I rebuilt my website to be read by AI, not just Google

A year ago I rebuilt my own website around a question most sites still aren't asking: what happens when your reader isn't a person scrolling Google, but an AI writing the answer for them? I've spent seventeen years watching marketing channels rise and die, search, social, email, each in turn, so I recognize the pattern when a new one arrives. This is one.

Why rebuild a website to be read by AI, not Google?

The numbers are stark. Something like 80%¹ of Google searches now end without a single click. AI Overviews have cut clicks to the² top result by roughly a third. ChatGPT alone fields hundreds of millions³ of questions a week. People increasingly get the answer, not the list of links, which means "rank #1" is quietly becoming the wrong goal. In seventeen years of buying and earning attention, I've learned to spot a channel whose economics are inverting. This is that moment for search.

Optimizing to be the source, not the result

So I built for the machine that answers. Concretely, on my own site that meant: JSON-LD structured data describing who I am, what I've done, and every article I've published, in a form models can parse; an llms.txt file that greets AI crawlers and tells them what matters; a robots file that explicitly welcomes GPTBot, ClaudeBot and PerplexityBot instead of blocking them; and prose written to be quotable, clear claims, one idea at a time, easy to lift into an answer. The whole site now runs through the same discipline I'd sell a client.

The dual-market angle nobody's pricing in

Here's where it gets interesting for anyone outside the US. English content dominates what these models learned, so being the clear, well-structured source in an underserved space, a market, a language, a niche the giants ignore, is wide-open ground. It's the same opening I wrote about with Bangla-first AI: the model's blind spots are a marketer's opportunity. The move that helps a personal site get cited is exactly the move that lets a Bangladeshi brand become the answer in a category no one has claimed yet.

The specifics are the tell

None of this games anything, and that's the point. The pages that get cited are the ones that make a checkable, specific claim a model can trust and reuse. "We help businesses grow" gets ignored; "launched the Galaxy Note9 to 1.54× year-on-year growth, No. 1 globally in its category" gets quoted, because it's concrete and attributable. Answer engines reward the same thing skeptical humans do: specificity backed by a name.

This isn't a hack: it's a discipline

Be clear, be structured, be genuinely useful, and make it easy to quote you. That worked for search engines. It works even harder for answer engines, because a model has to *choose* your sentence over everyone else's. The brands and people who win the next five years will be the ones who understood, early, that they were writing for a reader who never scrolls.

The short version

• ~80% of searches now end with no click; AI answers instead of listing links.
• Optimize to be cited: structured data, llms.txt, crawler-friendly, quotable, specific prose.
• Underserved markets and languages are open ground to become the answer first.
• Specificity backed by a name is what both AI and skeptical humans actually trust.

When an AI answers a question in your field, whose words is it using, and could they be yours?

Md Shafaat Ali Choyon (MPH, CHES®, MBA, MCIM) is a growth, marketing and public-health strategist who builds and runs AI in production, with 17 years across telecom, fintech, e-commerce, consumer tech and healthcare in the US and Bangladesh. See the essays or the portfolio.