SEO: robots, sitemap, canonical, link previews and structured data #48

New Issue

2026-09-02T01:19:06Z

paulhitt commented

2026-09-02 01:19:06 +00:00

(Migrated from gitlab.com)

Ask: make the public pages findable.

What is actually there today (audited 2026-09-01)

layout.leaf has charset, viewport, <title>, description, and three Open Graph
tags — og:title, og:description, og:type. That is the whole of it.

Missing, all of it:

	State
`robots.txt`	404
`sitemap.xml`	404
`<link rel="canonical">`	absent
`og:image`	absent
`og:url`	absent
`twitter:card`	absent
`application/ld+json`	absent
`hreflang`	absent

So every link shared to Signal, WhatsApp, Slack or anywhere else renders with no preview
image, and search engines are given no sitemap, no canonical, and no structured data.

Worth doing, roughly in order of return

robots.txt and sitemap.xml. The cheapest possible win and currently a 404. The
sitemap should be generated from the same route list the pages come from, not hand-kept
— a hand-kept one is wrong within a month.
og:image. Link previews are how this actually spreads: somebody shares
flamenet.io in a group chat and it renders as a bare URL. A single well-made card
would change every share.
canonical and og:url. The same content is reachable on more than one host —
relay.flamenet.io, flamenetmessenger.com, and every self-hoster's domain. Without a
canonical, those compete with each other.
Structured data (SoftwareApplication, Organization, FAQPage for the /get
questions). /comparison and /get are the pages most likely to earn a rich result.
Titles and descriptions per page. Check they are distinct and written for a person
searching, not lifted from the <h1>.
hreflang, once Spanish exists — see #47. Doing SEO and localization in the wrong
order means redoing part of it.

Two things to be careful about

⚠️ Do not let SEO pull the copy toward keyword filler. The pages are plainly written and
avoid jargon on purpose — that is a product decision recorded in several places, and
"optimising" a sentence into search bait would undo it. Better titles and real structured
data beat stuffed prose.

⚠️ Do not index what should not be indexed. /downloads/* and any self-hoster's copy
of these pages should not compete with the canonical site, and robots.txt is where that
gets decided rather than discovered.

ℹ️ Related: #47 (localization — Spanish first, and six of the seven countries sign-up is
open in are Spanish-speaking), #44 (features page shape).

**Ask:** make the public pages findable. ## What is actually there today (audited 2026-09-01) `layout.leaf` has `charset`, `viewport`, `<title>`, `description`, and three Open Graph tags — `og:title`, `og:description`, `og:type`. That is the whole of it. **Missing, all of it:** | | State | |---|---| | `robots.txt` | **404** | | `sitemap.xml` | **404** | | `<link rel="canonical">` | absent | | `og:image` | absent | | `og:url` | absent | | `twitter:card` | absent | | `application/ld+json` | absent | | `hreflang` | absent | So every link shared to Signal, WhatsApp, Slack or anywhere else renders with **no preview image**, and search engines are given no sitemap, no canonical, and no structured data. ## Worth doing, roughly in order of return 1. **`robots.txt` and `sitemap.xml`.** The cheapest possible win and currently a 404. The sitemap should be generated from the same route list the pages come from, not hand-kept — a hand-kept one is wrong within a month. 2. **`og:image`.** Link previews are how this actually spreads: somebody shares `flamenet.io` in a group chat and it renders as a bare URL. A single well-made card would change every share. 3. **`canonical` and `og:url`.** The same content is reachable on more than one host — `relay.flamenet.io`, `flamenetmessenger.com`, and every self-hoster's domain. Without a canonical, those compete with each other. 4. **Structured data** (`SoftwareApplication`, `Organization`, `FAQPage` for the `/get` questions). `/comparison` and `/get` are the pages most likely to earn a rich result. 5. **Titles and descriptions per page.** Check they are distinct and written for a person searching, not lifted from the `<h1>`. 6. **`hreflang`**, once Spanish exists — see #47. Doing SEO and localization in the wrong order means redoing part of it. ## Two things to be careful about ⚠️ **Do not let SEO pull the copy toward keyword filler.** The pages are plainly written and avoid jargon on purpose — that is a product decision recorded in several places, and "optimising" a sentence into search bait would undo it. Better titles and real structured data beat stuffed prose. ⚠️ **Do not index what should not be indexed.** `/downloads/*` and any self-hoster's copy of these pages should not compete with the canonical site, and `robots.txt` is where that gets decided rather than discovered. ℹ️ Related: #47 (localization — Spanish first, and six of the seven countries sign-up is open in are Spanish-speaking), #44 (features page shape).

paulhitt commented

2026-09-08 15:02:28 +00:00

(Migrated from gitlab.com)

Done. The audit table from the description, re-measured against production 2026-09-08:

	State then	State now
`robots.txt`	404	200, names the sitemap
`sitemap.xml`	404	200, 27 urls, generated from the `pages` table
`<link rel="canonical">`	absent	present, built from the requesting host
`og:image`	absent	present, absolute, from `#(origin)`
`og:url`	absent	present
`twitter:card`	absent	present
`application/ld+json`	absent	present (WebSite + Organization)
`hreflang`	absent	present where a translation exists

Shipped across !83 (robots/sitemap/canonical/og), !85 (structured data), !86 + !87 (three routes were rendering href="" because their contexts never set the chrome, and ReleasesContext encodes by hand and dropped it), and !91 (Spanish + hreflang).

hreflang reciprocates properly and only appears where it can:

/get      → en /get · es /es/get · x-default /get
/about    → none (no translation, so nothing to advertise)

⚠️ /es/get is live and reachable but deliberately not linked from the nav — it wants a native-speaker read first, since the copy is legal-adjacent in places and an overclaim in Spanish is the same FTC problem in another language. Remaining page translations are tracked in #47, not here.

Two things the tests now hold in place: every render call site must set the canonical, and a hand-written encoder must not drop it. Both were checked by reintroducing the bug.

Done. The audit table from the description, re-measured against production 2026-09-08: | | State then | State now | |---|---|---| | `robots.txt` | **404** | 200, names the sitemap | | `sitemap.xml` | **404** | 200, 27 urls, generated from the `pages` table | | `<link rel="canonical">` | absent | present, built from the requesting host | | `og:image` | absent | present, absolute, from `#(origin)` | | `og:url` | absent | present | | `twitter:card` | absent | present | | `application/ld+json` | absent | present (WebSite + Organization) | | `hreflang` | absent | present where a translation exists | Shipped across !83 (robots/sitemap/canonical/og), !85 (structured data), !86 + !87 (three routes were rendering `href=""` because their contexts never set the chrome, and `ReleasesContext` encodes by hand and dropped it), and !91 (Spanish + hreflang). `hreflang` reciprocates properly and only appears where it can: ``` /get → en /get · es /es/get · x-default /get /about → none (no translation, so nothing to advertise) ``` ⚠️ `/es/get` is live and reachable but **deliberately not linked from the nav** — it wants a native-speaker read first, since the copy is legal-adjacent in places and an overclaim in Spanish is the same FTC problem in another language. Remaining page translations are tracked in #47, not here. Two things the tests now hold in place: every render call site must set the canonical, and a hand-written encoder must not drop it. Both were checked by reintroducing the bug.

paulhitt (Migrated from gitlab.com) closed this issue

2026-09-08 15:02:29 +00:00

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: extant2000/flamenet-relay#48