How to Do a Website Content Audit: The Search & AI Playbook

A website content audit inventories every published URL and evaluates each one against performance, quality, business goals, and search visibility. Every page ends with one verdict: keep, update, consolidate, or delete.

Most audits end as abandoned spreadsheets because someone crawls the site, exports six months of analytics, color-codes 400 rows, and then the quarter changes before anything ships. A useful audit produces fewer findings, clearer decisions, and a backlog a writer can start on Monday. It also accounts for a layer many audits still skip: how content appears in AI-generated answers.

The Website Content Audit Playbook: SEO, AI & Visibility

The work follows a repeatable sequence:

1. Crawl the site and create the URL inventory.
2. Add GA4, Search Console, backlink, and AI-search visibility data.
3. Evaluate each URL against a consistent set of criteria.
4. Assign a keep, update, consolidate, or delete verdict.
5. Ship the highest-impact changes first.
6. Measure the results after 30, 60, and 90 days.

Collecting the data is easier than turning it into decisions that someone owns and ships. An audit that produces 300 annotated rows but no published changes has consumed two weeks without improving the site.

Discovery has also moved beyond standard search results. Neither GA4 nor Search Console shows whether ChatGPT, Perplexity, Gemini, or other AI systems are accessing your pages, citing them in answers, or relying on competitors instead.

Scrunch adds that missing layer to the audit.

Its Site Diagnostics view provides a page-level view of AI performance, including Audit Score, Agent Traffic, and Citations. Agent Traffic also shows which AI agents are visiting the site and which pages they rely on. That gives teams another signal when evaluating a page: a URL with modest Google traffic but regular AI citations may be worth protecting, while a strong Google performer with no AI visibility may need a different type of update.

‍

What a Content Audit Actually Is (and Which One You’re Running)

The audit’s objective determines how pages are judged and which spreadsheet columns carry the most weight:

• In an SEO improvement audit, look for underperforming pages, keyword cannibalization, and lost organic visibility, giving the most attention to clicks, impressions, average position, and query overlap.
• A conversion audit focuses on pages that attract traffic but produce no measurable action, so engagement time, conversion events, and assisted conversions matter more than rankings.
• A redesign or migration audit decides what moves to the new site, what needs rewriting, what should merge, and what should be removed; referring domains and internal link counts can determine the final decision.
• An AI visibility audit checks whether AI-generated answers represent your content for important topics and whether competitors are cited when your brand is absent.

AI visibility increasingly belongs within the other three audit types instead of operating as a separate exercise. A page can hold position 4 in Google and remain absent from AI answers on the same topic, a gap that will not appear in a Search Console export.

Choose one primary objective before opening the spreadsheet. Trying to audit for every possible goal at once usually leaves teams with too many findings and no clear order of work.

Building Your Audit Stack: Search, Analytics, Backlinks & AI Visibility

Which tool is best for a website audit? No single tool covers every job, so the best audit stack combines tools that can:

1. Crawl the site and create the raw URL inventory.
2. Measure engagement and conversions for each page.
3. Measure traditional search visibility by page and query.
4. Check backlink authority before pruning content.
5. Measure AI-search visibility for the brand and its priority topics.

How do you perform a website audit? Crawl the site, merge analytics, search, backlink, and AI visibility data into the URL inventory, evaluate each page, assign a verdict, ship the prioritized changes, and measure again after 30, 60, and 90 days.

Evaluate tools on four practical criteria: whether they supply URL-level data instead of site-level totals, whether the data can be exported, how much history is available, and whether the same view can be rerun against the baseline. A polished dashboard has little audit value if its data cannot be exported.

The fifth job requires a separate data source. GA4 and Search Console measure clicks that reach the site, but they cannot identify whether an AI assistant summarized your comparison page, cited a competitor, or relied on a third-party listicle that omitted your brand.

1. Scrunch

‍

Scrunch fills the AI-search layer that traditional audit tools cannot. Search Console tells you how a page performs in Google, GA4 tells you what visitors do after arriving, and backlink tools show the authority a URL has accumulated. Scrunch shows how AI systems interact with the site and how often its pages appear in AI answers.

Use it at three points in the audit:

• Diagnosis: Site Diagnostics provides page-level AI performance data, including Audit Score, Agent Traffic, and Citations, and helps identify issues that may affect how AI systems retrieve or interpret a page.
• Prioritization: Agent Traffic shows which AI agents visit the site and which pages they rely on, helping identify pages that matter to AI systems even when traditional traffic is modest.
• Measurement: After a page is rewritten, consolidated, or restructured, Scrunch can be used to monitor changes in AI visibility, citations, and agent activity.

