---
title: "Help AI Help Your Users"
date: "2026-03-15T06:01:17+00:00"
url: "https://invis.net/help-ai-help-your-users/"
author: "invisnet"
license: "CC BY-ND 4.0"
license_url: "https://creativecommons.org/licenses/by-nd/4.0/"
site: "invis.net"
copyright: "Copyright 2024-2026 Charles Lecklider. All rights reserved."
disclaimer: "Personal website. Opinions are my own."
categories:
  - "Article"
tags:
---

# Help AI Help Your Users

Last month, [**48%** of all documentation visitors](https://x.com/handotdev/status/2013295444898783543) across sites powered by Mintlify were AI agents—not bots scraping for training data, but agents actively retrieving information on behalf of users.

This is *good news*, if you're ready for it.

Right now, those agents are working harder than they need to. They fetch your landing page, parse out the navigation, follow links to features, pricing, documentation—dozens of requests to piece together the information they're looking for. Every page arrives wrapped in scripts and styling that the machine must execute and discard. That costs you bandwidth and server load. It costs whoever's paying for the AI tokens—*real money*—every time.

The Consolidated Content specification addresses this using standard HTTP content negotiation—mechanisms that have existed since the 1990s—*with a twist*. The Accept header requests a machine-readable format; the Prefer header asks for consolidated content—related information brought together, structured for machine understanding. A single request returns clean, contextual information. No JavaScript execution, no parsing chrome, no reassembling fragments.

Publishers benefit from dramatically reduced request volume and lower infrastructure costs. AI systems get better information at lower cost, with a massively improved security posture from not executing arbitrary JavaScript on every site they visit.

The day after I published the spec, Dries Buytaert—founder of Drupal—[posted about adding markdown representations to his site](https://dri.es/the-third-audience) with link-based discovery. Within an hour he had hundreds of AI crawler requests. What Dries implemented—format conversion via the Accept header—is exactly where the spec suggests starting. The Prefer header goes further: not just serving markdown, but consolidating related information so machines don't have to infer what belongs together.

AI systems are already looking for these signals.

The specification is now an [IETF Internet-Draft](https://www.ietf.org/archive/id/draft-consolidated-content-00.html). No new protocols, no new infrastructure—just a way to help the machines help your users.


---
Copyright 2024-2026 Charles Lecklider. All rights reserved.
Personal website. Opinions are my own.