Generative engine optimization (GEO): how to get your site cited by ChatGPT, Perplexity, and Google AI Overviews

Search is no longer only ten blue links. ChatGPT, Perplexity, Gemini, Copilot, Claude, and Google's AI Overviews now answer many questions directly, and they cite a small number of sources when they do. Generative engine optimization (GEO) is about being one of those sources.

The term comes from a 2023 research paper by Aggarwal et al. at Princeton and collaborators, which tested content changes against generative search and found that adding citations, quotations, and statistics could increase a source's visibility in AI answers by up to around 40%. Keyword stuffing, by contrast, did not help.

GEO vs SEO: what actually changes

The important point: GEO builds on SEO, it doesn't replace it. AI search products retrieve pages from search indexes before they summarise them. If your site is slow, blocked, or poorly indexed, no amount of rewriting will get it cited.

The technical checklist

1. Allow AI crawlers. Check robots.txt for rules that block GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot, ClaudeBot, or Google-Extended. Decide deliberately which you allow.
2. Publish a sitemap at /sitemap.xml with accurate last-modified dates, and reference it from robots.txt.
3. Get indexed by Bing. Several AI assistants rely on Bing's index. Submit your sitemap in Bing Webmaster Tools, not only Google Search Console.
4. Add JSON-LD structured data. At minimum: Organization, WebSite, Article on posts, FAQPage where you have real FAQs, and BreadcrumbList.
5. Render content on the server. Many crawlers do not run JavaScript. Static or server-rendered HTML is the safest option.
6. Make pages fast. Core Web Vitals still matter for the retrieval step, and slow pages get crawled less.
7. Consider an llms.txt file. It is a proposed convention that gives language models a curated map of your site. Adoption by AI providers is still uneven, so treat it as low-cost insurance, not a ranking lever.

The content checklist

• Lead with the answer. Put a direct, quotable answer in the first two or three sentences, then expand.
• One question per section. Use headings that match the way people ask — "How much does X cost?", "X vs Y".
• Be specific. Replace "significantly faster" with the actual number, and say where it came from.
• Cite sources. Link to primary research, official documentation, and standards.
• Use lists and tables for steps and comparisons — they are easy for models to extract accurately.
• Show who wrote it and when. A named author with real experience, a published date, and an updated date.
• Keep it fresh. Revisit high-value pages regularly and update the date only when the content actually changes.

What this looks like on a real site

When we rebuilt the CombineHealth frontend, the site had PageSpeed scores around 20–30% and no JSON-LD schema at all. We rebuilt the missing sections in Next.js, took PageSpeed to 100% across all pages, added comprehensive structured data, configured robots.txt and dynamic sitemaps, added an llm.txt file, and set up both Google Search Console and Bing Webmaster Tools.

At a larger scale, internal linking becomes the bottleneck. For a 400+ page site we built an AI agent that maps and recommends internal links whenever a new page is published — a job that is impossible to do well by hand.

How to measure GEO

1. List 20–30 questions your buyers actually ask.
2. Run them through ChatGPT, Perplexity, Gemini, Copilot, and Claude. Record whether you are cited, and who is cited instead.
3. Track AI referrals in analytics (referrers such as chatgpt.com and perplexity.ai).
4. Repeat monthly after each round of changes.

If you want an audit of your site's SEO and AI-search readiness, get in touch. We will send back a prioritised list of fixes.

Frequently asked questions