Skip to content
Semantec SEO
WorkAuditMIRENAFor agenciesLearningAboutContact
Start a project

Information Gain

Missing Proof Type Audit: Match the Evidence to the Claim

A missing proof type audit checks if the evidence can support the kind and strength of claim a page makes. A respected source can be relevant yet still leave a claim about identity, functions, comparison, causality, experience, policy, compatibility or results unproved.

Compare claims with their proof needs

By Kevin Maguire

Stacked reference books beside a small lock opener and lock examined through a magnifying glass.
Evidence must fit the specific claim; a large reference collection alone does not establish it.
Source notebook and version folders beside a pencil, magnifier and distinct record cards.
Inspect the source, method, date and scope before accepting an assertion.

Start with what the sentence asks someone to believe

Do not begin by counting citations. Read the sentence under review as a claim, identify its class, and ask what kind of observation, record or test would support its assertion. The source’s authority and the evidence’s fit are separate questions.

An editorial evidence fit matrix, not a universal proof hierarchy
Claim classEvidence to seekLimit the wording to what it shows
Identity or existenceA controlling official record, direct source or observation tied to the exact entity and relevant date.A similar name or an old record may not establish the present identity. Separate existence from current status.
Attribute, function or processA specification, documented method or observed test with the relevant model, version and conditions.A documented feature does not prove successful operation in every environment.
ComparisonComparable measurements or an independent assessment using a stated method, sample and criteria.Do not compare different dates, tasks or populations as if they were equivalent. State the dimensions being compared.
CauseA study or analysis designed to distinguish the proposed cause from plausible alternatives.Temporal order and correlation alone do not establish causality. Describe the design and remaining attribution limits.
ExperienceDirect observation or an attributable account with role, context, date and access to the event.One account supports a bounded experience claim; it does not establish a typical outcome for every reader.
PolicyThe responsible authority’s current policy text with relevant jurisdiction, scope and effective date.A summary may omit an exception. Separate the written rule from an interpretation or an observed enforcement decision.
CompatibilityA provider statement or a test that identifies both components, versions, environment and tested operation.Documentation of intended support and evidence of tested behavior are different claims. Neither covers untested combinations.
Outcome or relationshipA relevant dataset, observed result, record or expert assessment with a method suited to the specific relationship.A result for one sample does not establish a guarantee. A connection between entities does not by itself show its direction or cause.

A single sentence can contain more than one claim. “This certified system works with every device and increases revenue” combines identity or status, compatibility and an outcome. Split it into assertions before assessing the proof. One source may support only one of them.

Worked example: did the rewrite cause the result?

A page says that a content rewrite caused higher conversions. Its only evidence is a screenshot from before and after the change, with no record of other changes. The missing proof concerns attribution, not the absence of another citation.

Open educational book beside a contrasting ceramic cube and magnifying glass.
A contrasting case can expose an unsupported inference from an observation.

The screenshot may show a reported difference between two periods. It does not isolate the rewrite from changes in audience, campaign spend, seasonality, pricing, tracking or product availability. Even the observation needs dates, a defined conversion event, comparable denominators and a check that measurement stayed consistent.

Choose wording at the level the evidence can support
Proposed sentenceEvidence gap or limitResponsible next step
“The rewrite caused higher conversions.”The available periods do not isolate the rewrite from concurrent changes or sampling variation.Seek a suitable experiment or stronger attribution design. Describe its limits; do not promise that one design removes all uncertainty.
“Recorded conversion rate was higher in the later period.”A screenshot needs its date range, rate definition, denominator and tracking context.Verify the underlying data and state the observation with those limits. Do not imply the rewrite explains the change.
“The rewrite was intended to improve the purchase decision.”Intent is not a measured commercial result. The text should identify the editorial change it made.Describe the revised decision rule or evidence placement and label the purpose as intent.
No reliable measurement survives reviewNeither the outcome nor its attribution can be supported from the available record.Remove the performance assertion or defer it until evidence can be recovered. Keep any supported description of the work separate.

An independent source is not automatically causal evidence. An expert can assess a proposed mechanism without proving that it produced this particular result. Stronger evidence means better fit to this question, not simply a more impressive source label. The appropriate study design depends on the claim, practical constraints and consequence of error.

An open folio holding source notes and page instructions beside an owner badge and checklist.
Keep the proof plan and permitted wording together so the next editor preserves the boundary.

Build a claim level proof plan

Use the audit when a page contains sources but makes an assertion stronger, broader or different from what those sources establish. Work from the exact claim to the missing evidence, then decide how the page should change.

