Skip to content
Receipts, not vibes. See how your website actually scores against ours on our new website audit comparison tool!
The headline of a page, with the parts only search engines and AI read called outH1One per page, and it means it.Meta descriptionStructured dataLive sitemapllms.txt for AIWebMCP tools

How does a Loam website work behind the scenes?

June 3, 2026

Every Loam example site has an annotated tour of itself. Here is what it shows: the headline, the invisible search and AI markup, the security and the build.

Behind every page of a Loam website sit several layers, each written for a different reader. Visitors see the words, pictures and buttons. Search engines read a second layer of labels in the page code: a headline outline, a description, the page's one official address, structured data and a sitemap. AI assistants read those too, plus a plain-language guide to the site and, in browsers that support it, a set of tools they can call directly. Underneath it all, an editor and a design system keep every page on brand, and a security setup keeps out what does not need to be there.

You never type any of the hidden layers yourself; the site writes them from your content. But you should know they exist, because every one of them can be checked on any website, including the one you have now. This article goes layer by layer, explains what each does for your business, and tells you what to look at and what to ask.

What are the layers of a web page, and who reads each one?

Quick answer: people read what is on the screen, search engines read the labels in the code, and AI assistants read both plus files written especially for them. A good website serves all three from the same content.

The layers of a Loam page: what visitors, search engines and AI assistants each readVisitorsFast pages, real headings, keyboard and contrast checked,light and dark modeSearch enginesMeta descriptions, canonical addresses, a live sitemap andbreadcrumbsAI assistantsStructured data, FAQ markup, an llms.txt guide andWebMCP toolsNobody outsideIntegration keys, private fields and the content APIstay on the server

Think of it as one set of facts told three ways. An answer on your questions page appears as text for a visitor, as structured data for a search engine, and as FAQ markup an AI assistant can quote. When those three agree, every reader gets the same, correct answer. When the hidden layers are missing or out of date, a search result or an assistant may describe your business from guesswork.

Every Loam example site has a page called "How this site works" that points at each layer on the page itself. The example businesses are fictional, so nothing private is on show, but the markup the tour points at is real and you can inspect it.

What is an H1, and how do you check yours?

Quick answer: the H1 is the page's main headline in the code, not just in looks. There should be exactly one per page, and it should say what the page is about.

Search engines, AI answer engines and screen readers all use the H1 to decide what a page is about. A page with none leaves them guessing, and a page with five sends five different signals. Under the H1, H2s split the page into topics the way chapters split a book, and H3s sit beneath those. The order matters: a heading should never jump from H2 straight to H5 because the smaller text looked nicer.

On a Loam site, every layout sets its own heading levels, so the outline is correct on every page, including pages your team adds next year.

To check your own site: open a page, right-click and choose the option to view the page source, then search for "h1". Count them. Then look at the order of the other headings. If your site has a page builder, ask your web team whether its headings are real headings or styled text that only looks like one. Screen readers can only use the real kind.

What do search engines read that visitors never see?

Quick answer: six things: a meta description, a canonical URL, structured data, a share image, a live sitemap and breadcrumbs. On a Loam site they are generated from your content.

Here is what each one does, and how to check it:

  • A meta description is the two-line summary under your page's link in search results. It is not a ranking factor, but it decides a lot of clicks. On a Loam site, if an editor leaves it blank, the site writes one from the page's own content. Check yours by searching for your business and reading what appears under each link.
  • A canonical URL tells search engines which address is the real one, so the www and non-www versions of a page, and preview copies, never compete with each other. Ask your team whether every page declares one.
  • Structured data is machine-readable labels, written in a format called JSON-LD, that describe your business, your breadcrumbs, your questions and answers, your products and your articles. Search the page source for "ld+json" to see whether yours has any.
  • A share image is the picture that appears when someone sends your page in a text message or posts it on LinkedIn or Facebook. Text a link to yourself and see what shows up.
  • A live sitemap is an always-current list of every page for search engines. A Loam site updates it the moment something is published. Ask your team where yours lives and whether it updates on its own or only when someone remembers.
  • Breadcrumbs are the trail at the top of inner pages, marked up so search results show where a page sits on your site.

