Skip to content
Writing
·6 min read

Make your business readable to AI, not just to people

Drafted through my n8n + AI pipeline, edited by me.

By the end of this you'll understand why a buyer's AI assistant now reads your site before they do, and what makes your business legible to that reader.

The mess

A buyer asks an AI to shortlist three vendors and summarise each one before they click anything. If the assistant cannot cleanly parse what you do, it skips you or describes you wrong. You lose in a round you never saw, to competitors who simply read more clearly to a machine. This is already happening, quietly, in front of every page you publish.

The wrong way people solve it

They keep optimising only for search keywords and pretty visuals, writing for a reader that is no longer the only one. Some stuff the page with buzzwords and hope something sticks. The most common mistake is locking the facts that matter inside images and PDFs, where a model cannot read them at all, then wondering why the AI got their business wrong.

The system view

Treat the AI as a new kind of visitor with its own intake process. A buyer asks an assistant about you. It decides whether it can parse what you do, who it is for, and your proof. It either includes you in the shortlist or skips you. The buyer reads the summary, you watch how you are being described, and you keep structured data in place that the model can cite with confidence.

Trigger (a buyer asks an AI about you) → Decision (can it parse what you do, who for, and your proof?) → Action (include or skip you) → Human review (the buyer reads the summary) → Alert (watch how AI describes you) → Record (structured data it can cite).

What I would build

Plain text for the facts that matter, what you do, who it is for, and your proof, not buried inside graphics. Structured data that labels your pages so a model is not guessing. An llms.txt file, a simple plain-text map that tells AI tools what your site is and where the important parts live. And for the technical edge, exposing your data through a shared protocol so an assistant can actually use it, not just read about it. My own site carries this layer, and it took an afternoon.

What can break

Facts trapped in images, so the model never sees them. Claims that contradict each other across pages, which teaches a model to distrust you. No structured data, so it guesses and guesses wrong. And the privacy slip: exposing more than you meant to a tool that will happily repeat it. Readable is not the same as wide open. Decide what is public on purpose.

What the business gets

When a buyer's AI reads you, it comes away able to describe what you do and why you are credible, in your own words. You show up in the shortlist instead of being quietly skipped. And you get there early, while being early is still cheap.

Search was about being found. The next version is about being understood by the thing that now does the finding.

Bring me your site or the workflow behind it. I'll tell you what I'd make legible to AI first.

Building something this should run inside?

Book a systems call