How to do
generative engine optimization, AEO & AIO.

Generative engine optimization starts from questions, because an AI answer is written per question. List the 20 to 50 prompts a buyer types while choosing in your category: “best [category] for [use case]”, “[you] vs [competitor]”, “how to [solve the problem] in [industry]”, “is [product] worth it”. Take them from sales calls, support tickets and your own Search Console, where long, question-shaped queries are the ones people now paste into assistants.

This list becomes your prompt panel. Every later step is judged against it.

Run the panel in ChatGPT, Gemini, Claude, Perplexity and Copilot, and read the AI Overviews on Google. For every answer, record whether you are named, whether you are linked, how you are described, and which sources the answer cites instead. Run each prompt more than once, because answers vary from run to run.

The cited sources are your map. If the answers lean on a review site, a trade publication, a subreddit or one competitor’s comparison page, that is where the work goes. Prompts where the answer is vague, outdated or cites nobody in particular are your fastest openings.

An answer can only cite a page its engine can fetch. Check that robots.txt lets in the AI crawlers you care about (GPTBot, ClaudeBot, PerplexityBot, and Google-Extended for Gemini), that your key content renders without waiting on client-side JavaScript, and that pages load fast.

Clean HTML and structured data help machines parse a page. Google is explicit that its AI features need nothing extra: “Structured data isn’t required for generative AI search, and there’s no special schema.org markup you need to add.” An llms.txt file is optional documentation for the AI tools that read it; Google Search ignores it.

Answer engines lift the cleanest block on the page, and generative engines quote sources that state things plainly. Give each URL one job. Put the answer in the first lines, then the steps, the numbers, the conditions and the exceptions. Use a table for comparisons, a definition box for terms, and an FAQ written from the real questions in your panel.

Give every fact a source or a date. A page that says exactly what it knows, and where it learned it, is a safer quote for a model than a page of adjectives.

The highest-leverage asset in generative engine optimization is information that exists only on your site: your benchmarks, pricing data, product usage statistics, a validated method, a calculator. It earns editorial links and gives the model a reason to cite you over ten pages that repeat each other.

For a startup this usually exists already, inside the product. The work is getting it out of the database and onto a page with a method note.

Assistants that search before they answer draw on search indexes, so ordinary rankings still decide much of what gets cited. Google says its generative AI features are “rooted in our core Search ranking and quality systems.” Cover Bing too: it powers Copilot, and Bing Webmaster Tools reports your AI citations for free.

Then earn mentions where models read about your category: trade publications, analyst and comparison pages, podcasts, and communities such as Reddit where your buyers talk. Earn them honestly. Google’s guidance warns against chasing inauthentic mentions, and bought placements get discounted.

Re-run the prompt panel every month, track citations in Bing’s AI performance report and AI Overview citations in a rank tracker, and compare against a period when nothing changed. Double down on the page types that get cited. Rewrite or merge the ones that never do.