This gives teams four ways to evaluate the same URL:

• Search: Does it attract impressions, clicks, and rankings?
• Business: Does it generate conversions or support a commercial goal?
• Authority: Does it have backlinks and internal links worth protecting?
• AI: Can AI agents access it, do they use it, and do they cite it?

AI visibility should influence the verdict, not automatically determine it. A page with low Google traffic and no AI visibility may be an obvious deletion candidate. A page with low Google traffic but frequent AI citations may be worth protecting or improving.

Standout capabilities: Page-level Site Diagnostics, Audit Scores, Agent Traffic, AI citations, AI referrals, and monitoring of brand and competitor presence in AI answers.

Role in the audit: Adds AI-agent behavior and AI-search visibility to the URL-level decision process.

Best for: Teams that want content audits to account for both traditional organic search and AI-generated answers.

2. Screaming Frog

‍

Screaming Frog creates the site’s inventory by crawling it as a bot would. Its export supplies the structural columns for the master spreadsheet, including each URL’s status code, title, H1, meta description, word count, canonical, and indexability.

• Standout capabilities: Configurable desktop crawling, include and exclude rules, full technical exports by URL, directory-specific crawls, and GA and Search Console API connections.
• Role in the audit: Builds the master URL list to which every other dataset is joined.
• Best for: Small to mid-size sites and section-by-section crawls of larger sites.
• Pricing: Free for up to 500 URLs per crawl. The paid SEO Spider license costs $279 per user per year (£199 for 1 to 4 seats).

3. Google Search Console

‍

Search Console reports the impressions and average positions that pages receive in Google results, data that third-party estimates cannot reproduce. The Pages and Queries reports are the primary audit exports. Filters for individual pages and directories also make large-site audits easier to scope.

• Standout capabilities: Clicks, impressions, CTR, and average position by URL and query; 16 months of history; query-to-page mapping for cannibalization analysis; and page indexing status.
• Role in the audit: Supplies traditional search visibility by URL and the query data needed to find competing pages.
• Best for: Any site, regardless of size.
• Pricing: Free.

4. Google Analytics (GA4)

‍

‍

GA4 shows what visitors do after clicking through to the site, including how long they engage and whether the page contributes to a conversion event. Those metrics can prevent a low-traffic page with an 8% conversion rate from being deleted.

• Standout capabilities: Pages and screens reporting with views, average engagement time, and event counts; conversion events by landing page; and traffic-source segmentation for non-organic value.
• Role in the audit: Measures whether a page contributes value after someone arrives.
• Best for: Any site with conversion goals or forms to protect.
• Pricing: Free.

5. Ahrefs Site Audit

‍

Ahrefs adds backlink context to pruning decisions. A page receiving 12 visits per month looks expendable until the audit shows that it holds 40 referring domains. That evidence can change a delete verdict into a redirect or update.

• Standout capabilities: Referring domains and backlink counts by URL, scheduled cloud crawls with health scoring, a Top Pages report that ranks URLs by organic value, and internal link data for orphan detection.
• Role in the audit: Protects accumulated authority and serves as a crawl alternative to desktop software.
• Best for: Established sites with links that need protection.
• Pricing: Lite starts at $129/month. Standard costs $249/month with monthly billing.

Step 1: Crawl the Site and Build the Master Spreadsheet

Start with a crawl that creates one row per URL, then layer analytics, search, backlink, and AI visibility data onto the same sheet.

1. Crawl the site. In Screaming Frog, export internal HTML URLs with the core technical fields: status, indexability, title, meta description, H1, word count, canonical, crawl depth, and internal links.
2. Add GA4 data. Pull the previous 12 months of views, engagement, and relevant conversion data by page.
3. Add Search Console data. Export the same 12-month period for clicks, impressions, CTR, and average position. Export queries separately for cannibalization analysis.
4. Add backlink data. Use Ahrefs to pull referring domains by URL.
5. Add AI visibility data. Use Scrunch to establish a baseline for priority topics, including brand presence, competitor presence, citations, cited URLs, and AI-agent traffic.

Use a normalized URL path as the join key. Strip the protocol and domain, standardize trailing slashes and capitalization, and remove query parameters unless they represent distinct pages. Then join the datasets with XLOOKUP or VLOOKUP.

