Turn an on-page SEO audit into a verified remediation plan, not a list of tool warnings

An on-page SEO audit should start with a reproducible inventory of URLs and observed signals, not with a sitewide score. The purpose of this stage is to establish what exists, what each page is intended to do, and which findings can be verified from crawl data, Search Console, analytics, or the rendered page.

Evidence required: export the crawlable URL set, response status, index directives, canonical targets, page titles, descriptions, primary headings, internal-link counts, and the relevant Search Console or analytics fields used for prioritization. Keep the raw export so another reviewer can reproduce the finding.

Severity: assign severity from the consequence of the issue. A blocked or canonicalized-away priority page is more severe than a cosmetic metadata inconsistency. Do not use current traffic as the only severity signal because a page can have low traffic precisely because it is not eligible to perform.

Owner: name the person or team that can change the underlying source, such as content, engineering, product, or SEO. Avoid assigning every issue to SEO when the fix lives in a template, CMS rule, or deployment process.

Corrective action: define the desired end state at the URL level. Examples include restoring index eligibility, aligning a canonical with the intended destination, rewriting a misleading title, improving the content brief, or adding a contextual internal link from a relevant source page.

Validation: rerun the same check after deployment and compare the new output with the baseline. The source historically treated pages around positions 5-20 as a useful review segment because they already showed search visibility; use that range as a prioritization example, not as a universal promise of faster gains.

At the end of this stage, every issue should have a URL, evidence, severity, owner, corrective action, and validation method. If one of those fields is missing, the finding is not ready for implementation.

This stage checks page-level signals that can usually be observed directly in the HTML, rendered page, or crawl export. The goal is not to make every page look identical; it is to identify signals that are missing, contradictory, duplicated without intent, or inconsistent with the page's search purpose.

Title tags

Evidence required: capture the current title, target query or page purpose, search impressions where available, and whether the title is missing, duplicated, or visibly truncated in the audit tool. The source used 60 characters as a historical screening threshold; treat it as a review cue rather than a fixed search-engine limit.

Severity: high when the title is absent or clearly misstates a strategically important page; medium when duplication creates ambiguity across similar pages; low when the title is merely stylistically imperfect.

Owner: content or SEO for copy changes, with engineering ownership when a template generates the title incorrectly.

Corrective action: write a concise title that accurately represents the page and its search intent. Do not force exact-match repetition or change a strong-performing title without a documented reason.

Validation: confirm the deployed HTML contains the intended title, then monitor search-result presentation and click behavior as observational follow-up rather than proof of guaranteed improvement.

Meta descriptions

Evidence required: record whether a description exists, whether it is duplicated, and whether the page receives enough impressions for snippet testing to be meaningful.

Severity: usually medium or low because Google can rewrite snippets, but raise priority when the current description is inaccurate, missing on a high-visibility page, or generated incorrectly at scale.

Owner: content or SEO for copy, engineering for template defects.

Corrective action: provide a truthful summary that aligns with the page and gives users a clear reason to consider the result without making unsupported claims.

Validation: verify the live description in source or rendered output and record later snippet behavior separately from implementation success.

Heading structure

Evidence required: capture the rendered heading outline and note whether the page has no H1, multiple H1s, or a hierarchy that jumps from H1 to H3 without a relevant H2. These are structural review signals, not automatic ranking failures.

Severity: medium when headings make the page hard to understand or reveal a template defect; low when the outline is unconventional but still clear and accessible.

Owner: content for editorial structure, engineering for template-generated headings.

Corrective action: use headings to reflect the content hierarchy and user questions. Do not add headings solely to satisfy a crawler.

Validation: inspect the rendered outline after publication and confirm the primary heading represents the page purpose.

Canonical tags

Evidence required: compare the declared canonical, final response URL, indexability, internal links, and intended search destination. A high-value page whose canonical points elsewhere needs immediate investigation.

Severity: critical when the canonical contradicts the intended indexable URL; high when template rules create broad mismatches; lower when the canonical is already aligned and the crawler warning is cosmetic.