How does a website make itself understood by AI assistants?

Quick answer: with structured data and FAQ markup it can be quoted from, an llms.txt guide written for AI crawlers, and WebMCP, which lets AI agents ask the site for information directly. A Loam site also counts which pages AI crawlers visit.

People increasingly ask an assistant a question instead of searching. Making your content easy for those assistants to understand, trust and quote is sometimes called AEO, answer engine optimization. On a Loam site it comes from four pieces:

  • An llms.txt file, a plain-language guide to the site for AI crawlers, the way robots.txt is written for search crawlers. On a Loam site it lives at /llms.txt. Type that after your own domain and see whether anything is there.
  • FAQ markup. Questions and answers marked up as FAQ data, so an assistant can quote the answer directly. The "Why this website?" page on every example site publishes it.
  • WebMCP, an emerging web standard. In browsers that support it, AI agents can ask the site for its services, articles and contact details directly, rather than reading pages and guessing.
  • AI crawler counts. Visits from ChatGPT, Claude, Perplexity and other AI crawlers are counted page by page, so you can see which pages they read.

None of this guarantees an assistant will mention you; nobody honest can promise that. It removes the technical reasons it would not. The rest is what you publish, which is why we think every business needs a blog.

How does your team edit the site without breaking it?

Quick answer: each page is a stack of sections that open to plain fields: a heading, a paragraph, a button, a picture. Your team changes the words, previews the page and publishes, and the design stays put.

Editors can drag a section higher or add a new one from the same designed layouts. Team members, services, questions and articles are each kept as a single entry, so adding one updates every page that lists it. A change you have not published stays a draft. Nothing needs code.

What keeps it on brand is the design system. Colors, type, spacing, buttons and cards are designed in Figma and defined once as design tokens, named brand values every page reads. Nobody has to remember the exact shade of the buttons. Light and dark mode are both designed as part of that system, and every text and background pair is contrast-checked in both. Until a visitor picks one, the site follows their device. And when a business needs something the existing layouts do not cover, we design and build it rather than bending the business to fit.

To check your own site: ask whoever edits it to change a price while you watch. Note whether there was a draft and a preview, and whether the page still looked right afterward.

What keeps a Loam website secure?

Quick answer: mostly what is not there. No plugin marketplace, no public data API, no secrets in the page, no default passwords and no open door for spam, plus a person who reviews it every month.

Most website break-ins come through something that did not need to be there. So a Loam site leaves those things out:

  • No plugin marketplace. Every feature is built in and maintained together.
  • No public data API. Anonymous visitors get the pages, not a download of the site's content and structure.
  • No secrets in the page. Integration keys stay on the server, and data is scrubbed before it reaches the browser.
  • No default passwords. Admin accounts are never created with a known password, not even on day one.
  • No open door for spam. Forms carry a hidden honeypot field, a same-site check and a per-visitor rate limit.
  • No guesswork. A person reviews security and dependencies every month, and you get the report.

Questions to ask about your current site: how many add-ons does it depend on, and who updates them? Can someone who is not logged in pull your content out through an API? When did anyone last review it, and where is the report?

How do you check whether your site works for every visitor?

Quick answer: look at contrast, headings and labels, keyboard use, motion and phone speed. On a Loam site these are checked in the design system before a page exists, not added later.

