CiteUnseen

Rubric

Citation readiness

Rubric v0.8 · 5 checks · 20 of 100 points

Can a passage from your site be quoted on its own?

This is the largest single block of points because it is the mechanism. Answer engines reuse passages that stand on their own; copy that needs the surrounding page to make sense is rarely quoted.

Last updated · Written by Chris Brock, who wrote the rubric and signs the audits that use it.

Why this dimension exists

The largest block of points in the rubric, twenty of a hundred, because this is how everything else becomes a citation.

Engines quote passages, not pages. A paragraph that needs the three above it cannot be lifted, however well the page ranks. The unit of work is the self-contained answer: a direct response in the first sentence, a claim that carries its own support, a number that says where it came from, and headings stable enough that the passage still has an address next month.

Long build-ups, deferred conclusions and pronouns pointing at earlier paragraphs are normal prose and costly here. The five checks do not ask you to write for machines instead of people. They ask you to put the answer first.

The short version, for somebody asking how to get quoted rather than how we score it, is <a href="../../get-cited/">on its own page</a>.

The guide for this dimension.

  • Getting cited: what decides whether a passage can be lifted, attributed and checked.

The 5 checks

2 of these 5 are deterministic, which means the free scan settles them without anybody reading the site.

CI-01 — Direct answer passages

5 points · Proposed by machine, rated by a person

What is tested. Sample priority customer questions and determine whether pages contain concise, self-contained answers.

What is recorded as evidence. Question, page, answer passage, surrounding context.

What fails it. Important answers buried in slogans; videos without transcripts; vague copy that requires inference.

How to fix it.

  1. Write down the questions a buyer actually asks, in their words.
  2. Give each one its own heading, phrased as the question.
  3. Answer it directly beneath in two to four sentences that make sense with no surrounding page.
  4. Put the answer first and the elaboration after it, not the other way round.

How to confirm it is fixed. Read one answer out of context. If it needs the paragraph above it, it is not a passage yet.

Reference.

CI-02 — Claim support

5 points · Proposed by machine, rated by a person

What is tested. Identify material quantitative, comparative, security, compliance, performance, and customer-outcome claims, and record for each whether its provenance is stated: an external source, or a first-person attestation naming the asserting party, the period, the scope, and the basis.

What is recorded as evidence. Claim text and its supporting source, methodology, case evidence, or disclosure, Where no external source exists, the attestation stated beside the claim and which of party, period, scope and basis it names.

What fails it. Unsourced percentages; unsupported superlatives; vague customer counts; claims contradicted elsewhere; a private operating figure asserted with no stated period, scope, or basis.

How to fix it.

  1. List every number, superlative and outcome claim on the site.
  2. For each, add the support next to the claim: the date, the source, the sample, or the method.
  3. Delete or soften anything you cannot support. An unsupported claim is a liability, not a weak asset.
  4. Where the support is a document, link it.

How to confirm it is fixed. Pick three claims at random and try to check them using only the page they sit on.

Reference.

CI-03 — Stable passage structure

4 points · Settled by machine, from the evidence

What is tested. Review headings, anchors, page titles, summary blocks, definitions, tables, and FAQ organization.

What is recorded as evidence. Representative sections and anchor behaviour.

What fails it. One monolithic text block; unstable generated URLs; misleading headings; content hidden in inaccessible widgets.

How to fix it.

  1. Give each page exactly one H1 that says what the page is.
  2. Use H2 and H3 in order, with no skipped levels, and make each heading describe the section beneath it.
  3. Break long sections so no passage runs past a few hundred words without a heading.
  4. Avoid headings that are only labels ('Overview', 'More') where a specific one would do.

How to confirm it is fixed. Read the headings alone, top to bottom. They should read as an outline of the page.

Reference.

CI-04 — Source attribution

3 points · Proposed by machine, rated by a person

What is tested. Inspect references to standards, research, statistics, quotations, and third-party assertions.

What is recorded as evidence. Citation links, titles, dates, source proximity.

What fails it. Unattributed research; circular references; dead sources; citations that do not support the nearby claim.

How to fix it.

  1. Attribute every external fact to its source in the sentence that uses it.
  2. Link the source. A named source with no link is half an attribution.
  3. Say when the source is from, so a reader can tell how old the fact is.
  4. Distinguish your own findings from other people's, explicitly.

How to confirm it is fixed. Take one external fact and try to reach its source in one click.

Reference.

CI-05 — Content freshness signals

3 points · Settled by machine, from the evidence

What is tested. Inspect published dates, modified dates, version history, release notes, and material-update practices.

What is recorded as evidence. Visible dates, structured dates, contradictions.

What fails it. Time-sensitive content without dates; stale product instructions; structured dates inconsistent with visible dates.

How to fix it.

  1. Publish a real date on content where recency matters, and update it only when the content changes.
  2. Mark it up with <time datetime> and with datePublished and dateModified in structured data.
  3. Say what changed, where a substantive revision is worth a note.
  4. Remove auto-updating dates that move without the content moving. They are noise and readers learn to ignore them.

How to confirm it is fixed. Compare the displayed date, the <time> attribute and the structured data on one page.

Reference.