Resolver precedence

Every effective SEO value — title, description, canonical, robots, images — is produced by SEOResolver merging six layers. Higher layers win, and null never overwrites a value from a lower layer, so every page always renders something sensible.

The six layers

From lowest (always present) to highest (always wins):

#	Layer	Source	Typical use
1	Site config	`config/seo.php` (`site_name`, `title_suffix`, `default_og_image`, `default_robots`, …)	Brand-wide defaults
2	Global DB defaults	`seo_defaults` rows with no model type	Editable site-wide defaults without a deploy
3	Model-type defaults	`seo_defaults` rows scoped to a model class	"All products get this OG image"
4	Route defaults	`seo_defaults` rows scoped to a route name	Static pages (`home`, `contact`) with no model
5	Computed values	Derived from the model's own attributes	Fallback title from `title`, description from `excerpt`/`body`, …
6	Explicit values	The model's `seo_meta` row (`saveSEO()`)	What editors set by hand

php

use Rankbeam\Seo\Facades\SEO;

$seo = SEO::resolve($post);          // model page: layers 1-3 + 5-6
$seo = SEO::forRoute('contact');     // route page: layers 1-2 + 4

The result is an immutable SEOData value object consumed by every renderer (Blade, array, Inertia).

Computed fallbacks (layer 5)

When no explicit value exists, the resolver derives one from the model:

Title — the model's title/name attribute.
Description — the first attribute in seo.computed.description_fields (default chain: excerpt, summary, description, intro, lead, teaser, content, body, text, article) that contains meaningful text. HTML is stripped, entities are decoded, and the text is truncated at a word boundary (seo.computed.description_max_length, default 160 — no ellipsis).
Robots — from a model's getSEORobots() hook or an is_indexable attribute (see Controlling robots and indexability).
URL-derived values — canonical and og:url from getUrlForSEO().

Controlling robots and indexability

Per-model noindex is built in — no extra package or column ceremony. The HasSEO trait doesn't declare a robots method (it's optional), so it's easy to miss, but the resolver already honours three sources, highest priority first:

Priority	Source	Example
1	Explicit `seo_meta.robots`	`$page->saveSEO(['robots' => 'noindex,follow'])`
2	A `getSEORobots(): ?string` hook on the model	return `'noindex, nofollow'`, or `null` to fall through
3	An `is_indexable` attribute (column or accessor)	falsy ⇒ `noindex, nofollow`; truthy ⇒ `index, follow`

php

class Page extends Model
{
    use HasSEO;

    // Option A: let the resolver derive robots from a boolean flag.
    //   Schema::table('pages', fn ($t) => $t->boolean('is_indexable')->default(true));

    // Option B: compute it from your own state.
    public function getSEORobots(): ?string
    {
        return $this->status === 'draft' ? 'noindex, nofollow' : null;
    }
}

// Option C: set it explicitly per page (wins over A and B).
$page->saveSEO(['robots' => 'noindex, follow']);

What actually renders

The resolved directive is filtered by the emit policy before it reaches the <head>: the <meta name="robots"> tag is emitted only when the directive deviates from default_robots (default index,follow). So:

an indexable page (resolves to index, follow) emits no robots tag — its absence is exactly what a crawler reads as index,follow;
a non-indexable page emits <meta name="robots" content="noindex, nofollow">;
any deviating directive (noindex, max-snippet:-1, unavailable_after, …) is emitted verbatim, preserving the spacing you typed.

Set seo.robots.emit_default = true to always render the tag. Full details in the robots rendering policy.

Policies applied after resolution

These run regardless of which layer produced the value:

Title suffix — title_suffix is appended unless the resolved title already ends with it. If a route-default template already contains your brand, end the template with the suffix to avoid "Brand — X | Brand".
Canonical query-strip — derived canonicals (model URL / current URL) get the query string stripped, except any keys in canonical.query_whitelist (e.g. page for paginated archives), which are kept; explicitly set canonicals are preserved verbatim.
Absolute social images — og:image and twitter:image are always emitted as absolute URLs (the Open Graph spec requires it), even when the stored value is a relative path.

Inspecting which layer won

The Filament package surfaces this per field (Manual / Content fallback / Model-type default / Global default / Site config / Derived from URL). In code, SEOWarningEvaluator exposes the same manual-vs-fallback distinction for building your own admin indicators.

Resolver precedence ​

The six layers ​

Computed fallbacks (layer 5) ​

Controlling robots and indexability ​

What actually renders ​

Policies applied after resolution ​

Inspecting which layer won ​