  1. Classify the assertion. Record the sentence, entity and reader decision. Identify identity, attribute, relationship, process, comparison, cause, policy, compatibility or result; split mixed claims.
  2. Inspect the current evidence. Record source, method, date, sample, independence and scope. Read the controlling passage and its limitations rather than relying on a title or search snippet.
  3. Compare scope and strength. Check that the claim refers to the same version, market, population, time and measured outcome. Distinguish reported observations from interpretations and intended functions.
  4. Name the missing proof. Specify the record, observation, controlled test, independent comparison, expert review, policy text or dataset required. Explain why that evidence would address the gap.
  5. Choose a gap action. Acquire the evidence, qualify the sentence, replace it with a supported claim, remove it, defer the decision or route it to an accountable specialist.
  6. Read the complete answer again. Check that the direct answer, terms, evidence, exception and onward destination no longer imply the removed or narrowed claim. A cautious body paragraph cannot repair an unsupported headline.

Prioritize claims that control a choice, carry a large consequence or recur across many pages. A minor descriptive detail and a strong causal recommendation should not receive the same review effort simply because both have one citation. Place the evidence or qualification where it changes the reader’s interpretation.

Record the evidence boundary, not just a source link

The useful output is a proof plan for each affected claim, including accepted sources, remaining gaps and permitted wording. A bare bibliography leaves the next editor to reconstruct the reasoning.

  • Accepted support. Keep the exact claim, source passage, proof class, method, sample, date and independence together. Explain which part of the assertion it supports.
  • Scope and uncertainty. State the relevant market, version, conditions and limitations. Preserve conflicting evidence instead of combining incompatible findings into a confident sentence.
  • Gap action and wording limit. Name the evidence still needed, the interim wording allowed and any action that must wait. An unresolved record is not an approval.
  • Owner and review trigger. Name the person responsible, accepted or deferred state, next destination and a trigger such as a changed policy, new version, corrected dataset or stronger study.

When no feasible proof can support the desired claim, change the claim. Do not label a weak source as a controlled test, turn a provider’s documentation into an independent comparison, or turn the presence of two things together into a directed relationship. The audit is allowed to end with removal rather than a longer answer.

Useful information gain can be a new evidence class, a precise limitation or a supported distinction. Added breadth, rearranged competitor headings and a new URL do not establish that contribution. Put the proof boundary in the existing answer or component when that is where the reader needs it.

Questions about evidence fit

Can a reputable source be insufficient?

Yes. Reputation concerns the source; fit concerns the claim. A reliable specification can support a documented function without proving a particular customer outcome. Read the scope and method needed for the exact assertion.

Must every claim use independent research?

No. An official record may be the right source for identity or policy, while direct observation can support a bounded experience. Independence becomes relevant when incentives or the claim type demand it. Record the source relationship rather than treating one evidence class as best for every question.

What if sources conflict?

Compare definitions, methods, dates, populations and versions. The disagreement may reflect different scopes. Keep the unresolved boundary visible, reduce the wording or defer the affected decision instead of choosing the source that favors the desired assertion.

How does this differ from a relationship audit?

This audit asks if the evidence can support the assertion. A relationship audit asks which connection between entities or ideas must be stated, in which direction and where. A missing relationship may reveal a proof gap, but writing the connection does not supply its evidence.

Related decisions and sources

Information Gain

Choose a useful contribution with evidence appropriate to the reader’s decision.

Missing Limitation and Risk Audit

Place material constraints and consequences beside the action.

Missing Relationship Audit

Name and support the connection that a page leaves implied.

Google: creating helpful content

Use the public editorial self assessment questions as a broader reference.

The examples are teaching scenarios, not records of a client project. The illustrations explain record separation and revision; they are not evidence of the assertions described. These editorial methods do not guarantee rankings, indexing, traffic, rich results, citations or AI feature inclusion.

Semantec SEO

Semantic SEO agency
All English speaking markets

support@semantecseo.com
Discuss an audit

Services

  • Website audits
  • Topical maps
  • Writing instructions
  • Writing & rewriting
  • Internal linking
  • Entity analysis
  • Structured data
  • Page design
  • Location content
  • Measurement

Learning

  • Semantic SEO
  • Reader intent
  • Entities and facts
  • Topical mapping
  • Semantic writing
  • Evidence & sources
  • Internal links
  • Structured data

Company

  • About us
  • For agencies
  • Project pricing
  • Contact us

Policies and trust

  • Legal info
  • Terms
  • Privacy
  • SEO & AI disclaimer
  • Citation policy
  • Claims policy
  • Security

Founded by Kevin Maguire. Kevin Maguire on LinkedIn.

AI supports our internal analysis, writing and review. Search rankings and AI citations are external outcomes we cannot promise.

© 2026 Semantec SEO · HTML sitemap