Your audit sheet should cover five areas: URL and technical data, search and analytics performance, backlink and AI visibility signals, page-quality assessments, and final decisions. Keep verdict, ownership, priority, and effort fields empty until the diagnosis is complete.

For a 300-page site, the evaluation pass alone can take roughly 25–50 hours at 5–10 minutes per page. The main cost of an audit is usually the analysis and remediation, not the tools.

Step 2: Diagnose Search and AI Visibility

Before assigning verdicts, examine the full dataset for places where visibility is being lost. Traditional search and AI search should be analyzed separately first, then compared.

For traditional search, use this three-item checklist:

• Positions 11 to 15. These pages already have relevance signals and may need a stronger introduction, better internal links, or updated data to reach page one.
• High impressions and below-average CTR. This pattern often points to a title or meta description that does not match the query.
• Declining clicks with flat impressions. Stable impressions combined with falling clicks can indicate that competitors are overtaking the page.

For AI visibility, use Scrunch alongside the same priority URLs and topics:

• Brand presence. Does the brand appear in AI answers?
• Accurate description. Is the brand described accurately?
• Sources in its absence. Which sources appear when the brand is absent?
• Cited pages. Which pages from the site are being cited?
• Agent activity. Are AI agents accessing the page?

‍

Citation analysis can reveal reliance on review sites, forums, industry roundups, competitors’ comparison pages, or other third-party sources. Those gaps will not appear in Search Console and may point toward an authority-building opportunity.

The most useful findings often come from disagreement between the datasets. A guide can hold position 3 with steady clicks yet receive no citations in AI answers about its own topic. The page may bury its answer beneath 600 words of preamble or state facts in a form that an AI system cannot extract cleanly.

Record each AI visibility finding as an opportunity and attach the evidence before moving to page-level verdicts.

Step 3: Judge Every Page on Performance and Quality

Performance data identifies struggling pages, but reading the content explains why they struggle. Use a named checklist so different reviewers apply the same standards:

• Accuracy and staleness. Look for redesigned interfaces, changed pricing, discontinued tools, statistics without dates, outdated strategies, and incorrect company information.
• Search intent match. Pull the page’s top three queries from Search Console and compare the content with those needs.
• Content quality. Assess depth relative to current ranking pages, original data or perspective, clarity, and whether readers can skim to the answer.
• User experience. Check mobile rendering, long text blocks, missing subheads, interstitials, and loading behavior.
• Thin and duplicate content. Find sub-300-word pages without a clear function, near-identical pages, and indexed tag or archive pages.
• Conversion performance. On commercial pages, check whether the next step is visible and appropriate for the visitor’s stage.
• AI-search readiness. Determine whether the page states its main answer early, uses unambiguous entity names, follows a structure models can parse, and includes the numbers, definitions, and comparisons needed for accurate representation.

Performance and quality must be read together. High traffic does not guarantee a keep verdict. A page attracting 4,000 monthly visits from a commercially irrelevant query, with a 12-second engagement time, may belong within a page that better serves the intended audience.

If multiple people are reviewing content, score each check from 1 to 3 and record the total. The scoring is basic, but it prevents one reviewer’s “needs work” from carrying a different meaning than another’s.

AI tools can assist with intent assessment, stale-claim detection, query clustering, and drafts of revised content. They cannot independently crawl the site, access Search Console or GA4, or identify which pages hold backlinks. Use them during page evaluation, not as a substitute for the inventory and visibility data.

Finding Keyword Cannibalization With Search Console Queries

Cannibalization occurs when multiple URLs collect impressions for the same query and split relevance signals between pages that should be combined. Search Console exposes the pattern:

1. Export the Queries tab for the same 12-month window. Filter by query and inspect the Pages tab for each one.
2. Flag queries associated with two or more URLs that receive meaningful impressions.
3. Confirm that the intent overlaps. A how-to guide and a template download can rank for “content audit” while serving different needs.
4. Choose the canonical page using query-intent match, click and position history, referring domains, and proximity to a business outcome.
5. Record consolidation candidates in the master sheet, including the canonical URL, absorbed URLs, and sections from the losing pages that should be retained.

AI visibility can break a close tie. If three pages address one topic but AI answers consistently cite only one, that page has evidence of being treated as the authoritative source. Use that evidence alongside intent, traffic, links, and business value when deciding which page should become the canonical source.

Step 4: Give Every URL One of Four Verdicts

Every row needs one verdict, an owner, and a target date. Do not leave blank or “maybe” cells.

