Last time we said the goal is shifting from being found to being quoted — becoming the source an answer engine pulls from when it replies to someone. So how do you become that source? You make your site genuinely easy for a machine to read. The good news, and the quiet joke of all this, is that it is mostly the same work that was always good practice.
why being readable matters now
If AI is summarising the web, you want to be inside the summary — cited, linked, recommended. The new "rank number one" is "the answer mentions you, by name, with a link". You cannot buy your way into that. You earn it by being clear, consistent and genuinely useful — to people first, and to the machines reading on their behalf. Treat the machine as a slightly literal reader who never guesses: if something is implied but not stated, it doesn't exist.
the basics that still win
- Clean, semantic HTML. Headings that mean something, real text instead of words baked into an image, buttons that are actually buttons. If a machine can't parse the structure, it can't trust the content.
- Structured data (schema.org). Spell out what you are in a language machines read: a studio, a product, an article, a price, an FAQ. This is how you stop being "some text" and start being "a business in Madrid that does X".
- Speed and crawlability. If a bot can't load the page, it can't quote it — which is one more reason your site should be fast.
- Plain, factual copy near the top. Say what you do, and for whom, in normal words before the clever stuff. The first paragraph does most of the work.
the newer bits
llms.txt— a small file that points AI models at your most useful pages. It is early days, costs almost nothing, and there is no real reason not to.- Consistent facts everywhere — same name, services and location across the whole site, so you are easy to summarise without contradictions.
- Clean meta and Open Graph tags, so a link to you looks legitimate when it is shared or cited.
a ten-minute audit you can do today
You don't need a developer to spot the worst problems. Open your homepage and try four things. First, hit Ctrl-F and search for your main service — if the words a customer would use aren't on the page as real text, that is a gap. Second, select your headline and key facts with your mouse; if you can't highlight them, they are probably images, and invisible to a machine. Third, paste your URL into ChatGPT and ask it "what does this business do and who is it for?" — its answer is roughly what every other AI sees. Fourth, check your prices, services and contact details say the same thing on every page. Most sites fail at least one of these, and all four are fixable.
what quietly works against you
- Text trapped inside images, so the only readable version is a caption that isn't there.
- Walls of vague, keyword-stuffed copy that say nothing a machine can repeat with confidence.
- Key facts hidden behind a script or an interaction a crawler never triggers.
what you can't control (and shouldn't chase)
Be realistic: you cannot make an AI cite you, any more than you could force Google to rank you. The model decides, its rules change, and no one can promise you a spot in the answer. What you can do is stack the odds — be clear, fast, consistent and genuinely useful — so that when the machine goes looking for a trustworthy source on your topic, you are the obvious pick. Anyone selling you a guaranteed "AI ranking" is selling you snake oil. The honest play is to be the best-prepared source and let the quality speak.
the mindset
Write for a smart human in a hurry, and the machines follow. That is the heart of it: the fundamentals did not change, the audience just added a robot. Do the boring things well — clear structure, honest facts, fast pages — and you are readable by both.
This is the stuff we bake into every site we build, not as an add-on but as the baseline. 👋
