The Web Answer Citation Protocol (WACP): Optimizing Web Architecture for Token-Efficient AI Retrieval

What Is the Web Answer Citation Protocol?

The Web Answer Citation Protocol, or WACP, is a proposed method for publishing information in a form that remains useful to human readers while making important answers easier to identify and reference independently.

WACP does not replace a normal webpage. The detailed page remains the primary document. Instead, WACP adds a small number of concise, publisher-authored Answer Objects that are directly associated with the sections that explain and support them.

Each Answer Object receives a stable identifier such as #a1. A corresponding Answer Citation such as [a1] appears with the supporting content, and the same answer can be represented in a machine-readable JSON-LD layer.

The result is a dual representation of the same information: a normal document for people and an explicit semantic representation for compatible retrieval systems.

What Problem Does WACP Address?

Webpages are designed primarily as documents. A single page may contain navigation, introductions, examples, supporting explanations, images, footers and many separate facts. That structure is useful for a human reader but can make an individual answer difficult to reference precisely.

Search engines and retrieval systems already perform their own extraction, ranking and passage-selection processes. WACP does not attempt to replace those systems. It gives the publisher a consistent way to state:

This is my concise answer, this is its stable identifier, and this is the visible content that supports it.

A consuming system remains free to ignore that declaration, compare it with other information, retrieve additional context or decide that a different source is more appropriate.

How WACP 3.0 Works

WACP 3.0 is built around four connected pieces:

1. Answer Citation: a human-visible identifier such as [a1].
2. Stable Fragment: a URL-addressable identifier such as #a1.
3. Answer Object: a concise publisher-authored response to the question or topic represented by that section.
4. Supporting Content: the ordinary visible HTML containing explanation, qualification, evidence, examples and context.

The semantic JSON-LD layer mirrors the Answer Objects and their identifiers. It does not contain a different set of claims from the visible page.

Answer Citations and Stable Fragment IDs

The Answer Citation is the visible reference point for WACP. The notation is intentionally simple:

[a1], [a2], [a3] and so on.

Each citation corresponds to a normal HTML fragment identifier:

#a1, #a2, #a3.

This makes an answer addressable using ordinary web architecture. A URL such as:

https://ericstrate.com/wacp/#a4

identifies a specific WACP Answer Object without requiring a new URL, file format or proprietary endpoint.

Publisher-Authored Answer Objects

The Answer Object must be publisher authored or publisher approved. Automated tooling may assist in drafting it, but the publisher remains responsible for the final answer and for keeping it consistent with the supporting content.

WACP 3.0 recommends keeping an Answer Object short enough to function as a direct response. As a practical guideline, an answer should normally remain within a single paragraph and approximately 50 words or fewer.

That limit is a publishing guideline rather than a claim about how many tokens a search engine or language model must consume.

The Answer Object must not materially contradict, exaggerate or omit a qualification that would change the meaning of its supporting content.

Supporting Content and Provenance

WACP deliberately does not replace long-form content with a collection of short answers.

The supporting document is important because concise answers often require explanation. A statement may depend on methodology, dates, exceptions, evidence or the specific meaning of a term.

A retrieval system that needs more information should be able to move from the Answer Object directly to the normal page section that supports it.

This relationship between concise answer and visible context is one of the primary differences between WACP and publishing disconnected hidden summaries solely for machines.

The Semantic JSON-LD Layer

The WACP semantic layer mirrors the answers that already exist in the visible page.

The recommended WACP 3.0 representation uses standard Schema.org vocabulary:

ItemList → ListItem → DefinedTerm

The ItemList describes the ordered collection of answers. Each ListItem establishes its position, while the DefinedTerm carries the individual answer identifier and concise description.

A WACP implementation should maintain a one-to-one relationship between the visible answer and its semantic counterpart.

Why WACP Uses Standard Schema.org Types

WACP does not require Schema.org, search engines or AI companies to recognize a new schema type called WACPAnswer.

Instead, WACP 3.0 uses existing semantic concepts where they accurately describe the information being published.

The WACP behavior comes from the relationship between:

• the visible Answer Citation,
• the stable fragment ID,
• the publisher-authored answer,
• the supporting HTML, and
• the matching semantic object.

This approach keeps the protocol compatible with ordinary web publishing rather than requiring a proprietary content format.

Token Efficiency as a Design Goal

One motivation for WACP is the possibility that a compatible retrieval system could identify a relevant publisher-authored answer before passing a much larger amount of page content into an expensive processing step.

That possibility should not be presented as a guaranteed reduction in tokens, latency, server cost or energy consumption.

