Free private workbench – No sign-up

Knowledge Base Content Audit Workbench

Turn an article inventory and explicit evidence into a reviewable action queue for keeping, reviewing, updating, merging, redirecting, or retiring content. Every recommendation shows why it appeared and what still needs human approval.

By Knowledge Base Software Editorial Team–Methodology KCAW-1.0–Updated August 4, 2026

Start the content audit

No black-box quality scoreRisk, demand, outdated signals, confidence, and effort stay separate.

Evidence before actionCandidate findings route to review; consequential changes are never executed automatically.

Private and portableFiles are processed in this browser. Save a project JSON, action CSV, redirect CSV, or PDF.

Step 1 of 4Set boundary

Your inventory remains in this tab unless you download a project JSON. Nothing is sent to an AI model or content platform. Use identifiers and aggregate metrics rather than confidential article text.

Define the audit boundary

Freeze the evidence date, measurement window, review-age rule, and demand thresholds. The workbench will preserve them with every export so another reviewer can reproduce the queue.

Audit name *

Evidence snapshot date *

Metric window start *

Metric window end *

Review-age threshold, days *Age is a review signal, never proof that content is wrong.

Evidence basis *

High-demand thresholds for this window

Qualified views

Failed searches

Assisted contacts

Negative feedback

Demand is normalized against your thresholds and reweighted across available metrics. A blank remains Unknown; zero means an observed zero.

Scope and exclusions

Boundary rule: this workbench audits one row per existing article. Uncovered searches and repeated tickets can inform a separate knowledge-gap prioritization process; they are not silently converted into a Create recommendation here.

Review the evidence rows

Inspect the imported inventory, correct evidence states, or add a row manually. Any edit invalidates the previous report until you rebuild it.

Articles0

Named owner0%

Source recorded0%

Review trigger0%

No article rows yet. Go back to import a CSV, load the example, or add an article manually.

Swipe or scroll horizontally to inspect the complete inventory table.

Content audit action queue

Audit queue ready

Urgent gates0Mandatory containment or correction evidence

Action-ready rows0Consequential action with corroborated or confirmed evidence

Blocked decisions0Missing target, retention, scope, or evidence prerequisite

Decision-input completeness0%Coverage of declared audit fields, not a content quality score

Keep0

Review0

Update0

Merge0

Redirect0

Retire0

Prioritized article queue

Ordering is deterministic: mandatory gate, risk tier, normalized demand, outdated-signal points, then Article ID. Evidence strength and effort remain separate from priority.

Find article

Action

Priority

Readiness

Swipe or scroll horizontally to inspect the full action queue.

All audit recommendations
Priority	Article	Action	Readiness	Demand / outdated	Why	Owner / effort

Deprecation and redirect validation

Merge, redirect, and retire recommendations are checked for missing targets, self-redirects, chains, loops, target-scope mismatches, and canonical conflicts. The workbench does not crawl or change any URL.

Swipe or scroll horizontally to inspect the complete transition plan.

Proposed URL transition plan
Source	Action	Proposed target	Validation	Acceptance checks

Governance and evidence coverage

Human decision required: the workbench produces a queue recommendation, not proof that an article is accurate, duplicated, obsolete, safe to remove, or compliant. Verify consequential findings with an accountable reviewer and approved source before publishing a change.

Next in the workflowTest whether the action queue fits the qualified review capacity available.

Plan governance capacity View all free tools

What the Content Audit Workbench decides

The workbench turns declared evidence into an owned queue. It does not let one traffic number, age field, or similarity label make every editorial decision.

Accuracy

Update verified defects

Confirmed or corroborated factual defects, broken steps, and source-version gaps can enter an Update queue. Candidate signals remain Review.

Uniqueness

Merge only with a destination

A declared non-primary duplicate needs a relevant final destination and scope checks before it is action-ready.

Lifecycle

Retire with retention controls

An obsolete task can become a retirement candidate only after evidence, retention, successor, and URL behavior remain visible.

Governance

Route unresolved evidence

Missing owners, sources, review triggers, and candidate findings become review work rather than silent assumptions.

