docs
/
AI skills

Site identity, SEO tags and link sharing

The site's title, description, keywords, robots, logo, share image, favicon and social profiles, and each page's own SEO fields — exactly how each becomes the page's title, meta description, canonical, Open Graph and Twitter tags, icons and JSON-LD, which fields are ignored, and how to check the result on the site.

Every page on a hosted site gets its <title>, meta description, canonical link, Open Graph and Twitter tags, icon links and JSON-LD from two records: the site record and the page record. The renderer writes them. The page's own HTML can't override them.

This page covers what those tags say. Which pages are in the sitemap, robots.txt, noindex and site search are in Sitemap, robots and search indexing. Building the page itself is in Build a website page. The site record's fields are in Sites, and generating favicon files is in Set a logo and favicon.

The fields, and where each one goes

Site record (site.data)

Studio: Site Config › Identity (site-config-form.tsx 743–781) and Site Config › Tracking & Social (938–1050). Save changes writes the whole record back, keeping only features from the server copy (177–187).

FieldStudioBecomesNotes
nameIdentity › Name (read only)og:site_nameThe internal key, such as phasite. Tested: it shows as og:site_name. It is never used in <title>.
titleIdentity › Titlethe brand half of <title>; Organization and WebSite name in JSON-LD; Product brandThe brand. Without it, the tab shows the page's title alone.
descriptionIdentity › Descriptionmeta description, og:description and twitter:description for any page without its own; Organization and WebSite descriptionA site made by the API starts as Site <name> (appengine site.service.ts 241, 1732). Tested: phasite still says "Site phasite" in its JSON-LD. Replace it.
keywordsIdentity › Keywords<meta name="keywords"> for any page without its ownA string, comma separated.
robotsIdentity › Robots<meta name="robots"> for any page without its ownA string, such as index, follow. A value containing noindex or none also closes robots.txt to the whole site. See Sitemap, robots and search indexing.
logoIdentity › Logo/logo.png, which is also the og:image fallback; JSON-LD Organization logoStudio stores the picked file object. A string URL or { url } both work.
imageIdentity › Share imagenothingTested: og:image ignores it. See the warning below.
faviconIdentity › Favicon (stores the URL as a string, 768)/favicon.ico, /favicon.png, /apple-touch-icon.pngA string or { url }. Both work now; see the note below.
faviconAssets.ico / .pngnone (API only)/favicon.ico and /favicon.png, ahead of faviconWhat the generator pair is kept as.
appleTouchIconnone (API only)/apple-touch-icon.png, ahead of logo
defaultLanguage, else languages[0]Localisationog:locale (fr-FR → fr_FR; Studio stores two-letter codes, so fr stays fr; default en_US)Tested. Nothing else: <html lang> is always en, and no hreflang is emitted.
social[] { name, handle, url }Tracking & Social › Social profileseach url → JSON-LD Organization sameAs; a row named twitter (or x) with a handle → twitter:site and twitter:creatorname is one of facebook, twitter, linkedin, instagram, tiktok. Tested: @skilltestseo on both tags, and both URLs in sameAs.
tracking.googleSiteVerificationTracking & Social › Google site verification<meta name="google-site-verification">Tested.
tracking.bingSiteVerificationTracking & Social › Bing site verification<meta name="msvalidate.01">Tested.
seo.noIndexClassic form onlyevery page noindex, nofollow; robots.txt Disallow: /See Sitemap, robots and search indexing.
seo.googleSiteVerification, seo.bingSiteVerificationnone (declared in base-app's copy of the model, lib/models/site.ts 625–632)nothingTested: the tag carries the tracking.* value, never this one.
post.images (top level, not data)Classic form › Postsite-wide share image, after the page'sAn array of { url, meta: { width, height }, path }.

Identity › Share image does nothing. It saves site.data.image, and nothing on the site reads that field (base-app meta.tsx 69–87). Tested: with a share image set, og:image was still /logo.png. For a site-wide share image, set the top-level post.images on the site record, or give each page its own (step 4). Until then, a site with only a small favicon shares that favicon as its card.

