How many customers asked ChatGPT — and never saw your site?

AI search optimization — often called GEO, generative engine optimization — is the practice of structuring a website so AI assistants like ChatGPT and Perplexity can read it, quote it, and cite it as a source. It differs from classic SEO because answers cite passages, not ranked pages. Here is what actually works.

The Yani team ·

Start building free

What changes when people ask AI instead of Google?

Google gives a ranked list of links and the visitor chooses. An AI assistant gives one synthesized answer and chooses for them — pulling a handful of passages from a handful of sources and citing those. Ranking fourth on Google still earned clicks; being the fourth-best source for an AI answer earns nothing, because the answer only quotes two or three.

That shrinks the unit of competition from the page to the paragraph. The engines are not asking "which page is best for this query?" — they are asking "which few sentences answer this question cleanly enough to lift?" A page full of good information written in a way no single passage can be quoted is, to an AI engine, a page with nothing to cite.

What should you change on your site?

Five changes, all doable this week, all worth doing regardless of AI — they make pages better for humans too.

Answer first, explain second

Open every section with a sentence that stands alone as the answer — definition, number, or verdict — then elaborate. Engines lift self-contained passages; a conclusion buried in paragraph four does not get cited.

Headings phrased as real questions

People ask assistants in full questions — "can I export the code?", "how much does it cost?". A heading that matches the question, followed directly by its answer, is the easiest passage for an engine to match and quote.

FAQ blocks with FAQPage schema

A question-and-answer block is pre-chunked citation material, and FAQPage structured data labels it explicitly. Make the first sentence of each answer the answer — not a warm-up.

Comparison tables for hard facts

Prices, plans, feature differences, spec lists: engines lift structured facts far more reliably than facts scattered through prose. If a fact matters, give it a row in a table.

An llms.txt map of your site

llms.txt is an emerging convention: a plain-text file at your site root that tells AI systems what your site is and where the important pages live. It costs minutes to add and makes your best pages easier to find and attribute.

Should you let AI crawlers onto your site?

If you want to appear in AI answers, yes — a robots.txt that blocks GPTBot, ClaudeBot, or PerplexityBot removes you from those answers entirely. It is a real trade: your content helps train and feed systems you do not control, in exchange for citations that are becoming a meaningful referral channel. For most businesses trying to be found, being quotable beats being invisible.

One technical detail decides everything else: most AI crawlers do not run JavaScript. If your site only renders content in the browser, an AI crawler sees a nearly empty page — every answer-first paragraph and FAQ block you wrote might as well not exist. Whatever matters must be present in the static HTML the crawler first receives.

How do sites built with Yani apply this?

Sites Yani builds are made to be read by both kinds of searcher: dual-rendering keeps pages crawlable by Google and by AI engines like ChatGPT, so the content exists where crawlers actually look. And because Yani builds whatever you describe, the rest of this checklist is one instruction away — ask for FAQ sections with schema, comparison tables, an llms.txt, and answer-first page copy, and that is what gets built.

If you have not built the site yet, start with the full walkthrough at /blog/how-to-build-a-website-with-an-ai-agent, or see what a Yani-built web app includes at /features/webapp.

Frequently asked questions

Is GEO different from SEO, or the same work renamed?

It overlaps but is not the same. Classic SEO earns a rank on a results page; GEO earns a citation inside a generated answer. Good content helps both, but GEO rewards quotable passages, explicit question-answer structure, and machine-readable facts more than link-building and keyword placement.

Does llms.txt actually matter yet?

It is an emerging convention, not a guarantee — adoption among AI systems is growing but uneven. The honest case for it: it costs minutes, cannot hurt, and gives any system that does read it a clean map of your site. Treat it as cheap insurance, not a ranking lever.

Will allowing AI crawlers cannibalize my traffic?

Some answers will satisfy visitors who would have clicked through — that is the trade. But blocking crawlers does not bring those visitors back; it removes you from the answer while a competitor gets cited. If being found drives your business, citations are the channel you cannot opt out of.

Do AI crawlers run JavaScript?

Mostly no. GPTBot, ClaudeBot, and PerplexityBot largely read the raw HTML response and skip client-side rendering. Content that only appears after JavaScript runs is invisible to them — which is why server-rendered or statically generated pages are the foundation of AI visibility.

Build a site AI engines can quote

Describe your business to Yani and get a site that is readable by customers, Google, and the AI answering their questions.

Start building free