Owner: engineering or platform owners for generated canonicals, with SEO defining the intended destination.

Corrective action: align the canonical with the page that should represent the content, and correct any template or CMS rule that generated the conflict.

Validation: recrawl the page and confirm the live canonical, response status, internal links, and index directives now agree. Escalate a critical canonical issue as a P1 implementation defect only when the site actually uses that severity convention.

Content auditing requires judgment because mechanical thresholds do not tell you whether a page satisfies search intent. Use tool flags to identify candidates, then compare the page with the queries it receives, the information users appear to need, and the competing results that define the current search landscape.

Evidence required: gather the page's target query set, impressions, clicks, ranking pages, content outline, freshness indicators, and the gaps between the page and relevant competing results. When using competitors for comparison, record the specific missing topics or intent differences rather than copying their length.

Severity: high when a priority page is materially misaligned with the query intent or omits information necessary to answer the user; medium when coverage is incomplete but the core purpose is correct; low when differences are cosmetic.

Owner: content or subject-matter owners, with SEO supplying query and SERP evidence.

Corrective action: revise the brief and page around the information the target audience needs, preserving accurate existing material and adding only evidence-backed gaps.

Validation: review the published page against the updated brief, confirm crawlability and indexability, and monitor search behavior later without treating movement as guaranteed.

Thin-content flags

The source used 300-500 words as a historical tool threshold and contrasted examples near 150 words and 200 words. Keep those figures as screening examples only. A short contact page can be complete, while a longer informational page can still be thin if it fails to answer the query.

Evidence required: document page purpose, intent, substantive topics covered, duplication level, and whether the page contributes unique value.

Corrective action: expand, consolidate, redirect, noindex, or leave the page unchanged according to its purpose. Do not add filler to reach a word target.

Validation: confirm the chosen action is live and that the resulting page remains accurate and useful.

Keyword and entity alignment

Evidence required: check whether the title, primary heading, introductory copy, and major subtopics reflect the language users actually search. The source referenced an H1 and the first 100 words as review locations; use them as editorial checkpoints, not a placement formula.

Corrective action: revise wording where the page uses ambiguous, outdated, or market-misaligned terminology. Avoid stuffing repeated exact-match phrases.

Validation: inspect the live copy and confirm the main topic is clear to a reader without relying on hidden metadata.

Content decay

The source used a 12-18 month historical window to describe pages that once performed well and later declined. Treat that as a review example, not a decay schedule.

Evidence required: compare current and prior Search Console impressions, clicks, query mix, page changes, competitor changes, and content freshness.

Corrective action: update only the sections that are outdated or incomplete, preserve useful material, and fix technical issues if the decline is not primarily editorial.

Validation: verify the revised page, document the publication date, and monitor subsequent search data as evidence of response rather than as a promised outcome.

Internal links should be audited as user paths and crawl paths, not as a generic quantity. The key questions are whether important pages can be discovered from relevant sources, whether anchor text makes the destination understandable, and whether the architecture leaves priority pages unnecessarily isolated.

Orphan and weakly linked pages

Evidence required: compare crawl discoveries with sitemap and analytics URL lists, then identify pages with no meaningful internal links from crawlable content.

Severity: high when a strategically important page is isolated; lower when the page is intentionally excluded from navigation or search.

Owner: content, information architecture, or engineering depending on where the link should be added.

Corrective action: add contextually relevant links from pages that naturally support the user's next step. Do not create arbitrary links solely to increase counts.

Validation: recrawl and confirm the destination is discoverable from the intended source pages.

Click depth and navigation paths

Evidence required: record the shortest internal path from important entry pages and identify whether a deep path reflects a genuine hierarchy or accidental neglect.

Severity: medium when depth makes a priority destination difficult for users or crawlers to reach; low when deeper placement matches the site's information architecture.

Corrective action: surface the destination from an appropriate hub, category, breadcrumb, or contextual link when that improves navigation.