An object favicon works now. Older notes and docs say site.data.favicon must be a string or no icon is emitted. The code has changed. The icon links now point at the site's own /favicon.ico and /favicon.png, and those routes accept a string, { url }, { signedUrl }, { path } (only when it's an absolute URL) or an array (lib/brand-asset.ts 43–52). Tested: favicon: { url, path } still gave both icon links and a 200 image/png. Store a string anyway; it's what Studio and the generator write.

Page record

Studio: Build Studio › Page settings › Page Config (the page form) and Page settings › Post (the top-level post block, base-properties.tsx 162–168).

FieldBecomesNotes
data.titlethe page half of <title>; og:title, twitter:title; JSON-LD WebPage nameFalls back to data.name.
data.descriptionmeta description, og:description, twitter:description; WebPage descriptionFalls back to the site's.
data.keywords<meta name="keywords">Falls back to the site's.
data.robots<meta name="robots">Falls back to the site's. The site switch and the sitemap rules outrank it.
data.faviconone <link rel="icon"> to that URL, and no apple iconOnly an absolute http(s) URL. Anything else is ignored (meta.tsx 183). Tested. It isn't in the page form; write it by API.
post.images[]og:image (with og:image:width/height from meta, og:image:alt = path) and twitter:imageMust be an array. Tested: a single { url } object is ignored and the page shares /logo.png.
post.categories[], post.tags[]<meta name="category">, <meta name="tags">, plus apple-mobile-web-app-capableTested.
post present at all, or data.appType: "post"og:type article instead of websiteTested: giving a page a Post block turns it into an article.
data.slug / data.namethe canonical pathSee canonical below.

How each tag is built

All of this is base-app app/[[...slug]]/meta.tsx (generateSeoMetadata), called by the slug layout for every page ([[...slug]]/layout.tsx 21–43) and again by page.tsx for product pages.

<title>

meta.tsx 26–43. The site's title is site.data.title; the page's is the product name (on a product page), else data.title, else data.name.

Site titlePage title<title>Tested on
PhaZZerSkilltest SEO metaSkilltest SEO meta | PhaZZera test page
PhaZZerPhaZZer brand testPhaZZer brand testa test page
PhaZZerPhaZZer|Less Lethal WeaponsPhaZZer|Less Lethal Weaponsthe home page
PhaZZernone (a feature address such as /store)PhaZZer/store, /store/<product>
noneAboutAboutcode
nonenoneUntitledcode

The brand suffix is dropped when the page title already contains the site title, comparing letters and digits only. So a short site title is matched inside other words: with a site title Go, a page titled "Google reviews" loses its suffix.

Meta description, keywords, robots

TagFirst ofThen
descriptionthe product's description (product page) → data.description → site.data.descriptionthe platform's own Appmint description
keywordsdata.keywords → site.data.keywordsthe platform's own list of "AI website builder" keywords
robotsnoindex, nofollow when seo.noIndex is on, the page is the payment page, or the sitemap rules leave the page out → data.robots → site.data.robotsindex, follow, nocache

Leave a field empty and the platform's own text fills it. The base layout (app/layout.tsx 13–81) carries Appmint's marketing title, description, keywords, robots and share card. A site tag that has no value doesn't replace it. Tested on phasite, which has no keywords: every page says <meta name="keywords" content="AI website builder,AI content management,…">. Set keywords on the site (or on every page), and never leave the site's description empty.

Canonical

meta.tsx 59–66, lib/site-url.ts getCanonicalOrigin 51–63.

  • The origin is the host the visitor used, with two exceptions. When the site has a domain and the visitor came in on www.<domain> or on the platform host in site.data.hostName, it is https://<domain>. Any other host is left as it is, including a custom domain the record doesn't name and every *.localhost host.
  • The path is the address as requested, not the page's slug. Tested: a page named skilltest-seo-named with slug skilltest-seo-slugged answers at both addresses, and each one names itself as canonical. /skilltest-seo-parent/skilltest-seo-child showed the parent page with that whole path as its canonical.
  • The query string is dropped, except on a product page, which canonicalises to /product?slug=<slug>.
  • The home page's canonical is the bare origin (http://phasite-phazzer.localhost:3200, no trailing slash).

