Content Audit Template: Prune, Refresh, or Keep

Most content audits never ship the action queue. The auditor pulls the data, scores 200 pages, builds a beautiful spreadsheet, presents to stakeholders, and then the company writes new posts instead of fixing the old ones because new posts feel more productive than repairs. The result is sites carrying 30 to 60 percent of their indexed pages as quiet underperformers that nobody touches. The 2026 algorithm environment is unkind to that pattern. After the March 2026 core update, sites with too many thin or dated pages saw section-wide demotions even on their healthy pages, because Google's quality assessment runs at the section level as well as the page level. The audit is no longer optional. The audit that actually ships is the one with a small decision framework and a queue that someone is accountable for clearing.

Quick Answer: A content audit that actually ships uses a five-way decision matrix (keep, update, merge, prune, redirect) and a 100-point scoring rubric across five signals (EEAT, freshness, engagement, technical health, search performance). Pull page-level data from Search Console and Analytics, score each page in 60 to 120 seconds, and route into one of the five buckets. The action queue then gets cleared on a quarterly cadence with the highest-impact pages first. Sites that run this audit quarterly and clear at least 80 percent of the action queue outperform sites that publish more new content without auditing.

Why Content Audits Are Repair Plans, Not Writing Exercises

The mental shift that makes content audits work is treating them as inventory triage instead of editorial review. The auditor's job is not to read every page and assess literary merit. The auditor's job is to look at the data, decide what each page should become, and produce a queue that someone can clear. The reading and assessing happens during the fix work, not during the audit.

The reframe matters because content audits that try to do too much never finish. A 500-page site audit takes weeks if the auditor reads each page in detail. The same audit takes one to two days when the auditor scores pages from the data first and only reads the pages that fail the data scoring. The data-first approach gets you the same action queue with 80 percent less effort.

The other reframe is that audits should produce decisions, not recommendations. "Consider updating this page" is not a decision. "Update this page with new stats and 600 additional words by Friday" is a decision. The audit's deliverable is a queue of decisions with owners and deadlines, not a document of observations.

The Five Decisions, Keep, Update, Merge, Prune, Redirect

Every page on the site falls into one of five buckets. The decision matrix is small enough to apply quickly and big enough to cover real cases.

Keep. The page is performing, the content is fresh, and there is no meaningful improvement available within reasonable effort. Most sites have 20 to 40 percent of pages in this bucket.

Update. The page has good bones but specific elements are out of date or missing. The fix is typically 1 to 4 hours per page (refresh stats, add new sections, improve internal linking, regenerate hero image). Most sites have 30 to 50 percent of pages in this bucket.

Merge. The page covers a topic that another page on the site covers better, or two pages cover overlapping topics and should be combined into one stronger page. The fix is consolidate content from both pages into one, 301 the loser to the survivor, and update internal links. Most sites have 5 to 15 percent of pages in this bucket.

Prune. The page is not worth keeping. The topic is no longer relevant, the content is too thin to improve cost-effectively, or the page is dragging site quality down. The fix is either delete with a 410 status or apply noindex. Most sites have 10 to 25 percent of pages in this bucket.

Redirect. The page is not worth keeping in its current location but the URL has accumulated links or referrals worth preserving. The fix is to 301 redirect to the most relevant existing page on the site. Most sites have 5 to 15 percent of pages in this bucket.

The percentages above are not targets. They are typical ranges. A site with 70 percent of pages in Keep is probably skipping the hard cases. A site with 50 percent of pages in Prune probably went too aggressive and is throwing away pages with recovery potential.

The 100-Point Scoring Rubric Across Five Signals

The scoring rubric below is the one that produces decisions reliably in 60 to 120 seconds per page. Each of the five signals scores 0 to 20 points. Total scores route into decisions according to a small lookup table.

EEAT signal (0 to 20 points):

• 0 to 5: No byline, no author bio, no first-hand markers
• 6 to 10: Generic byline, generic bio, some first-hand markers
• 11 to 15: Authored, partial bio, multiple first-hand markers
• 16 to 20: Full author signal, verified bio with schema and sameAs links, strong first-hand markers throughout

Freshness signal (0 to 20 points):

• 0 to 5: Last updated more than 24 months ago, contains explicitly dated content (e.g., "in 2022") that contradicts the current year
• 6 to 10: Last updated 12 to 24 months ago, contains some dated content that should be refreshed
• 11 to 15: Last updated 6 to 12 months ago, content is mostly current
• 16 to 20: Updated within 6 months, all stats and screenshots are current

Engagement signal (0 to 20 points):

• 0 to 5: Bounce rate above 80 percent, average session under 30 seconds, no conversion proxies
• 6 to 10: Bounce rate 60 to 80 percent, session 30 to 90 seconds
• 11 to 15: Bounce rate 40 to 60 percent, session 90 to 180 seconds, some conversion proxy
• 16 to 20: Bounce rate under 40 percent, session over 180 seconds, multiple conversion proxies

