Jul 12, 20264 min read

If you publish it, make it readable by AI

We put our work online to be found. The audience now includes LLMs and agents — so making your content machine-readable is just the modern version of being accessible.

Why do we put anything online at all? To be found. A portfolio, a blog post, a project write-up — we publish it so that someone, somewhere, looking for exactly this, can actually reach it. Accessibility has always been the whole point.

For most of the web's life, the someone doing the looking was a person with a search bar. So we learned to be legible to search engines — titles, meta tags, sitemaps, the whole ritual of SEO. Not because we loved Google, but because that is where the audience was.

The audience changed

The audience is shifting again. More and more, the thing reading your work is an LLM, or an agent acting on someone's behalf — answering a question, comparing options, deciding whether you are the person or the project worth surfacing. If your content is not readable to them, you are quietly invisible to a growing slice of the people you were trying to reach.

We are moving into an era where being accessible means being readable by machines, not just people.

Describe yourself, or be guessed at

Here is the part I care about most: if you do not describe your own work for machines, they will describe it for you. They will scrape the HTML, strip the navigation, drop the nuance, ignore the images, and reconstruct a version of you from whatever survives. That version is out of your hands.

Making your content AI-readable is not giving it away. It is the opposite. It means you get to decide what the machine sees, how your work is summarized, which points matter, what every image actually shows. You are taking back the data layer of your own presence online.

So this blog does it on purpose

That is why this blog publishes its own machine-readable layer at /blog/llms.txt — a clean, self-authored description of every post, written in my words, with a text description of each image. Any model or agent can read it and understand what I meant, instead of what a parser could salvage. Same source of truth as the page you are reading; I just made sure the machines get an honest copy too.

The plumbing that makes it work

None of this is magic — it is a handful of small, boring files that have quietly become the interface between your site and the machines reading it. Four worth knowing:

robots.txt
who may crawl, and where the sitemap lives
sitemap.xml
every page worth indexing, as one machine-readable list
JSON-LD
structured facts (who you are, when a post was published) as Schema.org data
llms.txt
a self-authored, plain-text summary written for LLMs

The first three are SEO-era plumbing, repurposed. robots.txt and sitemap.xml have guided crawlers for twenty years. JSON-LD (Schema.org) is how a machine reads 'this is a Person, this is their job, these are their links' without parsing your markup at all. llms.txt is the new one — a convention for handing a model a clean summary instead of making it reverse-engineer your HTML.

This site ships all four. robots.txt points to the sitemap, the sitemap lists every page and post, each page carries JSON-LD, and there is an llms.txt at the root plus a detailed one for the blog. Put together, an agent can land here and get an honest, structured answer to 'who is this and what have they built' — no scraping, no guessing.

In short

  • —The whole point of publishing online is to be accessible to whoever is looking — and increasingly, who is looking is an LLM or an agent acting for a human.
  • —For a decade we optimized content to be found by search engines (SEO); the same shift is now happening for AI, and it is worth getting ahead of.
  • —If you do not describe your own work for machines, they will guess — scraping HTML, dropping context, ignoring images. You lose control of how you are represented.
  • —This blog publishes its own machine-readable layer at /blog/llms.txt: a self-authored description of every post, including text for every image.
  • —The practical toolkit is four small files: robots.txt (crawl rules + sitemap pointer), sitemap.xml (every indexable page), JSON-LD/Schema.org (structured facts like identity and publish dates), and llms.txt (a plain-text summary written for LLMs). This site ships all four.
  • —Making content AI-readable is not surrendering it — it is the opposite. You decide what the machine sees and how you are summarized.