Link to each page by one address, its slug. Its name, a different capitalisation and any extra path segment all serve the same page with a canonical naming that other address.

Open Graph and Twitter

meta.tsx 69–87 and 144–174. Tested on the home page, a test page and a product page.

TagValue
og:title, twitter:titlethe <title>
og:description, twitter:descriptionthe meta description
og:image, twitter:imagethe product's first image (product page) → the page's post.images[] → the site's post.images[] → <origin>/logo.png when the site has a logo, favicon or favicon pair → none
og:urlthe canonical origin, not the page's URL. See What it can't do yet.
og:site_namesite.data.name, the internal key
og:typearticle when the page has a post block or appType: "post", else website
og:localedefaultLanguage → languages[0] → en_US
twitter:cardalways summary_large_image
twitter:site, twitter:creator@handle of the social row named twitter or x; nothing if there's no handle

The fallback share card is your logo, cut to whatever it is. /logo.png serves logo, else the favicon PNG, else favicon (brand-asset.ts 64–65). Tested: phasite has no logo, so every page without a share image shares a 64 px favicon. A card wants 1200 × 630.

An SVG logo breaks the card. The brand routes always send the content type their name says. Tested: an SVG logo came back from /logo.png and /apple-touch-icon.png as SVG bytes labelled image/png. Social sites don't render that, and browsers won't render an SVG apple icon labelled PNG. Upload a PNG or JPG logo. Uploaded SVGs are also served as application/octet-stream, so they fail in <img> too.

Icons

meta.tsx 183–198 and lib/brand-asset.ts.

The site hasThe page's <head> gets
a page data.favicon that is an absolute URL<link rel="icon" href="<that URL>"> only
favicon, faviconAssets.ico/.png or logo<link rel="icon" href="/favicon.ico" sizes="any">, <link rel="icon" href="/favicon.png" type="image/png">, <link rel="apple-touch-icon" href="/apple-touch-icon.png">
none of themthe platform's defaults: /favicon.ico, /favicon-32x32.png, /apple-touch-icon.png, a mask icon /safari-pinned-tab.svg

The site serves those addresses itself, from the site record, by fetching the stored URL on the server (brand-asset.ts 54–69):

AddressServed from (first that is an absolute URL)Content typeNothing set
/favicon.icofaviconAssets.ico → favicon → logoimage/x-icon307 to /images/platform-favicon.ico
/favicon.pngfaviconAssets.png → favicon → logoimage/png404
/apple-touch-icon.pngappleTouchIcon → logo → faviconAssets.png → faviconimage/png404
/logo.pnglogo → faviconAssets.png → faviconimage/png404

Each response carries X-Brand-Asset: <name> and Cache-Control: public, max-age=604800, s-maxage=86400. A visitor's browser may keep an old icon for a week.

Page-level icon links are stripped. The renderer removes every <link rel="…icon…"> (icon, shortcut icon, apple-touch-icon, mask-icon) from a page's head (build-studio/html-format.ts 205–208). The site record, or a page's data.favicon, is the only way to set an icon.

Tags in the page's own <head>

A Build Studio page stores each screen as a full HTML document. The renderer copies that document's <head> into the page body, after first removing what the metadata system owns (html-format.ts 222–238, html-renderer.tsx 130). Tested with one of each:

In the page's headOn the site
<title>, <meta name="description|keywords|robots">, og:*, twitter:*, <link rel="canonical">, icon linksremoved
<meta name="google-site-verification">kept, but in the body. Google only reads it in <head>, so it doesn't verify. Use tracking.googleSiteVerification.
<link rel="alternate" hreflang>kept, in the body. Search engines ignore hreflang there.
<script type="application/ld+json">kept, in the body. Body JSON-LD is valid. This is how to add a FAQ, Event or LocalBusiness block.

