Generative Engine Optimization (GEO): A Practical Guide

Generative engine optimization (GEO) is the practice of shaping content so AI search tools such as ChatGPT, Perplexity, Claude, and Google AI can easily find, understand, and cite it. It is not a trick. It means content that is clear, structured, sourced, and reachable by crawlers. No tactic guarantees a citation.

Updated November 7, 2026

What is generative engine optimization, and how is it different from SEO?

Generative engines answer a question by summarizing information from many pages, then show some of those pages as links. Researchers formalized the idea in a paper titled "GEO: Generative Engine Optimization", first posted to arXiv in November 2023 (Aggarwal et al., arXiv:2311.09735).

Traditional SEO aims to get your page into a list of results. GEO aims to get your content chosen, summarized correctly, and cited inside an answer. Both rest on the same foundation: pages that can be crawled, content that helps people, and sources that can be trusted.

So do not treat GEO as a replacement for SEO. Treat it as a way of writing and organizing content so that pieces of it are easy to lift into an answer.

Does GEO actually work? What do the research and Google say?

The GEO paper reports that the optimization strategies it tested can raise visibility by up to 40% in generative engine responses, measured on a benchmark the authors built called GEO-bench (last revised June 2024). The authors also note that the effectiveness of strategies varies by domain, so the figure is not a promise for your business.

Google takes a simpler stance. In its guidance for site owners, Google says the best practices for SEO remain relevant for AI features in Search such as AI Overviews and AI Mode, and that no additional requirements or special optimizations are necessary to appear in them (Google Search Central, AI features).

The honest takeaway is that there is no secret formula. What you control is quality, clarity, and accessibility.

How do AI search tools decide which pages to cite?

For Google, a page must be indexed and eligible to be shown in Google Search with a snippet to appear as a supporting link in AI Overviews or AI Mode. Google also says these features may use a "query fan-out" technique, which means issuing several related searches across subtopics to build an answer.

In practice, one customer question can be split into several sub-questions. A page that answers each sub-question clearly, in its own section, has a better chance than a page that answers everything in one long paragraph.

ChatGPT search, Perplexity, and Claude each use their own crawlers for search. They offer controls through robots.txt, covered below.

What seven steps make a business more citable?

These steps are the author's practice, based on Google's guidance and the research above, not a proven recipe. Apply all of them to the pages you most want to be the reference.

1. Do not block search crawlers. Check robots.txt and your firewall, because a page that cannot be fetched cannot be cited.
2. Answer in the first paragraph. Open with a definition or answer in 40 to 60 words, then elaborate.
3. Write headings as questions people actually type, such as "How long does delivery take?"
4. Write short, self-contained paragraphs that still make sense if quoted alone.
5. Attach a source and a year to every number in the same sentence. A number without a source is hard to trust for people and machines alike.
6. Show the author and the updated date. Readers and engines both judge freshness and accountability.
7. Use tables, numbered lists, and an FAQ for information that is naturally a comparison, a sequence, or a question and answer.

Google sums up its guidance as "Who, How, and Why": who created the content, how it was created, and why (Google, helpful content). Answering all three on your page helps readers and engines.

How should you configure robots.txt for AI crawlers?

AI crawlers are generally separated by purpose, and each is controlled on its own. That lets you, for example, appear in AI search while declining to have your content used for model training. The table summarizes each provider's documentation as of November 7, 2026.

OpenAI says each setting is independent and that it can take about 24 hours after a robots.txt update for its systems to adjust (OpenAI, crawlers). Perplexity also says up to 24 hours (Perplexity, crawlers). Anthropic describes its three robots on its support page.

An example pattern for a site that wants to appear in AI search but stay out of model training (adapt it to your own policy):

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

User-agent: ClaudeBot
Disallow: /

Whether to allow training is a business decision. Do not paste this pattern without thinking through the consequences.

Do FAQ markup and structured data help you get cited?

Structured data helps machines understand a page's structure, but do not expect magic. For FAQPage markup, Google's documentation says the FAQ rich result is shown only for well-known, authoritative government and health websites (Google, FAQPage).

That means for most businesses an FAQ section is still worth having because it answers real reader questions in a compact form, not because it produces a special display in Google. Write FAQs for people, with answers that stand alone.

What does a quotable paragraph look like next to an unquotable one?

The difference is a clear subject, completeness, and brevity. Compare two answers to the question "What is WhatsApp opt-in?"

A habit that helps: cover everything except one paragraph of your article, read it alone, and ask whether it still makes sense to someone who has read nothing else.

What is a worked example on a single page?

Suppose a local courier company has a "Shipping rates" page that is one long paragraph with no section titles. This example is fictional and only illustrates the process. The fix can be done in one afternoon.

First, replace the opening with a direct answer: how rates are calculated, what affects them, and which cities are served. Second, split the page into sections titled as questions, such as "How are shipping rates calculated?" and "How long does delivery take?"

Third, put the rates in a table with the date they apply from. Fourth, add an FAQ with short standalone answers, plus an author name and an updated date at the top. Fifth, check robots.txt so search crawlers are not blocked.

The page may not be cited right away, but it is now easier for humans to understand and easier to cut into answers. You keep that benefit even if a particular engine has not cited you yet.

What checklist should a page pass before publishing?

Use this as a last gate for every important page. A "no" on any item is a reason to hold the page back.

1. Does the first paragraph answer the main question in 40 to 60 words?
2. Are most headings real questions?
3. Does every number have a source and a year in the same sentence?
4. Is there a table or numbered list where the content is a comparison or sequence?
5. Are the author, publish date, and updated date visible?
6. Can the page be crawled, and is it not blocked by robots.txt or a firewall?
7. Does it link to related pages and to authoritative outside sources?
8. Does it contain any claim you cannot prove? Remove it.

How can you check whether your business is being cited?

There is no single reliable tool for this yet, so the most honest method is a regular manual check. Write ten questions your prospects might type, ask them in ChatGPT, Perplexity, Claude, and Google, and note whether your site is cited and how.

Keep a simple table with columns for question, engine, date, cited or not, and the other sources cited. Repeat monthly with the same questions so you can see the trend. Generative answers vary over time, so read patterns rather than a single result.

Also check your server logs for visits from user agents such as OAI-SearchBot, PerplexityBot, and Claude-SearchBot. A crawler visit does not guarantee a citation, but its absence points to an access problem.

What mistakes should you avoid?

• Writing for machines instead of readers. Keyword-stuffed content lowers quality and trust.
• Inventing statistics or quotes to look authoritative. It destroys trust and can mislead readers.
• Burying the answer in the fifth paragraph after a long preamble.
• Blocking every AI crawler and then wondering why you never appear in AI answers.
• Treating one citation as a permanent win. Re-check every month.

How does Maira help?

Maira is not a search optimization tool, so we do not claim it makes AI cite you. What it does is relevant after the visit: if a reader finds you through search and then contacts you on WhatsApp, Maira's agent helps follow up with messages that ask permission first and records each consent.

Maira can also build a strategy from one sentence about your product, covering the ideal customer, market map, and outreach plan, which helps you decide which questions your content should answer. You can try Maira on the Free plan or read the product overview.

To keep going, read what is an AI sales agent and WhatsApp for B2B sales. Start with your single most important page, improve it, measure, and then repeat.