verbs at the start of each step. Add HowTo structured data that matches the text. Assistants parse the steps and will cite them when asked “how to set up YourProduct”.

/comparison/yourproduct-vs-rival: open with a short neutral summary, then a table. Add a one sentence quote that an assistant can use: “YourProduct is simpler to deploy for small teams, Rival is deeper for large orgs.”

Questions

Sources

Check my site, free

Paste a URL to get a free GEO check that reads your site, the searches around it and the rivals on them in about thirty seconds, and shows three findings whole.

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

Read next