JSON-LD

components/seo/JsonLd.tsx, placed in the body by page.tsx 351. Every page with a site title (or name) gets:

BlockWhat it says
Organizationname = site title (else its name), url = origin, logo = the stored logo (else favicon) URL as is (a storage-bucket URL, not /logo.png), description = site description, sameAs = every social url
WebSitename, url, description, and a SearchAction aimed at <origin>/search?q=…. Tested: /search is a 404 on a hosted site.
WebPagename = page title, else name, else site title; description; url = the canonical URL; isPartOf the WebSite
BreadcrumbListevery page except the home page. One crumb per URL segment, named from the segment (skilltest-seo-meta → "Skilltest Seo Meta"), not the page title. A product page adds the product.
Productonly with ?slug=<slug> on a page named product, an appType: "product" page, or an address with a product segment. name, description, first image, sku, brand = site title, and an Offer (price, priceCurrency default USD, availability from stock → quantity → in stock, url = the canonical URL) only when the price isn't 0.

No Article block is ever emitted: the component supports one, but no page passes it (page.tsx 351).

Pages that aren't page records

AddressTitle and tagsTested
/product?slug=<slug><title> = product name (plus | site title unless it contains it); description = the product's description, HTML tags and all; canonical /product?slug=…; og:image = first product image; Product JSON-LD. The WebPage block keeps the template page's title.yes
/store, /store/<product> and other feature addresses with no page record<title> = the site title alone; the site description; canonical = the path; no Product blockyes
/blog/<post> (the Blog feature's post page)the site's defaults, the same as /blog: <title> = the site title, og:type website, the site's share image. Not the post's title, summary or cover; generateMetadata has no post case. See Write and publish a blog postyes
/login, /register, /forgot-passwordAppmint's marketing title, description, keywords and share card (og:url https://appmint.ai, twitter:creator @appmintai), and indexableyes
/checkout, /checkout/success, /account…, /order/…the same Appmint title and card (/order/track: "Track Order"), noindex, nofollowyes
the payment page, with ?invoiceNumber=, ?invoice=, ?ref=, ?t=, ?token= or ?order=<title> Pay <that value>, noindex, nofollow (page.tsx 202–208)code
any 404status 404, <meta name="robots" content="noindex">yes

Ask first

  • Which site? All of this is per site.
  • The brand name for the tab and search results, and the one-sentence description. Use the person's words.
  • The logo, the share image (ideally 1200 × 630) and the icon. As files or URLs they give you, PNG or JPG. Don't generate a logo.
  • Social profiles: each network's full URL, and the X handle if they want it on cards.
  • Search Console or Bing: the verification code (the content value of the tag they were given, not the whole tag).
  • Which pages need their own title, description or share image? Every page they want found should have its own title and description.

Steps

1. Site Config: read the site record

repo_find on site ({ "data.name": "<site>" }), then repo_find_one on site with its id for the whole record. Note what's set: title, description, keywords, robots, logo, favicon, faviconAssets, social, tracking, seo, and the top-level post. Keep this copy: it's how you put things back.

Then read what the live home page says (step 5's command). Report what's wrong before changing anything: the platform keywords, an empty or "Site " description, a 64 px share card.

Hands on: the site's id, its current values, and the list of gaps. Check: the record you read has the same name as the site the person means.

2. File manager: the logo, share image and icon

Upload each file with POST /repository/file/upload (multipart: location = a folder such as <org>/brand, and file), or in the file manager. Keep the returned path and signedUrl.

For the icon, either:

  • use a square PNG directly as favicon; or
  • Site Config › Identity › Create Favicon (POST /tools/create-site-favicon/<site> with { "site": "<site>", "path": "<logo path>" }). It needs a logo, makes favicon.ico and favicon.png at the org's bucket root, and writes favicon = the .ico URL. It writes the whole site record back, so do it before your other changes, not during them. The body's site must equal the site in the address (422 "Site name mismatch"), and path is required (422 "Favicon path is required"). Tested.
  • or POST /repository/file/create_favicon with { "location": "<path without the org prefix>" }. It returns both files and changes no record. See Set a logo and favicon.