Different crawlers and retrieval systems use different architectures. Some may ignore WACP completely. Others could choose to retrieve an Answer Object first and supporting content only when additional context is required.

WACP therefore treats semantic efficiency as a hypothesis that can be measured by consuming systems rather than as a performance claim made by the publisher.

Optional Cryptographic Provenance

A machine-readable answer can be copied, altered or separated from the page that originally published it. Cryptographic provenance is one possible way to address that problem.

WACP 3.0 therefore allows future implementations to associate Answer Objects with signatures, content hashes or other established integrity mechanisms.

This is an advanced and optional extension.

A publisher does not need cryptographic infrastructure to implement WACP’s core relationship between Answer Citation, stable identifier, Answer Object, supporting content and semantic representation.

What WACP Does Not Claim

WACP is intentionally narrow in what it claims to do.

It does not instruct Google, Bing, ChatGPT, Perplexity, Claude or any other system which source to rank, retrieve or cite.

It does not create authority merely because a publisher labels something an answer.

It does not guarantee that an AI system will trust a publisher’s summary instead of independently analyzing the surrounding page.

It is not a Google ranking factor, a Schema.org standard or an industry-adopted specification.

WACP gives publishers a consistent method for exposing concise, addressable answers while preserving the context required to evaluate them. The consuming system remains responsible for deciding whether the source and answer are useful.

How WACP Relates to SEO, AEO and GEO

WACP is most relevant to Answer Engine Optimization because it deals directly with identifying and addressing concise answers.

It can also be considered within Generative Engine Optimization because generative systems frequently use retrieved source material when constructing responses.

Neither changes the underlying requirement for good SEO.

A WACP implementation on an inaccessible, unhelpful or untrusted page does not solve those problems. Technical SEO, content quality, entity clarity, internal architecture and real authority remain important independently of WACP.

This is why WACP is implemented on EricStrate.com as an additional information architecture rather than as a replacement for normal search optimization.

WACP 3.0 Implementation Example

A simplified visible answer unit can be implemented with ordinary HTML:

<section>

  <div
    id="a1"
    data-wacp="answer"
    data-answer-id="a1">

    <span>[a1]</span>

  </div>

  <h2>What is WACP?</h2>

  <p>
    Supporting content explaining and documenting
    the publisher-authored answer.
  </p>

</section>

The corresponding semantic representation can use standard JSON-LD:

{
  "@context": "https://schema.org",
  "@type": "ItemList",
  "@id": "https://example.com/page/#wacp-answers",
  "numberOfItems": 1,
  "itemListElement": [
    {
      "@type": "ListItem",
      "position": 1,
      "item": {
        "@type": "DefinedTerm",
        "@id": "https://example.com/page/#a1",
        "termCode": "a1",
        "name": "What is WACP?",
        "description": "A concise publisher-authored answer.",
        "url": "https://example.com/page/#a1"
      }
    }
  ]
}

In a production implementation, the description should contain the same substantive answer presented through the visible WACP Answer Object.

WACP 3.0 Conformance Requirements

Required

• Each WACP Answer Object must have a unique identifier.
• The identifier must correspond to a stable HTML fragment such as #a1.
• A human-visible Answer Citation must identify the associated answer.
• The Answer Object must be publisher authored or publisher approved.
• Supporting content must remain visible in the normal webpage.
• Machine-readable answer text must not materially contradict the visible answer or supporting content.
• The semantic layer must reference the same answer identifier used by the visible page.

Recommended

• Keep Answer Objects to one concise paragraph.
• Target approximately 50 words or fewer when the subject permits.
• Use established Schema.org vocabulary where appropriate.
• Make Answer Citations usable by keyboard and touch interfaces.
• Allow a reader to navigate from an Answer Object to the context that supports it.

Optional

• Cryptographic signatures.
• Content hashes.
• Interactive citation popovers.
• Hover behavior on pointer-based devices.
• Additional provenance metadata.

The visual design of an Answer Citation is not part of the protocol. A sidebar, inline citation, expandable element or other accessible interface can be used as long as the semantic relationship remains intact.

Versioning, Development and the Open Specification

The permanent canonical location of the Web Answer Citation Protocol is:

https://ericstrate.com/wacp/

Version numbers describe the specification, not the URL. Future revisions should therefore remain at the same canonical location instead of creating separate URLs for WACP 4.0, WACP 5.0 or later versions.

WACP is developed openly. Source documentation, implementation examples and the specification repository are available through the Web Answer Citation Protocol GitHub repository.

The repository should identify the currently published specification version and maintain historical versions through normal source-control history, releases or tags.