Validation: confirm the revised path in a fresh crawl and test the user journey manually.

Anchor context

Evidence required: export source URL, destination URL, anchor text, and nearby sentence context. Generic anchors are not automatically harmful, but they may be less descriptive than alternatives.

Corrective action: rewrite anchors when a clearer phrase helps users understand the destination. Avoid forcing keyword-rich anchors everywhere.

Validation: inspect the live source page and ensure the anchor remains natural in context.

Links from authoritative or high-visibility pages

Evidence required: identify pages that already attract external links, organic entrances, or strong internal visibility, then check whether relevant downstream pages are logically connected.

Corrective action: add a contextual link only when it improves navigation or topic continuity. Do not add links because a proprietary authority metric is high by itself.

Validation: recrawl to confirm the edge exists and monitor destination discovery separately from ranking effects.

The source suggested that a focused internal-link review can take roughly 30-60 minutes on some sites. Preserve that as an internal operating example, not a service-level expectation; site size, crawl quality, and architecture can change the effort substantially.

After the diagnostic stages, convert findings into a remediation queue. A tool may surface a 404 response on an irrelevant retired URL while a live priority page has a canonical conflict; issue count alone cannot tell you which deserves attention first.

Evidence required: for each issue, record the affected URL, observable problem, intended state, business or search context, and whether the finding is isolated or template-wide.

Severity: use consequence-based labels such as critical, high, medium, and low. Keep severity independent from effort so a difficult critical fix is not downgraded merely because it takes longer.

Owner: assign the team that controls the source of the defect and name a reviewer who can validate closure. This prevents issues from circulating between SEO, content, and engineering without a decision.

Corrective action: write the change as a testable end state. A title edit, canonical correction, internal-link addition, or redirect change should be specific enough that another person can confirm whether it was implemented.

Validation: run the same crawl or inspection used to create the finding after deployment. A small metadata edit may take about 10 minutes in one workflow, but effort varies by platform. For audit governance, the source used a 30-day follow-up crawl as an operating example; schedule validation based on deployment timing and data availability rather than treating that interval as mandatory.

Keep implementation validation separate from outcome monitoring. A fix can be correctly deployed even if rankings do not change, and rankings can change for reasons unrelated to the fix. Close the technical issue when the expected state is verified, then track search performance as a separate observation.

A tool-led audit works well when the site architecture is understandable, the team can reproduce findings, and owners are available to implement changes. Specialist review becomes more valuable when issue patterns are broad, the site has complex rendering or faceted navigation, traffic changed without an obvious cause, or earlier remediation failed because the wrong problem was prioritized.

Evidence required: before escalating, collect the crawl export, Search Console observations, analytics trend, deployment history, template rules, and examples of affected URLs. The more specific the evidence, the easier it is to determine whether the problem is on-page, technical, content-related, or external to the audit scope.

Severity: escalation is high when indexing, canonical, rendering, or template defects affect important groups of pages; medium when the team mainly needs prioritization support; low when the issue is isolated and the corrective action is already known.

Owner: keep internal ownership even when a specialist is involved. External review can diagnose and recommend, but the organization still needs someone accountable for implementation and validation.

Corrective action: use specialist time to resolve ambiguity, trace systemic causes, review migrations, or design a testable remediation sequence. Avoid paying for another generic issue list when the missing need is interpretation.

Validation: require the same closure standard as an in-house audit: evidence that the source defect changed, a fresh crawl or inspection confirming the new state, and a record of any remaining risks.

The source used sites above 500 pages as an example where pattern analysis can become more complex and sites under 200 pages as an example where a capable in-house owner may be able to complete much of the work. Treat those figures as historical scoping heuristics, not thresholds for whether specialist help is necessary.

If the team can explain the issue, produce the evidence, name the owner, implement the correction, and verify the result, a tool-led workflow may be sufficient. If the team cannot determine which findings are causal, systemic, or worth fixing, specialist review can help resolve that diagnostic gap.