Technical state

Expose routing conflicts

Status, canonical, redirect, indexability, and broken-asset inputs are checked without pretending the tool crawled the live site.

Restraint

Keep justified content

Low demand or old publication dates alone never prove that an article should be deleted or rewritten.

Transparent scoring and deterministic ordering

Scores help sort evidence; they do not replace the decision rules or become an invented overall quality grade.

Methodology KCAW-1.0
Dimension	Calculation	How it is used	What it cannot prove
Mandatory gate	Any candidate, corroborated, or confirmed harm / mandatory-correction signal.	Moves the row to the top for urgent containment or verification.	That the reported harm is real without source review.
Outdated-signal points	40 factual defect + 30 broken steps + 20 source gap + 5 stale screenshot + 5 overdue review.	Orders similarly risky rows and makes the contributing signals visible.	Article quality, accuracy, or the correct action by itself.
Normalized demand	Available metrics are capped against entered thresholds, weighted 20% views, 30% failed searches, 35% assisted contacts, and 15% negative feedback, then reweighted for blanks.	Sequences work among comparable risk and gate states.	Task success, prevented tickets, causation, or content value.
Evidence state	Candidate, Corroborated, Confirmed, Rejected, Unknown, or None for each finding.	Separates investigation from action-ready work.	Truth beyond the evidence and reviewer boundary entered.
Effort	User-entered estimated hours, kept outside priority.	Feeds capacity planning after the queue is accepted.	Actual delivery time, specialist availability, or approval delay.

Methodology basis: the accompanying knowledge base content audit guide separates outdated, duplicate, missing, demand, confidence, risk, effort, and action evidence. The workbench deliberately limits this version to existing-article decisions so a missing-content signal is not mistaken for a proven publishing opportunity.

Safe merge, redirect, and retirement planning

A copy decision is incomplete until readers, search, internal links, locales, and rollback are accounted for.

Before change

Confirm scope compatibility

Compare task, audience, locale, permissions, retention, and the final destination. Similar wording does not prove the same intent.

During change

Use a direct final route

Avoid self-redirects, loops, and chains. Preserve useful material before switching the source URL.

After change

Verify every surface

Test the status, destination, canonical, internal links, sitemap, search index, locale relationships, access behavior, monitoring, and rollback path.

Use the detailed knowledge base deprecation strategy before executing a retirement or redirect. This tool never deletes content, publishes redirects, or certifies recoverability.

Content Audit Workbench questions

Does the workbench crawl my knowledge base?

No. It analyzes the CSV or project JSON you provide in the browser. HTTP status, canonical, redirect, indexability, and demand values are declared evidence, not live crawl results.

Does an old last-verified date trigger an Update?

No. Age creates a Review signal. Update requires stronger evidence such as a corroborated or confirmed factual defect, broken procedure, source-version gap, or technical defect.

Can low-traffic articles be retired automatically?

No. Low demand does not prove that a task is obsolete, especially for rare but consequential procedures. Retirement requires task-obsolescence evidence, a retention decision, and human approval.

Why is there no overall content health score?

A single score can hide a critical defect behind strong traffic or governance coverage. This workbench keeps priority, demand, outdated signals, evidence confidence, effort, and data completeness separate.

Does the tool find duplicate articles with AI?

No. You supply authored duplicate candidates and evidence states. The workbench checks whether the proposed decision has a primary page, destination, compatible scope, and a safe redirect path.

How are blank metrics handled?

A blank is Unknown, not zero. Demand is calculated from available metrics and the weights are rebalanced. The report preserves which metrics were absent.

Where is my inventory stored?

It remains in the current browser tab. Use Save project JSON to keep a local reusable copy. The tool does not upload article rows to an AI service or content platform.

Can the report be used as an approval record?

It can support an approval record when you preserve the source evidence, scope, methodology version, reviewer, decision, acceptance test, and final outcome. The generated recommendation itself is not independent assurance.

Related knowledge base tools and guides

Published by the Knowledge Base Software Editorial Team. This vendor-neutral workbench contains no paid placement, product ranking, or guaranteed outcome.

Important: outputs are planning and editorial review aids. They do not replace qualified legal, compliance, security, accessibility, records-retention, technical, or subject-matter review.