---
title: "AI optimisation — Egenverk"
description: "Egenverk makes sites readable for AI assistants with llms.txt, Markdown and JSON-LD, and runs AI calls with the right model, batching and token logs."
canonical: https://egenverk.se/en/services/ai-optimisation
lang: en
updated: 2026-10-19
alternate: https://egenverk.se/tjanster/ai-optimering.md
---
# A site AI can read. AI calls you can count on.

We give your site llms.txt, Markdown versions and structured data so AI assistants find the right facts, and build AI calls with a fixed model per task, batch processing and logged tokens.

AI optimisation at Egenverk has two parts. First, your website becomes readable for AI assistants and AI search: llms.txt, Markdown for every page, JSON-LD and, for WordPress, an MCP server. Second, your AI calls become predictable: a fixed model and response format per task, batches for what can wait, tokens counted per call. We promise no placement in AI answers.

## What is included

- **llms.txt.** `/llms.txt` per language lists the site's pages with links to Markdown, and `/llms-full.txt` collects the pages' text in one file.
- **Markdown for every page.** Add `.md` to an address and an agent gets the page as plain text, without clicking.
- **Structured data.** JSON-LD with schema.org types for the organisation, products and articles on every page.
- **MCP for WordPress.** With Egenverk Sluice MCP, an AI client can read and, if you allow it, write in WordPress and WooCommerce. Pre-release.
- **The right model per task.** Each call type has a fixed model and a JSON schema for the response, set on the server. The client never picks the model.
- **Batches and token log.** Cases that can wait run through the provider's batch API, which costs less. Tokens are logged per call and counted against a monthly budget.

## How we have done it

- **egenverk.se.** Every page exists as Markdown, `/llms.txt` exists in Swedish and English, and the site has `/llms-full.txt`, Atom feeds and JSON-LD checked in the build.
- **Sluice MCP.** The MCP server runs at `/wp-json/sluice/v1/mcp` inside WordPress. No relay service: traffic goes only between the site and the client. ([Learn more](https://egenverk.se/en/plugins/sluice-mcp))
- **Egenverk Claims.** Model and response schema per call type in the server's configuration, batch processing through the provider's batch API and token counts per call. The service stores no content. ([Learn more](https://egenverk.se/en/plugins/customer-hub/claims))
- **Egenverk Babbla.** A daily token budget, a limit per user and an estimated cost per user and day. ([Learn more](https://egenverk.se/en/plugins/babbla))

## How we work

1. **Call.** Free. We listen and tell you honestly if we are the right fit.
2. **Written proposal.** Fixed scope, timeline and price.
3. **Small, reviewed steps.** Every change is reviewed and tested before it ships.
4. **Your code, your repo.** Ownership and licences are defined in the proposal. Documentation and a changelog are included.

## Common questions

### Will we rank higher in AI answers?

Nobody can promise that. We make it easy for AI assistants to read the right facts about you: plain text, an index and structured data.

### What is llms.txt?

A text file at the root of the site that lists the pages with links to their Markdown versions, for AI assistants that read the site.

### Do we need MCP?

Only if AI agents should be able to do things in your WordPress, not just read. Then Sluice MCP gives permissions, a log and human approval.

### How do you keep the cost of AI calls down?

With a fixed model per task, batch processing for what can wait and a token budget that stops the calls when it runs out.

### What does it cost?

Contact us for pricing.

## Can AI assistants read your site?

Send us the address. A short call, no sales script.

[Book a call](https://egenverk.se/en/contact)