• Keep. The page remains relevant, accurate, and effective against its objective.
• Update. Demand or visibility potential exists, but the page does not capture it.
• Consolidate. Multiple pages compete for the same intent. Merge them into one canonical page and redirect the others.
• Delete or redirect. The page has no meaningful traffic, visibility, authority, conversions, strategic purpose, or realistic path to any of those outcomes.

Avoid universal thresholds such as “delete anything under 100 visits.” A useful cutoff depends on the site’s traffic distribution, sales cycle, and whether the page is two months or five years old.

AI visibility and recoverable value can change the verdict. A page with modest organic traffic may place the brand in AI answers for a high-intent topic, even though that value never appears in a clicks column. If updating, consolidating, or repairing internal links can restore a page, that option retains the URL, links, and history.

How Should You Decide What to Do With Each Page?

Ask these questions in order, and stop when the answer produces a verdict.

‍

Before You Delete Anything: The Pre-Deletion Checklist

Deletion is difficult to reverse, so every candidate needs four checks:

• Referring domains. Pull the backlink count for the URL. Any page with external links has authority that deletion could discard.
• Internal links. Review inlink counts from the crawl. A page with 40 internal links may be structurally important regardless of traffic.
• Non-organic traffic and conversion paths. Segment GA4 by source. Pages with little organic traffic may still receive substantial email, paid, or sales-enablement use.
• AI visibility and topic importance. Check whether the page receives citations or AI-agent activity, or covers a topic where brand representation is already limited.

Map every deleted URL to the closest relevant live page and implement a 301 redirect. Do not redirect everything to the homepage, which Google frequently treats as a soft 404. The old assumption that 301 redirects lose roughly 15% of PageRank has been retired. As of 2025, Google confirms that correctly implemented 301 redirects pass full PageRank. Archive the content somewhere retrievable before removal.

Pruning is often over-prescribed because advice to delete underperforming content produced fast gains on bloated sites with thousands of auto-generated pages, then spread to sites where a few dozen dated posts only needed rewriting. If a page has links, internal relevance, or a topic the business still values, a few hours of updating preserves what it has accumulated instead of forfeiting those assets through deletion.

Step 5: Prioritize the Fixes That Can Move Visibility

Score each verdict on four dimensions before writing begins: potential impact, confidence in the proposed fix, effort in hours, and the topic’s business value. Sort by that four-factor score to turn the backlog into a working queue.

Ship quick wins during the first two weeks:

• Fix broken links and redirect chains found during the crawl.
• Rewrite titles and meta descriptions on high-impression, low-CTR pages.
• Add internal links from strong pages to those ranking in positions 11 to 15.
• Update stale facts, pricing, and screenshots on pages that already receive traffic.

Next, for AI visibility, Scrunch can take much of the prioritization work off your plate. Instead of manually deciding which pages need optimization, its Site Optimization recommendations identify page-level issues that may be limiting AI visibility and prioritize the changes with the greatest potential impact. This gives the team a ready-made queue of pages to optimize based on AI search performance, competitor performance, and Scrunch's data.

‍

‍

Use those recommendations alongside your business and search data. An established page with existing authority and traffic may be a better optimization target than creating something new, while commercially important gaps may justify more substantial content work. Group related updates by topic cluster so writers can tackle several pages efficiently.

Step 6: Ship the Changes and Prove They Worked

Work through the queue by impact, not spreadsheet order. For every change, log the URL, update, owner, and publication date so you can connect future performance changes to the work.

Measure organic and AI-search performance separately at 30, 60, and 90 days. For AI search, track brand presence, citations, cited URLs, competitor presence, AI referrals, and agent activity rather than looking for a single “AI ranking.” Scrunch makes it possible to compare these signals before and after an update and see whether important pages are being accessed, represented accurately, and cited.

A Four-Week Audit Sprint

For a mid-size site, the first audit can run as a four-week sprint:

• Week 1: Crawl the site and collect search, analytics, backlink, and AI visibility data.
• Week 2: Normalize and merge the datasets.
• Week 3: Evaluate pages, resolve cannibalization, and assign verdicts.
• Week 4: Prioritize the queue and ship the highest-impact fixes.

The sprint gets you from inventory to action; larger refresh and consolidation projects can continue afterward. Future audits should get faster as your templates, data structure, and decision criteria are established.

‍

Ready to see what your content audit reveals about AI search? Scrunch helps you identify where your brand and content appear, which pages get cited, where competitors are winning, and what to optimize next.

Try Scrunch free for 7 days