AQA Guide: Evidence, FAQ Maintenance and Practical Limits

A useful agency FAQ answers a real buyer's question and lets a reviewer check the answer. AQA, short for AI Question Answer, is an open specification published by Davy Abderrahman of AI Labs Solutions. It proposes a convention for recording provenance and maintenance information around Q&A content. It does not demonstrate that a search engine adopted the convention, or guarantee a citation.

Start with the answer and its evidence

Choose a question that matters to the client's actual offer. Assign someone who can verify the response. Link the supporting source, record when it was checked and identify conditions that would make the answer obsolete. A date should describe a real review or change; moving the displayed date forward without checking the claim adds no evidence.

For an agency, an answer register can be a compact table with these columns: stable question ID, language, public question, answer, evidence URL, reviewer, review date and next review trigger. Keep unpublished customer details out of that register's public version. The stable ID should remain the same when a heading or translated slug changes.

What the published AQA convention adds

The AQA publisher's site describes a JSON-LD context that extends existing types such as Article, FAQPage, Question and Answer. Its Basic level records dates, a source citation, an author and a language; deeper levels add maintenance and revision details. These are declarations made by the publisher. A validator can inspect a declaration's structure, but it cannot establish that a cited source supports the answer or that a review actually happened.

Check the exact version used in a project. On 16 November 2026, the website displayed v1.2.2-draft, while the linked specification document's header still said 1.2.0-draft. That mismatch is a reason to pin the reviewed document and context, and to record the implementation's scope. It is not a basis for claiming independent certification.

A small example an agency can audit

Suppose a public software FAQ asks, "Does the report describe the models actually tested?" A defensible answer describes the report's verified contents and links a public example. Before adding any markup, check the example, its date and the exact wording. The following is a fictional content record, not evidence about a customer's report:

question_id: report-model-provenance
language: en
question: Does the report describe the models actually tested?
answer: Check the model and mode identifiers in the attached example.
evidence_url: [replace with the reviewed public example]
reviewer: [person responsible for checking this answer]
reviewed_on: [date the source was actually checked]
review_trigger: a change to the report format or model selection

Replace every placeholder before publication. If the evidence is private, obtain permission to publish it or use a genuinely public demonstration. Do not turn a demonstration into a customer success story. If the format cannot support the claim, correct the answer instead of decorating it with metadata.

When implementing the published convention, follow its versioned specification and validate the resulting document against the reviewed version. Keep the questions and answers visible on the page and consistent with the markup. Extension properties should not be described as universally understood by third-party consumers.

Keep specification conformance separate from search results

Google says that generative search does not require special structured data or additional AI files. It also documents that FAQ rich results are no longer shown in Google Search; this does not prevent a useful FAQ from serving readers. Read the official AI search guidance and the Search documentation changelog. Neither an AQA declaration nor a local conformance score is an official Google metric.

An internal content-quality check, successful retrieval, an explicit citation and a commercial conversion answer different questions. Report them separately. An assistant answering a prompt that already names a brand is not the same experiment as an unprompted recommendation of that brand.

A practical acceptance test before delivery

  1. Open the public page in its explicit language with a conflicting browser preference. Confirm that the intended answer, source and language are still served.
  2. Compare the visible answer with its structured representation, including dates and attribution. Parse JSON locally without executing scripts or fetching arbitrary URLs supplied by the document.
  3. Check each cited source manually and record what it supports. Reject fabricated references, unverifiable reviewers and copied testimonials.
  4. Save a baseline for the actual prompts and surfaces being measured. After the approved update, compare equivalent observations and record uncertainty rather than promising a ranking lift.

AI Labs Audit is an infrastructure for agencies and consultants using their own scenarios and expertise. Use its measurement methodology to explain what was tested. AQA can be evaluated as a publishing convention within that workflow; its usefulness and any visibility effect must be demonstrated independently.

Editorial review: 16 November 2026. AQA is published by the founder of AI Labs Audit, the publisher of this guide. This commercial and authorship relationship is disclosed so readers can assess the advice. The worked record is illustrative and makes no claim of engine endorsement or measured client gains.

Questions and answers

How long does it take to implement AQA Basic?

There is no fixed implementation time. Work depends on the existing FAQ, the reliability of its dates and sources, and validation. Use actual verification dates and identified sources; do not invent freshness to satisfy a format.

Is AQA automatically valid Schema.org markup?

AQA is a publisher-defined specification. Validate the underlying Schema.org markup and any extension separately; an ignored property does not prove that every document is valid or that a search engine supports AQA. Publication does not establish engine adoption or ranking effects.

Does AQA guarantee citations in AI answers?

No citation or ranking guarantee is demonstrated. AQA describes a way to structure FAQ information. Useful answers, reliable sources and accurate dates can be evaluated independently, while any internal readiness score must be identified as an internal metric with its scope and limits.

About the author

Davy Abderrahman

Founder of

As a specialist in visibility in AI answers (AEO/GEO/LLMO), I help agencies and consultants measure and analyze their clients’ presence on ChatGPT, Claude, Gemini, Perplexity and other AI answer engines.

AEO GEO LLMO AI visibility AI audits
How do your brands appear in AI answers?

Agencies, consultants and in-house teams: compare the answers generated for your own scenarios, languages and selected models. Our infrastructure, your scenarios, your expertise.

Was this article helpful?

- (0 votes)