Create Favicon leaves /favicon.png serving the .ico. It sets favicon to the .ico and doesn't set faviconAssets (appengine tools.service.ts 316–324). /favicon.png then sends the ICO bytes labelled image/png. To serve real PNG bytes, also write faviconAssets: { ico, png } from create_favicon's result.

Hands on: the URLs: logo, share image, favicon (and the .ico/.png pair, if made). Check: each URL opens as an image. A logo, share image or icon that's an SVG: stop and ask for a PNG.

3. Site Config: identity, social profiles, verification

Write only what changed, with repo_update on site: { "datatype": "site", "id": "<sk>", "patch": { … } }. repo_update merges patch into data one level deep (repository.skill.ts 177). So a key you send replaces that whole field.

{
  "title": "PhaZZer",
  "description": "Less-lethal conducted energy weapons for law enforcement and security teams.",
  "keywords": "conducted energy weapon, less lethal, law enforcement",
  "logo": { "path": "<org>/brand/logo.png", "url": "<signedUrl>" },
  "favicon": "<favicon URL>",
  "social": [
    { "name": "facebook", "handle": "", "url": "https://facebook.com/<page>" },
    { "name": "twitter", "handle": "<handle>", "url": "https://x.com/<handle>" }
  ],
  "tracking": { "…every key it already had…": "…", "googleSiteVerification": "<code>", "bingSiteVerification": "<code>" }
}
  • tracking and social are replaced whole. Send tracking with the analytics, pixel and cookies it already had, or they are wiped. Send the whole social list.
  • Verification codes go in tracking, never in seo.
  • A site-wide share image is the record's top-level post.images, which repo_update can't reach: call_appengine POST /repository/update-partial/site/<sk> with { "post.images": [ { "url": "<url>", "path": "<path>", "meta": { "width": 1200, "height": 630 } } ] }.
  • The same writes by API: POST /repository/update-partial/site/<sk> with dot-path keys, such as { "data.title": "…", "data.tracking.googleSiteVerification": "…" }.

Never send { "data": { … } } to update-partial. It replaces the whole data object, and the site loses its name, home page, domains and every other setting.

Stop: this changes what every page of the live site says in search results and when shared. Show the person the before and after for the tab title, description and share card, and wait for yes.

Hands on: the saved site record. Check: repo_find_one on the site shows the new values and everything else as it was. Site Config › Identity and › Tracking & Social show them.

4. Build Studio: each page's own SEO

For each page the person named:

  • Title and description: page_write { "id": "<page sk>", "title": "…", "description": "…" }, or Build Studio › Page settings › Page Config.
  • Keywords, robots, a page icon: repo_update on page with patch: { "keywords": "…", "robots": "…", "favicon": "https://…" }. Set robots only when the person asked for that page to be hidden from search. Hiding pages is covered in Sitemap, robots and search indexing.
  • Share image, categories, tags: the top-level post. Use call_appengine POST /repository/update-partial/page/<sk> with { "post.images": [ { "url": "…", "path": "…", "meta": { "width": 1200, "height": 630 } } ] }. Tested: the page's og:image changed at once. images must be an array. Giving a page a post also makes it og:type article.

Don't put <title>, description, Open Graph or icon tags in the page's HTML: they are removed. A JSON-LD block in the page's head is kept.

Stop: on a site that serves drafts, these are live as soon as they are saved. Say which pages change, and wait for yes.

Hands on: the page ids and what each now says. Check: repo_find_one on each page shows data.title, data.description and, when set, post.images as an array.

