Is every web resource a web page?

No. Pages are common resources, but services, collections, physical things, and abstract concepts can also be identified.

Can one resource have several representations?

Yes. Different formats or languages may represent the same resource under the site's identification conventions.

Does a stable URL guarantee a stable quotation?

No. Content can change while the identifier remains the same; exact evidence may require a version or preserved representation.

AI Glossary / Graphs and symbolic AI / Intermediate

Web resource

Definition

A web resource is something identified within the Web's architecture, commonly by a Uniform Resource Identifier, or URI. It may be a document, image, service, collection, or something described through the Web, such as a person or a physical object. The resource is distinct from the particular data returned when someone follows its identifier.

What it means

A useful distinction separates the resource, its identifier, and its representations. A catalog entry is a resource; its URI identifies it; an HTML page or JSON response can represent its current state. The same resource may have several representations, and its state can change over time.
RFC 3986 deliberately uses a broad definition. A resource does not have to be a downloadable file or even something accessible through the Internet. Consequently, identifying a person is conceptually different from identifying a web page about that person. Applications need conventions that make the intended identity clear.

How it works

A client uses an identifier to request information or interact with a service. A server may return a representation together with metadata describing its media type and other properties. Content negotiation can select an appropriate representation, such as HTML for a browser or JSON for a software client.
The returned bytes are evidence about the resource under the conditions of that request. They need not be a permanent copy or its only possible representation. A later request may return updated information. A URI may also redirect, become unavailable, or require access rights. Identity, availability, and the exact contents of a response are related design concerns, but they are different questions.

Three practical examples

A product catalog can identify an item consistently while offering a human-readable page and a machine-readable description.
A knowledge graph can use identifiers for organizations and separate identifiers for documents describing those organizations.
An AI research assistant can record the resource URI, retrieval time, and relevant version information when collecting supporting material.

Think of it this way

Think of a library catalog record and the different views printed from it. The underlying record, its catalog number, and each printed view serve different purposes. The analogy has limits: Web identifiers can identify far more than records, and a server's responses do not automatically establish that every client agrees about the resource's identity.

Worked example

Consider the fictional identifier https://example.org/catalog/item-42. In this example, the organization defines it as identifying one catalog entry.
At 09:00, a browser requests HTML and receives a page describing the item. A software client then requests JSON and receives structured fields for the same entry. At 17:00, an editor changes the availability description, and a new HTML request returns updated text.
There are three observed responses in this scenario. That does not imply three different resources. The organization has kept one resource identifier while serving different representations and reflecting a changed state.
Now suppose a researcher needs to support a quotation from the morning page. Recording only the stable catalog URI may not reproduce those exact words later. A version-specific reference or preserved copy, together with appropriate retrieval details, provides better evidence of what was actually consulted.
Conversely, two different URIs could be configured to refer to the same resource. Identical response bytes alone would not prove this: two separate catalog entries might temporarily contain identical placeholder descriptions. Identity follows the system's intended identification and linking conventions, not a simple count of files or matching strings.

Illustrative case study

Fictional case: a learning website initially stores the identifier for an AI method and the address of its explanatory article in one field. When it adds translations and revised articles, staff become unsure whether each translation represents a new method.
The team separates the method's identity from the documents describing it. Article records then carry language and revision information, while relationships connect them to the method. The change improves the clarity of the content model; it does not establish that the articles are accurate or that external websites use compatible identifiers.

Common misconceptions and mistakes

Do not assume resource means file. Do not assume one URI can return only one format or that unchanged identifiers guarantee unchanged text. Do not treat a web page about a company as automatically identical to the company itself.
Also avoid assuming that an identifier guarantees successful retrieval. Resources may be temporarily unavailable, access-controlled, or described indirectly. For AI systems, successful retrieval is only one step: relevance, provenance, freshness, and the reliability of the returned information still require assessment.

Business use

Clear resource identification supports content management, data integration, API design, and evidence tracking. Teams can update presentation formats without confusing the underlying objects they describe.
Before building an AI knowledge service, decide what each identifier denotes, how versions are represented, and how references should survive updates. Preserve response metadata when exact evidence matters. Test whether integrations accidentally create duplicate entities from language variants or merge distinct entities because their descriptions look similar. These decisions are especially useful when a website serves both readers and automated retrieval systems.

Check your understanding

A catalog URI returns HTML in the morning and JSON to another client. Must these responses identify two separate resources?

Answer: No. They can be different representations of the same resource. The intended identity depends on the application's identification model.

Related terms

Representation
Linked data
URI
URL
HTTP
Content negotiation
Knowledge graph
Resource identifier
Provenance

Learn next

Choose one fictional course with two language versions and three revisions. Draw separate boxes for the course, its descriptive documents, their identifiers, and their representations. Explain which reference a reader should use to identify the course and which should support a quotation from an earlier revision.

Sources

Primary sources: W3C, Architecture of the World Wide Web, Volume One: https://www.w3.org/TR/webarch/ ; IETF, RFC 3986, resource and identifier definitions: https://www.rfc-editor.org/rfc/rfc3986 . The catalogue and learning-site scenarios are original and fictional.

Previous: Web query classification Next: Web scraping