Technical health signal (0 to 20 points):

• 0 to 5: Page has indexing issue, broken images, or Core Web Vitals fail
• 6 to 10: Page indexed but slow, poor mobile experience, or missing key schema
• 11 to 15: Page is technically healthy with minor improvements available
• 16 to 20: Fully optimized with schema, fast load, clean mobile experience

Search performance signal (0 to 20 points):

• 0 to 5: Zero or near-zero impressions and clicks in last 90 days
• 6 to 10: Some impressions but low CTR, ranking outside top 30 for primary keyword
• 11 to 15: Ranking 10 to 30 for primary keyword, moderate CTR
• 16 to 20: Ranking top 10 for primary keyword, healthy CTR

The decision lookup:

• Total 80 to 100: Keep
• Total 60 to 79: Update
• Total 40 to 59 and another similar page exists: Merge
• Total 40 to 59 and topic still worth covering: Update with major rewrite
• Total 20 to 39: Prune or redirect
• Total 0 to 19: Prune

The lookup is heuristic. Judgment overrides the score in obvious edge cases (a page that scores 95 but covers a topic the business no longer cares about probably gets pruned or redirected anyway).

Pulling Page Data From Search Console and Analytics

The data the rubric needs sits in two places. Search Console gives you impressions, clicks, CTR, average position, and which queries each page ranks for. Analytics gives you sessions, bounce rate, average session duration, scroll depth, and conversion proxies.

The export workflow:

1. Search Console: Performance report, last 90 days, group by Page. Export to spreadsheet. This gives one row per page with the search performance signal data.
2. Analytics: Pages and screens report, last 90 days. Export to spreadsheet. This gives one row per page with the engagement signal data.
3. Site crawler (Screaming Frog or similar): Export the full URL set with technical health columns (status code, indexability, word count, internal link count, schema present, Core Web Vitals scores).
4. Merge the three exports on URL into a master sheet.

The master sheet then has one row per indexed page with all the signals visible. Scoring is faster from a dense data view than from individual page reviews.

For sites without analytics or with limited Search Console history, the audit can still run on the technical and search-performance columns from the site crawl and Search Console alone. The EEAT and freshness scoring shifts to a quick page-level review (60 seconds per page) for the borderline cases.

EEAT Scoring, the Experience and Trust Axis

EEAT scoring in 2026 leans heavily on the Experience signal because that is what the March 2026 update enforced most aggressively. The fast scoring checklist:

• Is the author named with a real human byline (not "admin", "team", "editor")
• Does the byline link to a bio page with photo, role, and credentials
• Does the bio link to verifiable external profiles (LinkedIn, GitHub, ORCID, professional website)
• Does the article contain first-hand markers (original screenshots, personal observations, "we tested", "in my testing")
• Does the article contain original data, original images, or original quotes
• Does the article cite external sources with links
• Is there author schema in the JSON-LD

The Experience markers are the section that most existing content fails. A 2022 blog post about a tool might have an author byline and a bio, but lack any "I tested this" language and use only stock screenshots. The fix during the update phase is to add a specific testing observation, swap one stock screenshot for a real one, and reference one concrete outcome.

The companion guide on E-E-A-T 2026 ranking signals covers the broader strategy that pairs with content audit work.

Freshness Scoring, What Stats and Screenshots to Swap

Freshness is the easiest signal to improve and one of the highest-leverage. The fast scoring checklist:

• Does the article reference a specific year (2022, 2023) that is no longer current
• Are any of the cited statistics from more than 24 months ago
• Are any screenshots from product UIs that have visibly changed
• Does the article reference deprecated features, pricing, or company names
• Is the publishedDate or modifiedDate older than 12 months

The fix during the update phase:

• Replace dated year references with current year or undated phrasing
• Update statistics with the most recent equivalent (or remove if no current data exists)
• Regenerate screenshots from the current UI (Apple's Preview, Cleanshot, or similar)
• Update product names, pricing, and company references
• Update both publishDate and modifyDate in the frontmatter (or only modifyDate if the article was substantially the same with updates)

Refresh cadence depends on topic volatility. SaaS reviews need refresh every 6 to 12 months because pricing and features change frequently. Evergreen explainers need refresh every 18 to 24 months. News and trend pieces should be considered for archive (pruned with redirect) after 24 to 36 months.

Engagement Scoring, Dwell Time, Scroll Depth, Conversions

Engagement signals tell you whether users find the page useful when they land. The fast scoring checklist:

• Bounce rate under 60 percent (varies by page type, but the 60 percent threshold is a useful default)
• Average session duration above 90 seconds (under 30 seconds usually means the page did not deliver on the title promise)
• Scroll depth above 50 percent on at least 50 percent of sessions
• Conversion proxy events firing (newsletter signups, internal link clicks, time on page above three minutes)

The engagement signal is noisier than the others because page intent affects what "good" looks like. A definitional explainer page might have a 90-second session and that is fine because the page answered the question quickly. A long-form guide should target three plus minutes. Use the page-type peer comparison rather than absolute thresholds when possible.

