SEO for a marketing site: what works and what is myth
A small site has few levers, and all of them are cheap. What to do once, and what not to do at all.

A ten-page marketing site has no SEO strategy, and that is good news: nearly everything that works is done once and then left alone.
What matters
**A title and a description per route, all different.** The same `<title>` on five pages is five pages of which search will show one. The description does not move the ranking, but it moves whether the line is clicked, which comes to the same thing.
**One `h1` per page, about what the page is about.** Not "Welcome", not the company name.
**A canonical and language versions.** The same page reachable at three addresses competes with itself. Language versions not tied by `hreflang` get shown to the wrong people.
**First-screen speed.** Not "optimisation" but specifically: how long somebody on a phone waits.
**A sitemap with everything in it.** Especially what appeared recently: a map edited by hand after each publication eventually loses the latest article.
What does not
Keyword density. Meta keywords — nothing has read them for fifteen years. Copy written for a robot: a language model reads it now, and it can see it was not written for a person. The endless block at the bottom of the page listing every city you serve does not lift the page, it sinks it.
And above all: numbers you will not keep updating. "62 sections" in the description of a site that has 66 is not an SEO problem, it is a promise that no longer matches the product. We caught exactly that on our own site last week.
What the template does about it
Meta comes from one place per route and language, and so do canonicals and `hreflang`. The sitemap walks static routes and the articles on disk, so a new article enters it by existing. Before publishing, a command walks every route and **refuses to draw the social cards** if it finds a duplicate title, a missing description or a broken language set.
None of this puts a site at the top. It removes the whole class of mistakes usually found six months later, by way of traffic that never came.
More from the blog
Two, never more. A longer strip turns the end of an article into a second index page, and the reader has already chosen once.

