Do I need to rebuild my site to be found by AI?

Not always. If the site is fast, well structured and easy to update, work on content, structured data and configuration is often enough. If it is old, slow or built on a closed platform, rebuilding on clean foundations usually costs less than patching it.

Is the SEO work I have already done wasted?

No, quite the opposite: almost all good SEO work, clear content, clean structure, a fast site, counts for AI assistants too. GEO adds a layer. It does not start again from scratch.

Traffic from AI is still small, so there is no need to panic. But behaviour is shifting, and preparing costs little precisely because few are doing it: arriving early here is a cheap advantage.

A website is not enough: Google and AI

Getting found today means two things at once: showing up in Google results and being legible to the AI assistants that answer in place of those results. The good news for a small business is that the two share most of the same work. You do not need two websites. You need one built properly.

What is happening, without the hype

Start with the figure that keeps this in proportion: traffic arriving at websites from AI is still around 1% of the total. Anyone telling you Google is dead is selling something.

Inside Google, though, things really are changing. The Pew Research Center measured that when an AI generated summary appears at the top of the results, users click through on 8% of visits, against 15% when no summary is shown: close to half the clicks disappear. A randomised field experiment run in 2026 by researchers at ISB and Carnegie Mellon measured a 38% drop in organic clicks.

And the people who do arrive from an AI answer are worth more: according to Adobe Digital Insights (March 2026), AI referred retail traffic converted 42% better than non AI traffic. Fewer visits, but decided ones. Someone who already has the answer arrives ready to act rather than to browse.

The honest summary is this: the shift is early, but it is real. Answers are replacing some of the clicks, and the clicks that remain carry more weight. Preparing now is cheap. Catching up later is not.

Classic SEO: what still holds

Google is still the main door, particularly for local businesses. The SEO work that matters has not changed: a fast site, a clear structure, careful titles and descriptions, pages that answer what customers actually search for, a correct presence on Google Maps. Neglecting that to chase AI is a way of losing twice.

GEO: what it adds

GEO, generative engine optimisation, is the set of practices that make a site legible and quotable to generative engines. Concretely:

Answer-first writing: every page opens with the answer, in complete sentences an assistant can lift as they are.
Structured data: the business, its services and its content described in a standard format machines read without ambiguity.
llms.txt: a file summarising the site for AI engines. I have written a practical guide to it.
AI crawlers allowed: many sites block GPTBot or ClaudeBot without knowing it. A well configured robots.txt opens the door explicitly.

Where SEO and GEO overlap

Almost everywhere, which is the best part of this. A fast site pleases Google and AI crawlers. Clean semantic structure serves both. A page that answers a question well works in search results and in generated answers. Even structured data, invented for SEO, has become the map assistants use to work out who you are.

For a small business that means one thing: you do not have to choose between Google and AI. You have to do one job well, and it serves both.

Where they genuinely differ

There are three real differences, and knowing them is how you avoid being sold nonsense:

The unit of measure. SEO thinks in pages climbing rankings. GEO thinks in sentences getting picked up. An assistant does not deliver a visitor to your page, it extracts a piece of it. Every sentence that matters has to stand on its own, out of context.
How you measure. Google has mature tools like Search Console. For assistants there is no official dashboard telling you when and how you were cited. You measure by querying the AI systematically, which is why I built VisibAI.
The pace. Search rules change slowly and are documented. Assistant behaviour changes quickly and without notice. The best defence is not to chase tricks: clear, readable facts survive every update.

What to do first, in order

Fix the foundation. A fast site, readable on a phone, open to crawlers. Without this, nothing else starts.
Write pages that answer. Every service gets its page, every page opens with the facts: what you do, where, for whom.
Add structured data. Invisible to visitors, decisive for machines.
Publish llms.txt and open up robots.txt. Minutes of work, and hardly anyone has done it.
Then, and only then, content. Articles answering your customers' real questions, added over time onto a technical base that makes them count.

There is one more reason not to put this off: across the 1,344 audits run with VisibAI between February and August 2026, the average AI visibility score was 39 out of 100, on a self selected sample skewed towards B2B. With scores that low, in many sectors whoever gets organised now still finds very little competition.

That path, from technical foundation to content, is exactly what I do for my clients: a site built once, working on both fronts.

A website is not enough: getting found on Google and on AI