The fixes during the update phase if engagement is the weak signal:

• Move the answer earlier in the article (Quick Answer box in the first 200 words)
• Add Key Takeaways callout for skimmers
• Break long paragraphs into shorter ones
• Add subheadings every 200 to 400 words
• Add specific examples that hold attention
• Add internal links that give the user a next step

Building the Action Queue (Priority and Effort)

Once every page has a decision, the queue has to be ordered and assigned. The two axes that matter are impact and effort.

Impact rough proxy: pages with high impressions and low CTR have the highest update impact because moving CTR on existing impressions is fast traffic. Pages with high CTR but low average position have high update impact because moving position on high-CTR pages compounds well. Pages with low impressions and low rankings have low update impact because there is little to lift.

Effort rough proxy: Keep is 0 hours. Update averages 1 to 4 hours depending on depth. Merge averages 2 to 6 hours. Prune is 0.1 hours. Redirect is 0.1 hours.

The queue prioritization that ships:

1. Prune and redirect first (low effort, immediate quality lift through subtraction)
2. High-impact updates next (highest ROI per hour)
3. Merges third (medium effort but high signal cleanup)
4. Low-impact updates last (or never, if the queue is long)

A useful working rule is to size the quarterly action queue at 80 percent of available hours, leaving 20 percent for emergent work. Queues at 110 percent of available hours never finish.

Quarterly Audit Cadence That Actually Gets Done

The cadence pattern that ships is a quarterly audit on a fixed calendar date. The audit takes one to two days. The action queue clears across the following 8 to 11 weeks. The next audit starts on the calendar date for the next quarter regardless of whether the prior queue fully cleared.

Why quarterly works:

• Frequent enough that drift gets caught before it dominates
• Infrequent enough that the audit overhead does not crowd out other work
• Aligns with normal business planning cycles
• Allows for seasonal pattern detection (quarterly comparison across the same season)

Why monthly does not work:

• Audit overhead becomes the work instead of the input to the work
• Action queue does not have time to clear before the next audit
• Drift is small enough monthly that the marginal value of the audit is low

Why annually does not work:

• Twelve months of drift is too much to address in one cycle
• Sites that audit annually usually skip the audit by month 13 because the queue is too daunting
• Search performance has moved enough in 12 months that the data is partially stale

For broader content strategy that pairs with audit work, the keyword cannibalization fix guide covers a specific audit-driven cleanup pattern. The what is content audit primer covers the basics. Astro SEO Blog publishes audit-related tools and worked examples regularly.

FAQ

How big a site is too big to audit manually? Above 1000 pages, manual page-by-page scoring is no longer practical and you need to score from data first. Above 10,000 pages, even data-first scoring needs sampling (audit 500 pages stratified across the categories rather than every page).

What is the difference between pruning and noindex? Pruning typically means deleting the URL and returning 410 (or 404, or redirecting). Noindex leaves the URL accessible but removes it from the index. Use noindex when the page still serves a user purpose (a thank-you page, an admin page) but should not appear in search. Use pruning when the page has no purpose.

Should I refresh the date on every updated page? Update the modifyDate on every meaningful update. Update the publishDate only if the article was substantially rewritten. Bumping publishDate on minor edits looks like manipulation and Google has gotten better at detecting it.

How do I know if a merge will hurt the surviving page? Generally it does not. Combining two thin pages into one stronger page usually lifts the survivor because the content depth and inbound link equity from the merged page consolidate. The exception is if the two pages target genuinely different intents, in which case the merge can dilute relevance for one of the intents.

Is there a tool that automates content audits? Several. Surfer's Content Audit, Clearscope's Content Inventory, ContentKing's continuous audit, Ahrefs' content explorer, and Semrush's content audit all approach the problem differently. None fully automate the decision step; they automate the data pull and surface candidates for human review. The decision step still benefits from human judgment.

What happens to my site after a major prune? Short-term traffic dips on the pruned URLs (obviously). Site-wide quality signal usually improves over weeks. Pages that were dragged down by section-wide quality issues often see ranking lift. The net effect within 90 days is usually positive for most sites.

Should I audit before or after a Google core update? Both, with different goals. Pre-update audit prevents the update from hitting harder by addressing quality issues proactively. Post-update audit interprets the damage and routes recovery actions. The how to recover from a Google core update guide covers the post-update version specifically.

Sources and Further Reading

Astro SEO Blog covers adjacent strategy work in the topical authority playbook, the E-E-A-T 2026 guide, and the internal linking guide, all of which intersect with audit-driven cleanup.

External references for further reading:

• Google Search Central's quality rater guidelines summary for the content-quality framework.
• Ahrefs' content audit guide for industry-standard data pull patterns.
• Search Engine Land's content audit coverage for analysis and case studies.