LLM Optimization for Websites: Cut What Models Read
Services page, bytes an agent reads 89.6%
Measured demo service page: 13,581 → 1,412 bytes
01 What optimising for an LLM actually means
Agents can retrieve pages through direct requests, browsers and search systems. Support the relevant paths with clear content and compact source-backed responses.
| Property | Why it helps | How to check it |
|---|---|---|
| Fetchable | A blocked request cannot be optimised | Request the URL as an agent would |
| Parseable without JS | Direct clients may not render JavaScript | Read the raw response, not the DOM |
| Small | Context is finite and shared | Compare bytes with and without negotiation |
| Front-loaded | Truncation and summarisation cut the tail | Find the offset of the first real fact |
02 What subtraction is worth
The demo page captures show two page types: a 91,226-byte product page becomes 1,315 bytes, and a 13,581-byte services page becomes 1,412 bytes. Each retains its source buying facts; compare your own pages separately.
These captures show different page types. Compare your own before-and-after responses to measure the improvement for your site.
03 How the subtraction happens
Content negotiation, done at the edge. A request asking for markdown receives markdown assembled from values you already store; browsers and search crawlers receive exactly what they received before.
Code Example
import { NextResponse } from 'next/server'
import { createGatewayProxy } from '@rebilder/gateway/next'
import { gatewayConfig } from '@/lib/gateway-config'
// Configure storeId and source resolvers in gatewayConfig.
export default createGatewayProxy(gatewayConfig, () => NextResponse.next())
Frequently asked questions
What is LLM optimization?
In a website context, making pages that language models can retrieve and afford to read: reachable without a blocked fetch, parseable without JavaScript, small in tokens, and front-loaded with the facts. In a machine-learning context the same phrase means making models cheaper to run, which is unrelated.
How do I optimize my website for LLMs?
Check access and source completeness, make key facts easy to find and test both direct and browser responses. Add a compact Markdown representation where it serves the clients you support.
How many tokens should a page be?
Use enough content to answer the question clearly. Compare completeness and response size on your own pages instead of targeting one universal token count.
Does this require changing my CMS?
Choose an adapter for your platform and connect the relevant business sources. The core SDK is framework-independent and free to self-host.