# Your website is a blank page to ChatGPT

_People increasingly ask an assistant, not a search box, "where should I stay in Hội An?" If the assistant can't read your site, you're not in the answer. For a lot of sites, it can't — and the reason is mechanical, not mysterious._

`2026-06-30·building·agent-ready`

## The failure is one line in a log

The AI crawlers — GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot — fetch your HTML and read it as-is. They **do not run JavaScript**. So if your content is rendered client-side by a single-page app, what they receive is an empty shell: a `<div id="root"></div>` and a script tag. Beautiful in a browser, blank to an agent.

> A site can look perfect to every human visitor and be, to the models now recommending businesses, a completely empty page.

This isn't an SEO nuance to optimize later. It's a binary: readable or invisible. And it's invisible to exactly the channel that's quietly becoming the front door.

## What "agent-ready" actually means

Being readable is table stakes; being _understandable_ is the goal. The signals that move the needle are unglamorous and well-defined:

- **Server-rendered content** — the words exist in the HTML, no JS required.
- **robots.txt & sitemap.xml** — say who's welcome and what exists.
- **schema.org JSON-LD** — Hotel, Restaurant, LocalBusiness: structured facts a machine can trust.
- **llms.txt** — a plain-language index written for models.

None of it is exotic. Most of it is a weekend. The hard part isn't doing it — it's knowing whether you did, and proving it changed anything.

## A scanner with no black box

So the tool is deliberately transparent: a deterministic core fetches a site the way a crawler would, scores it against a rubric you can read line by line, names the specific gaps, and — the part that matters — re-scans after fixes to show the number went up. No "trust the AI"; the rubric is the contract.

```
agent-ready https://example.com        # scan & score
agent-ready ./dist --json              # machine-readable report
# → fix the flagged gaps → re-scan → prove the delta
```

## The uncomfortable, honest part

Here's what the research kept insisting, and I won't hide it to sell the tool: for a local hotel or restaurant, the single biggest lever often _isn't_ web tech at all. It's the Google Business Profile and the sheer count of Google reviews. The assistant frequently recommends the place with the most reviews, not the one with the cleanest markup.

So agent-ready is necessary, not sufficient. It gets you _legible_ to the machines. Being legible and being _chosen_ are different jobs — and a tool that pretends otherwise is just another black box with better manners.