Public and reproducible

# How LinkGlimpse checks link previews

LinkGlimpse turns the initial HTML and public share image for a URL into deterministic metadata checks, modeled previews, and copy-ready repairs. This page explains what the result measures—and what it cannot prove.

Methodology reviewed 2026-09-18.

## Inspection workflow

### 1\. Validate the submitted URL

LinkGlimpse accepts HTTP and HTTPS URLs without embedded credentials. It resolves the hostname before every fetch and rejects local, private, reserved, and documentation-only network ranges, including unsafe redirect targets.

### 2\. Fetch the public HTML

The checker follows redirects explicitly, records each response, accepts HTML pages, and reads a bounded response. It does not sign in, bypass crawler controls, or execute the page's client-side JavaScript before extracting metadata.

### 3\. Extract metadata signals

The parser reads the document title plus Open Graph, Twitter Card, description, canonical, robots, Googlebot, site-name, and author values from the returned HTML. Relative canonical and image URLs are resolved against the final page URL.

### 4\. Inspect the share image

When an image is declared, LinkGlimpse checks whether the public URL responds, whether its content type is an image, and—when detectable from the file header—its dimensions. The diagnostic also records HTTPS use, aspect ratio, and reported file size.

## How the diagnostic score works

The score is a weighted readiness summary of the checks returned for that run. A passing check receives its full weight, a warning receives half, and a failed check receives none. Page fetch, core Open Graph fields, Twitter Card type, canonical and robots directives, and image fetch quality carry more weight than supporting checks.

The score is not a search ranking, traffic forecast, accessibility grade, or guarantee that a platform has refreshed its cache. A high score means the inspected response contains the signals LinkGlimpse tests; it does not certify every platform-specific rule.

### Checks included when data is available

-   Successful page response and redirect path
-   Open Graph title, description, image, and URL
-   Twitter Card type, title, description, and image fallbacks
-   Canonical URL and noindex detection
-   Public image response, HTTPS, content type, dimensions, aspect ratio, and size

## How previews are modeled

Each preview maps the extracted title, description, domain, URL, and image into a representative card layout. Where appropriate, the model uses documented-style fallbacks such as Open Graph values when a dedicated Twitter value is absent. Rendering is deterministic: LinkGlimpse does not ask an AI model to invent missing page content.

Platforms operate private crawlers, caches, crop rules, and client interfaces that can change independently. LinkGlimpse therefore describes previews as representative rather than pixel-perfect, and it cannot force Facebook, LinkedIn, X, or another platform to refresh a cached card.

## Boundaries and known limitations

### Initial HTML only

Metadata added only after client-side JavaScript runs may not appear in the inspection. Server-render important tags for crawler reliability.

### Public resources only

Authenticated pages, local development hosts, private networks, and reserved addresses are intentionally unavailable.

### A current fetch, not cache state

The report describes the response LinkGlimpse received during the run. A social platform may hold an older cached response.

### Platform policies remain authoritative

A passing LinkGlimpse report does not replace each platform's current documentation, debugger, or post inspector.

## Privacy and reproducibility

The hosted checker fetches only the public URL a user submits. Metadata and preview API responses use private, no-store cache controls. Product analytics reduce a submitted URL to its hostname; the CLI, browser extension, and local MCP source do not add their own telemetry.

[Review diagnostic examples](/examples)[Use the metadata API](/api)[Inspect the source code](https://github.com/stnikolaus/linkglimpse)[Read the privacy policy](/privacy)

---

Source: https://www.linkglimpse.com/methodology