The vocabulary of copying a website

28 terms that decide whether a copy works: what the browser-rendered state is, why computed styles matter more than stylesheets, what a remote reference costs you, and how fidelity is measured rather than claimed.

Browser-rendered state
The page as it exists after the browser has parsed the HTML, applied every stylesheet and run the scripts that build the DOM. It is what you see on screen, and on a modern site it is very different from the HTML the server sent.
Computed style
The final value the browser arrived at for every CSS property on an element, after cascade, inheritance, media queries and any style a script set. Copying computed styles is why a capture survives without the original stylesheets.
DOM
The Document Object Model: the live tree of nodes the browser builds from the HTML and then lets scripts modify. Reading the DOM after scripts have run is the difference between copying a page and copying its source file.
Shadow DOM
A separate DOM subtree attached to an element and isolated from the page's stylesheets, used by web components and design systems. A copier that does not walk into shadow roots silently loses whole sections of a page.
Pseudo-element
A generated box like ::before or ::after that exists only in CSS and has no HTML node. Icons, decorative underlines and counters usually live here, so a copy that ignores pseudo-elements loses ornament that a reader notices immediately.
Standalone HTML file
A single .html file that renders correctly with no network access, because every stylesheet, font, image and script it needs has been inlined into it.
MHTML
An older single-file page archive format that wraps a page and its resources as a MIME multipart document. Chrome can save it, but it stores the source page rather than the rendered result and few tools can edit what comes out.
Data URI
A URL that carries the file's bytes inline, base64-encoded, instead of pointing at a server. It is how a font or image gets embedded into a standalone HTML file, at roughly a 33% size cost.
Hotlinking
Loading an asset from someone else's server instead of your own. A "copy" that hotlinks the original site is not a copy — it breaks the moment the original changes a path or blocks the referrer.
Remote reference
Any URL in a captured file that still points at the original origin. Counting remote references is the honest test of whether a capture is self-contained; ours is reported in the benchmark table.
Pixel diff
Comparing two screenshots pixel by pixel and reporting the percentage that match. It is the only fidelity measure that cannot be argued with, which is why every number on this site comes from one.
Lazy loading
Deferring an image or section until it is about to enter the viewport. A capture taken without scrolling the page first will simply not contain lazily loaded content.
Critical CSS
The subset of a stylesheet needed to paint what is visible without scrolling, inlined into the HTML so the first paint does not wait on a stylesheet request.
Largest Contentful Paint
A Core Web Vital measuring when the biggest element in the viewport finishes rendering. Fonts, hero images and render-blocking CSS are the usual reasons it is slow.
Cumulative Layout Shift
A Core Web Vital measuring how much content jumps around as a page loads. Images without width and height attributes and late-loading webfonts are the common causes.
Gutenberg block
The unit of content in the WordPress block editor: a piece of markup with a registered type and attributes the editor knows how to render and edit. Exporting to blocks rather than raw HTML is what makes a captured page editable in WordPress.
Full Site Editing
The WordPress model where templates, headers and footers are themselves built from blocks, so an imported page can be styled with the same tools as the rest of the site.
Manifest V3
The current Chrome extension platform: a service worker instead of a background page, declarative network rules, and no remotely hosted code. It is why a modern capture extension does its work in the page rather than in a hidden tab.
Service worker
The background script of an MV3 extension. It is event-driven and can be shut down at any moment, so capture state has to be persisted rather than held in memory.
Side panel
A Chrome surface that docks an extension UI beside the page instead of in a popup that closes on click. It is what lets a capture run while you keep interacting with the page.
Headless browser
A real browser driven by code with no visible window. Headless captures scale, but a page that behaves differently without a visible viewport — or that blocks automation — will render differently than it does for a person.
WARC
The web archive format used by institutional crawlers: it records the HTTP exchanges themselves, not the rendered page. Better for provenance, worse for getting an editable page back.
Robots.txt
A file at the root of a site telling crawlers which paths they may request. It governs crawling, not copyright, and it does not stop a browser you are driving yourself.
Canonical URL
The address a page declares as its own preferred version, so that duplicates and parameter variants consolidate onto one URL in search results.
Structured data
Machine-readable facts about a page, usually JSON-LD using schema.org types. It is how a page tells a search or answer engine what it is rather than hoping the text is parsed correctly.
Answer engine optimisation
Writing and marking up a page so an engine that answers in prose — an AI overview, a chatbot, a voice assistant — can quote it accurately: a self-contained opening answer, plain-text structured data, and claims with a checkable source.
Generative engine optimisation
The same goal for models that synthesise rather than link: being the source a model reaches for, which in practice means publishing numbers with a method, being consistent about your own identity across pages, and being crawlable by model crawlers.
Fidelity
How close a copy is to the original, measured rather than asserted. On this site it always means the pixel-match percentage of a rebuilt page rendered with the original server unreachable.

Where these terms get used