AI SEO, GEO, and AEO

More people now ask an assistant before they search. ChatGPT, Perplexity, Gemini, Claude, Microsoft Copilot, and Google's AI Overviews and AI Mode answer the question directly and cite a handful of sources. I find out what they say about you, where they're getting it, and what would make them say it more accurately and more often.

The vocabulary, briefly

The field has too many acronyms for one idea. GEO (generative engine optimization), AEO (answer engine optimization), AIO, LLMO, and AI SEO all describe making a site easy for AI systems to find, understand, trust, and cite. In practice it's good technical SEO, clear entities, and content written so a single paragraph can stand on its own as an answer.

What I test

• A prompt set built from your buyers' questions. "Best X for Y," "Is Z legitimate," "Who does A near B," and comparisons with named competitors, run across the major assistants.
• What each answer says. Whether you appear, how you're described, and whether the facts are current.
• Where the answer came from. The sources each assistant cites, and which of them you control, can influence, or should correct.
• Crawler access. Whether GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended, and Bingbot can reach your pages, through robots rules, Cloudflare, and your firewall.
• Machine readability. Whether your key content is in the server HTML, since most AI crawlers don't run JavaScript.

What I change

Entities

AI systems reason about things, not keywords: your organization, your people, your products, and your locations. I make sure each is described the same way everywhere, with Organization, Person, Product, and LocalBusiness schema tying them together and sameAs links pointing to the profiles that confirm them.

Answer-shaped content

Pages that open with a direct answer, define their terms, include the numbers people ask about, and keep each section self-contained. FAQs written around real questions. Glossaries. Comparison pages that are fair enough to quote.

Evidence elsewhere

Assistants trust what other sources confirm. I identify the directories, review sites, publications, and community discussions that answers lean on, and where your information there is missing or out of date.

Files for machines

An llms.txt file that summarizes your site and links to clean Markdown versions of key pages, XML sitemaps, and IndexNow, so changes reach Bing, and through it Copilot, quickly.

Measuring it

AI answers vary from run to run, so I measure with repeated prompts over time rather than one screenshot. Alongside that: AI assistant referrals in GA4, citations of your pages, and the share of tested prompts where you appear and are described correctly. Results are reported as trends, with the prompt set kept stable so the comparison is fair.

What to expect

Technical access problems can be fixed in days. Changes in how assistants describe you follow their crawl and retrieval cycles, which take weeks to months. Start with my AI search readiness checklist, or see a company that wanted accurate AI answers for job seekers.

Updated October 6, 2026