The search signals that cost nothing — and most expert sites still miss.

Why the boring layer matters

Machines don't read your site the way a person does. A person infers — from size, position, design — what matters on a page. A machine reads labels. The signals below are the labels. When they're right, every page you publish inherits legibility for free. When they're wrong, every page inherits the same handicap, and nothing downstream — content, schema, ads — performs to its potential.

None of this is advanced. That's the point. These are the signals the free tools check first, the ones AI engines lean on when deciding whether your page is usable source material, and the ones that are miscounted on a shocking share of expensive, professionally designed sites — because they're invisible in a browser and fail without an error message.

The seven signals

1. Exactly one H1 per page. The H1 is the page's one-sentence answer to "what is this?" Zero H1s means the page never answers. Two or more means it answers twice and splits the signal. Check: view source, count the <h1> tags.
2. Headings that step down in order. H2s under the H1, H3s under H2s — an outline, not a font-size picker. Machines reconstruct your page's logic from this hierarchy. Design tools that use heading tags for visual size produce pages whose outline reads as gibberish.
3. A title tag that says what the page is. The <title> is what shows in the browser tab, the search result, and the AI citation. "Home" is a wasted title. So is the company name alone on every page. Each page's title should say what that page is about, plainly.
4. A meta description that earns the click. One or two honest sentences on what the page covers. Machines often display it verbatim under your result — it's ad copy you either wrote or left to chance.
5. Link text that describes the destination. Machines read link labels as descriptions of the linked page. A site full of "learn more" and "click here" describes its own pages as nothing. "See what the Operating Map includes" beats "learn more" for machines, for screen-reader users, and honestly for everyone.
6. Alt text on every image that carries meaning. To a machine, an image without alt text is a blank space. If the image is decoration, an empty alt is correct. If it carries content — a chart, a pricing table, a headline baked into a graphic — the alt text is the only version the machine gets.
7. A canonical tag and a current sitemap.xml. The canonical tag tells machines which URL is the real one when a page is reachable several ways. The sitemap is the index of what you've published, so discovery doesn't depend on crawl luck. Both are one-time setups that quietly rot when nobody owns them — new pages that never enter the sitemap don't fully exist.

The afternoon pass

This is checklist work, not judgment work — which means it can actually get done. One pass, page by page: count the H1s, read the outline, rewrite the titles and descriptions, rename the "learn more" links, fill the alt text, confirm the canonical, regenerate the sitemap. A typical ten-page expert site takes an afternoon.

Two things this pass is not. It's not a growth strategy — these signals are the floor, not the ceiling; strangers come from a content layer answering their questions, and machines learn your business facts from schema. And it's not a one-time achievement — every future page can reintroduce every one of these mistakes, which is why the floor needs an owner, not a memory.

How to prove the pass worked

Don't grade your own homework by eye — the whole problem with these signals is that eyes can't see them. Three free checks, five minutes total: view source on your two most important pages and count the H1s yourself; run one page through Google's Rich Results Test, which reports structural problems alongside schema; and load yoursite.com/sitemap.xml to confirm it exists, parses, and contains the page you published most recently.

That last check is the one that catches the quiet failure. A sitemap that's missing your newest pages means the generation step broke at some point and nothing complained. It never does.

Why this is an operations story

Here's the pattern behind every one of these misses: the site looked perfect the day it launched, and no one was assigned to the parts a browser doesn't show. The H1s multiplied when someone new edited a template. The sitemap froze when the person who set it up rolled off. The alt text was "phase two."

Machine legibility isn't a project. It's a property the site either maintains or loses — the same way a site's truth decays as the business moves. The fix is structural: make the boring layer someone's explicit job, on a rhythm, with a check that runs whether or not anyone remembers.