Accessibility is not a plugin you add at the end. These are the six things a Loam site checks, and each one is something you can try:

  • Contrast. Every text and background color pair, in light and dark mode, is audited against WCAG contrast guidelines. Squint at your pale gray text on a white background.
  • Real headings and labels. Screen readers get a true outline and labeled form controls.
  • Motion you can stop. Anything that moves pauses on hover and respects the device's reduce-motion setting.
  • Keyboard friendly. Menus, dropdowns and accordions work without a mouse. Put your mouse aside and press Tab through your home page.
  • Fast on a phone. Images are sized for each screen in modern formats, and pages are cached close to visitors.
  • Lightweight by design. No slider libraries or plugin scripts; interactive pieces use the browser's own scrolling.

Google measures how fast a page loads, how quickly it responds and how much it jumps around, under the name Core Web Vitals. Ask your team how your pages do.

What is a Loam website built on, and what do you own?

Quick answer: established tools: Next.js, Payload CMS, MongoDB Atlas, Vercel, Vercel Blob and Figma. You own the domain and the database, and a care plan covers hosting, security updates and a monthly report.

In plain language, here is what each does for you:

  • Next.js prepares pages ahead of time and serves them instantly, which visitors and search engines both notice.
  • Payload CMS is the editor your team uses. It runs inside your own site, not on someone else's platform.
  • MongoDB Atlas holds your content, in an account you own, so it stays yours whoever looks after the site.
  • Vercel hosts the site on a global network, so pages load from a server near each visitor, and updates go live without downtime.
  • Vercel Blob stores your photos and documents, resized automatically for every screen.
  • Figma is where your brand is designed as a system of colors, type and styles.

The running costs are built so nothing is held hostage. Your domain is registered in your name with your own registrar. Your database is in your own account, so if you ever move on, your content goes with you. The Loam care plan covers hosting, security updates and the monthly report, and with a Loam retainer it is included at no extra cost. Without a care plan, the site runs on your own Vercel account, whose free plan usually covers a small business site, photo storage included.

Ask any web team: which accounts are in my name, and what would I need to take over if we parted ways?

How does a Loam website get built?

Quick answer: in five steps: discovery, design system, layouts, content, then launch and ongoing care.

How a Loam website gets built, in five steps1DiscoveryThe business, itscustomers and thejob of the site2DesignYour brand as adesign system,light and dark3LayoutsTested, accessiblelayouts, shaped tofit4ContentYour content,imported in onepass5CareLaunch, then amonthly review andreport
  1. Discovery. We learn the business, its customers and what the website has to do for both.
  2. Design system. Your brand becomes a design system in Figma: colors, type and styles as tokens, in light and dark.
  3. Layouts. Pages are composed from audited, accessible layouts, then customized wherever your business needs something new.
  4. Content. Content arrives from spreadsheets in one pass. Each example site was imported that way.
  5. Launch and care. After launch comes a monthly security and maintenance review, with a plain-language report.

Every step is spelled out, so you know where the build stands and what you are being asked for.

Where can you take the tour yourself?

Quick answer: open "How this site works" on any of our example sites. The businesses are fictional; everything the tour points at is real.

The fictional bakery's How this site works page is a good place to start. Its headline announces that it is an H1, and each feature in this article is labeled on the page that demonstrates it. Try the light and dark switch, open its /llms.txt, and view the page source with the checklist above beside you. Then run the same checks on your own site.

If you would like to see how your site measures up, share it on any example site's "Get a site like this" page for a free Website Showdown, or tell us about your business. We will read it, and reply as people, not a drip campaign.

Questions readers ask next

Yes. Paste a page address into a rich results test from a search engine and it lists the markup it finds and any errors. Run it after big content changes, not only at launch.

The submission is checked for spam, saved securely and sent to the people your team chose, by email or into the tools you already use. Nothing private is sent to services that do not need it.

Your colors are set up in pairs, one for light backgrounds and one for dark, so the brand stays recognizable and readable either way. The site follows each visitor's own device setting.

A person reviews the site's software every month, applies updates and checks that nothing broke, then sends you a plain report. You do not have to remember to log in and click update.

Very hard to do. Editors choose from designed layouts and fill in text and images, so spacing, fonts and colors stay consistent even when several people edit.