Be Visible
All Episodes
GEO and Beyond: Preparing Your Website for AI and Autonomous Agents

GEO and Beyond: Preparing Your Website for AI and Autonomous Agents

0:00|0:00

Hosts Ido Navarro and Ben Cohen break down Generative Engine Optimization (GEO) and what it means to build websites for AI search systems, crawlers, and autonomous agents.

They explore practical site infrastructure—including clean markdown delivery, schema, and emerging conventions like llms.txt—along with key steps to ensure machines can discover, understand, and interact with your brand.

This show was created with Jellypod, the AI Podcast Studio. Create your own podcast with Jellypod today.


Chapter 1

Be Visible to Machines, Not Just People

Ben cohen

A buyer asks an assistant which tool fits their team, gets a shortlist, a comparison, and a recommendation, and then books a demo with one company. Never opened a single website along the way. If that's how the decision gets made, where does your brand actually live?

Ido

It lives in whatever the machine understood about you. That's the whole episode. Welcome to Be Visible, I'm Ido Navarro.

Ben cohen

And I'm Ben Cohen. We're talking about how websites get discovered, understood, and used by AI systems, and what you can do about it this week.

Ido

Think of a modern site as having three audiences. The human who wants a polished experience. The machine that interprets, summarizes, and cites. And the agent that wants to do something, like fill a form or complete a purchase.

Ben cohen

But is this really different from SEO? Because a lot of people say it's just SEO with a new name.

Ido

Partly, and that's fair. Crawlability, clear content, and trust still matter. SEO was about ranking ten blue links through keywords and backlinks. GEO asks a different question: did a model find you, understand you correctly, trust you, and mention you? Being found isn't enough anymore. You have to be selected and summarized.

Ben cohen

And the research says these systems lean on third party sources too, not just your site. So you can't fully control the story.

Ido

Right, which is why your own content has to be the clear, unambiguous reference everyone else can point to. What you do, who you serve, why you're credible. Stated plainly, not buried in a slider.

Ben cohen

Let's get into infrastructure. Markdown first. I've looked at what sites look like once you strip the styling and scripts, and it's humbling.

Ido

Use the restaurant. Your website is the beautiful dining room, the lighting, the furniture. Clean markdown is the printed menu. A machine just wants to know what's on offer, not inspect the chandeliers. Navigation clutter, scripts, repeated interface bits, all of that is noise.

Ben cohen

So where does llms.txt come in? Does adding one suddenly make a site visible?

Ido

No, and let's be clear. It's an emerging convention, not a formal standard. It's a plain markdown file with a title, a short description, and links to your important pages, ideally the markdown versions. Think of it as a map. Its sibling, llms-full.txt, puts the complete content in one file.

Ben cohen

Which sounds great until the site is huge.

Ido

Exactly the limitation. A giant file can bloat or overflow what a model can take in. And adoption is uneven, so nobody should call it a ranking file. It's low risk, cheap, and part of a bigger readiness picture. Tools like LLMagnet for WordPress generate those files, serve markdown for each page, and log which AI bots visit. Useful, though you can build the basics by hand too.

Ben cohen

And schema?

Ido

Schema gives explicit context. Organization, product, service, FAQ. But it has to match what's visible on the page. Drift between the two creates confusion, and schema can't rescue weak content.

Ben cohen

Now the part that sounds futuristic. What happens when the visitor is an agent acting for a person?

Ido

Reporting suggests personal agents like Meta's Muse and OpenAI's dots reach sites through connectors or cloud browsers. That changes what robots.txt, on page SEO, and analytics can tell you. So a site an agent understands but can't use is incomplete. Clear forms, labels, predictable actions.

Ben cohen

So what can someone actually do this week?

Ido

Ask ChatGPT, Perplexity, and Claude the same question your customers ask. Compare the answers. Then fix the explicit basics: what you do, for whom, with evidence. Check crawler access, add accurate schema, consider llms.txt, publish original material, and test whether an agent can complete a key journey.

Ben cohen

And what not to do. Don't stuff keywords for AI. Don't mass produce low value pages. Don't block crawlers without knowing the consequences. And don't make the human experience worse.

Ido

Also don't count citations alone. Accuracy and business relevance matter.

Ben cohen

My prediction, and it's an opinion: within a few years, sites will routinely serve different representations to people and machines.

Ido

Mine: agent readiness becomes the next layer after mobile and accessibility. Something you simply expect.

Ben cohen

So the takeaway is that visibility isn't only about visits. It's about machines discovering, understanding, trusting, and eventually using your brand correctly.

Ido

Today, ask several AI systems one important customer question and compare what they say about you. If you enjoyed this, follow Be Visible.