5. Check it on the site

Read the tags the site really sends. Never trust the record alone:

# the tags in <head> only — a page's own head is copied into the body, so don't read past </head>
curl -s "https://<site host>/<slug>?cb=$RANDOM" | awk 'BEGIN{RS="</head>"} NR==1' \
  | grep -oE '<title>[^<]*</title>|<(meta|link) [^>]*>' \
  | grep -E '<title|name="(description|keywords|robots)|rel="canonical|og:|twitter:|rel="(icon|apple-touch-icon)|site-verification|msvalidate'
curl -sI "https://<site host>/favicon.png" | grep -iE '^HTTP|content-type|x-brand-asset'
curl -sI "https://<site host>/logo.png"   | grep -iE '^HTTP|content-type'
  • <title> is what the person approved, with the brand suffix where expected.
  • keywords is no longer the "AI website builder" list.
  • og:image is the share image, not /logo.png, wherever one was set.
  • /favicon.png is 200 with X-Brand-Asset: favicon.png. A 404 means no favicon, favicon pair or logo is set.
  • The verification tag is in this <head> list, not only further down the page.

Then the person can run the URL through the platforms' own preview tools (Facebook Sharing Debugger, LinkedIn Post Inspector) and Google's Rich Results Test. Facebook and LinkedIn cache cards, so they must ask for a re-scrape there.

Hands on: the tag list per page. Check: every value matches what was agreed, on the platform host and on the custom domain.

When it's done

Tell the person, in one line each: the tab title and description search engines will see, the share image and where it came from, the icon, the social profiles, and which verification codes are in place. Name the gaps from What it can't do yet that affect their site, such as the Appmint title on /login. Open Site Config › Identity and one page on the site.

What it can't do yet

  • og:url is always the site's origin, on every page (meta.tsx 151). Tested. Facebook and LinkedIn may group every page's shares under the home page.
  • og:site_name is the site's internal name (phasite), not its title (meta.tsx 150). Tested.
  • Identity › Share image (site.data.image) is ignored. Tested. Use post.images.
  • seo.googleSiteVerification and seo.bingSiteVerification are ignored; only tracking.* is read (meta.tsx 202–203). Tested.
  • Empty site fields fall back to Appmint's own marketing text (keywords, description, robots index, follow, nocache). Tested for keywords and robots.
  • /login, /register and /forgot-password carry Appmint's title, description and share card, and are indexable. /checkout and /checkout/success carry the same title (noindex). Tested.
  • <link rel="manifest" href="/site.webmanifest"> is on every page, and that address returns an HTML page (200 text/html). Tested.
  • Product page descriptions keep their HTML. <p class="p1">… goes into the meta description, the Open Graph and Twitter descriptions, and the Product JSON-LD. Tested. The product's summary is used only when description is empty.
  • The storefront's own product pages (/store/<product>) get the site title only and no Product JSON-LD. Only the /product?slug= page gets per-product tags. Tested.
  • The JSON-LD search box points at /search, which is a 404. Tested.
  • No hreflang, and <html lang="en"> on every site. languages only changes og:locale.
  • No Article JSON-LD and no per-post title or share card for blog posts at /blog/<post>, and breadcrumbs are named from URL segments, not page titles.
  • The Organization logo in JSON-LD is the raw storage URL, unlike the share card, which uses /logo.png.
  • Create Favicon doesn't check the logo exists (tools.service.ts 311 doesn't wait for the read) and doesn't write faviconAssets.
  • The site-wide SEO switches aren't on Site Config's main screen. seo.noIndex is in the Classic form only; seo.disallow has no screen at all.

Don'ts

  • Don't put a title, description, Open Graph, canonical or icon tag in a page's HTML. It is removed.
  • Don't put a verification <meta> in a page's head. It ends up in the body and doesn't verify.
  • Don't use Identity › Share image and assume the card changed. Check og:image.
  • Don't upload an SVG as logo, favicon or share image.
  • Don't write seo.googleSiteVerification. Write tracking.googleSiteVerification.
  • Don't repo_update the site with a partial tracking or social. Send the whole field.
  • Don't write a page's post.images as a single object.
  • Don't judge the result from the record or from Studio. Read the tags from the live page.