Generative engine optimization (GEO): how to get your business cited by AI assistants

Generative engine optimization (GEO) is the work of making your business easy for AI assistants to find, understand and cite. When someone asks ChatGPT, Claude, Perplexity, Gemini, Grok or Copilot "What's a good dentist near me that does Saturday appointments?" or "Which tool turns PDFs into a chatbot?", the assistant assembles an answer from sources it can access and trust. GEO is about being one of those sources, described accurately.

Why does generative engine optimization matter now?

More people ask AI assistants questions they used to type into a search engine, and many assistants now search the web and cite sources in their answers. That changes what "being visible" means:

• Answers, not lists. An assistant may name two or three options rather than showing ten links.
• Descriptions matter. The assistant summarises what you do, so you want that summary to be correct.
• Consistency matters. If your website, directory listings and reviews disagree about your services or prices, assistants may repeat the wrong version, or skip you.

The term "GEO" was popularised by a 2023 research paper, GEO: Generative Engine Optimization (Aggarwal and colleagues), which studied how content changes affect visibility in AI-generated answers. The practical lessons line up with good communication: be clear, specific, factual and well sourced.

How do AI assistants decide what to say about a business?

Broadly, from two places:

1. Training data: text the model learned from, up to a cut-off date. You can't edit this directly, but public, consistent information about your business over time helps.
2. Live retrieval: many assistants search the web or a partner index when answering, read the top results, and cite them. This is where GEO has the most immediate effect.

For live retrieval, an assistant needs to be allowed to crawl your pages, able to parse them (real text, not just images or scripts), and able to extract a clear, quotable answer.

What are the most effective GEO practices?

1. Write a canonical description of your business. One or two factual sentences: what you are, who you serve, where, and what makes you different. Use it consistently on your About page, homepage, directory listings and social profiles.

2. Answer questions directly. Use question-style headings ("How much does a cleaning cost?") and put the answer in the first sentence beneath them. Assistants quote concise answers.

3. Publish an FAQ page with real questions. Pull them from your inbox. Use <details> blocks or plain headings, and add FAQ structured data where appropriate.

4. Be specific and factual. Prices or price ranges, service areas, hours, policies, features and limitations. Vague marketing language gives assistants nothing to cite.

5. Keep facts consistent everywhere. Business name, address, phone, hours and services should match across your site, Google Business Profile, directories and social profiles.

6. Use structured data. Schema.org markup such as Organization, LocalBusiness, Product, FAQPage and BreadcrumbList helps machines read your pages unambiguously.

7. Let AI crawlers in. Check robots.txt. Many AI providers publish crawler user agents (for example OpenAI's GPTBot and OAI-SearchBot, Anthropic's ClaudeBot, PerplexityBot). Blocking them may keep you out of their answers.

8. Consider an llms.txt file. The llmstxt.org convention proposes a markdown file at /llms.txt summarising your site and linking key pages. It's cheap to add.

9. Be honest in comparisons. Fair, sourced comparison pages are more likely to be trusted, and quoted, than one-sided ones.

10. Earn mentions on trusted sites. Industry directories, local press, partner pages and genuine reviews give assistants independent sources that confirm what you say.

What should you avoid?

• Keyword stuffing and repetitive text: it reads badly to people and models alike.
• Fake reviews, invented statistics or awards: they can be checked, and they damage trust.
• Hiding key facts in images or PDFs only: put them in page text.
• Contradictory pages: old price lists and outdated service pages left online.

How does a website chatbot relate to GEO?

They're complementary, and they draw on the same work. The clear, question-and-answer content that helps AI assistants cite you is exactly what makes a knowledge base chatbot answer well. If you build an FAQ chatbot from your website FAQ, you'll naturally tighten that FAQ, and the chatbot's knowledge gaps report shows questions people ask that your content doesn't answer yet. Each gap you fill improves your chatbot, your FAQ page and your chances of being cited. See create a chatbot from your website FAQ.

How do you measure GEO?

There's no single dashboard yet. Practical steps:

1. List 10 to 20 questions a customer might ask an assistant, such as "best [service] in [city]" or "how do I [task your product solves]".
2. Ask them monthly in several assistants, and note whether you're mentioned, how you're described and which pages are cited.
3. Check referral traffic in your analytics from assistant domains.
4. Fix inaccuracies at the source: update your pages and listings where the wrong information comes from.

What does a GEO-friendly page look like?

A page that AI assistants can quote easily usually has:

• A clear H1 that names the topic.
• A two- or three-sentence summary at the top that directly answers the main question.
• Question-style subheadings with the answer in the first sentence below each.
• Specific facts: names, numbers, locations, dates, with sources where relevant.
• A short FAQ with real customer questions.
• A visible "last updated" date.

What does InstantKnowledgeBot do for its own GEO?

We follow the same advice: a consistent description of what InstantKnowledgeBot is, question-style explainers like this one, FAQ structured data, an llms.txt file, and comparison pages with sources and dates. Want to apply the same thinking to customer questions on your own site? Build a free knowledge base chatbot, or read what a knowledge base chatbot is. More articles on the blog.