Content Auditing with Claude: Finding the Pages That Hurt You

Every site with more than fifty pages has content problems it does not know about. Thin pages that rank for nothing and dilute crawl budget. Two or three articles targeting the same keyword cluster, splitting link equity and confusing Google about which one to rank. Pages written for one search intent that actually attract traffic from a completely different intent, leading to high bounce rates and poor engagement signals.

The traditional way to find these problems is to export your page data into a spreadsheet, manually review each URL, compare keyword targets, and make judgment calls page by page. Claude Opus changes this because of its context window. You can feed it your entire content inventory at once and ask it to reason across pages rather than evaluating each one independently.

Start by building a structured document. For each page on your site, include the URL, the title tag, the H1, the primary target keyword, the word count, and the first 200 words of body content. If you have ranking data from Google Search Console, include the top five queries driving impressions to each page along with the average position for each query. Format all of this as a structured list, one page per entry.

Paste this into Claude and ask three specific questions. First: which pages target substantially the same keyword intent and should be consolidated? Claude will identify pairs and clusters of pages that compete with each other, often catching overlaps you missed because the titles use different phrasing but the underlying intent is identical. Second: which pages have a word count under 500 and are not ranking in the top 20 for any query? These are your thin content candidates, pages to either expand significantly or redirect to a stronger page on the same topic. Third: for each page, does the content match the dominant search intent of its top query? A page targeting “how to do a content audit” that reads like a product pitch has an intent mismatch. Claude is good at catching these because it can compare the informational or commercial nature of the query against the actual tone and structure of the content.

The output is a prioritized list of problems. You will typically find that 10 to 15 percent of your pages need immediate attention, either through consolidation, expansion, or intent realignment. This is where the real ranking gains live, and it is work that most teams never get around to doing manually because the audit itself takes so long. If you want a structured starting point, our SEO audit service runs a version of this process across your full site inventory.

Bulk Title Tag and Meta Description Generation with Claude Code

Rewriting title tags and meta descriptions at scale is one of the highest-leverage SEO tasks you can do, and one of the most tedious. Every page needs a title under 60 characters that includes the primary keyword, communicates the value of clicking, and does not duplicate any other title on the site. Every page needs a meta description under 155 characters that expands on the title, includes a secondary keyword naturally, and gives the searcher a reason to choose your result over the nine others on the page.

Doing this for ten pages is manageable. Doing it for 200 pages is where most teams stall out, either leaving the old meta in place or writing generic descriptions that all sound the same. Claude Code turns this into a batch operation.

The input is a CSV or JSON file with one row per page. Each row should contain the URL, the current title tag, the current meta description, the primary target keyword, one or two secondary keywords, and a one-sentence summary of what the page covers. You can generate this export from your CMS or build it from a crawl.

Claude Code reads this file, processes each row, and outputs a new file with the original data plus two new columns: the proposed title tag and the proposed meta description. In the prompt, you specify your constraints: maximum character counts, brand name placement rules, tone guidelines, and any patterns to avoid. You can tell it to never start two titles with the same word, to always include the primary keyword in the first half of the title, and to end every meta description with a phrase that implies action rather than passive description.

The entire generation step for 200 pages takes a few minutes. The review step, which is the part you should never skip, takes about an hour. You are looking for three things in review: titles that are too similar to each other, descriptions that make promises the page does not deliver on, and any awkward keyword insertions that read like they were written by a machine. Plan to rewrite about 15 to 20 percent of the output by hand. The other 80 percent will be ready to deploy as-is, which still represents an enormous time saving compared to writing all 200 from scratch.

This workflow pairs well with a broader keyword strategy engagement where the target keywords for each page have already been validated through search volume and competition analysis. Generating titles against the wrong keywords is fast but useless.

Fact-Checking and Freshness Audits with Gemini

Content decay is one of the quietest ranking killers. A blog post published 18 months ago with a statistic that was accurate at the time now cites a number that has been updated, references a tool that has been renamed, or describes a Google feature that no longer works the way it did. The post still ranks, but it is slowly losing ground to fresher content that gets the details right.

Gemini is the right tool for this work because of its access to current search results. Claude is better at reasoning and analysis, but for tasks that require checking whether a specific claim is still accurate against today’s sources, Gemini has the advantage of being able to ground its answers in live web data.

The workflow is straightforward. Take any blog post or landing page that is more than six months old and that targets a topic where facts change, which includes most technology and marketing topics. Paste the full content into Gemini and ask it to identify every factual claim, statistic, date, tool name, and version number in the text, then verify each one against current sources. Ask it to flag anything that appears outdated, incorrect, or unverifiable.

Gemini will typically return a structured list: claim, source it found, whether the claim still holds, and what the current accurate version of the claim is. For a 2,000-word blog post, you might get eight to twelve flagged items. Some will be trivial, like a year reference that needs updating from 2025 to 2026. Others will be substantive, like a statistic from a study that has since been superseded by newer research, or a feature description that no longer matches the current product.

This matters for SEO because Google’s helpful content system evaluates whether content demonstrates first-hand expertise and up-to-date knowledge. A page full of stale statistics signals neglect. Running this audit quarterly on your top 50 pages by traffic keeps them competitive without requiring a full rewrite. Combine this with the content analysis from your SEO score calculator to identify which pages need freshness updates most urgently.

Quick-Win Targeting from Google Search Console Data

The most reliable source of SEO opportunity data is already sitting in your Google Search Console account. Most teams look at it occasionally, check their top queries, glance at the click trends, and move on. The real value is in the data you have to filter for deliberately.

Go to the Performance report. Set the date range to the last 90 days. Click into the Pages tab and export the full dataset. Now open the Queries tab and export that as well. You want two files: one showing each page with its total clicks, impressions, and average position, and another showing each query with the same metrics. If your site is connected to Bing Webmaster Tools, pull that data too for a more complete picture.

The quick-win filter is simple: queries where your average position is between 4 and 20 and your impressions are above 100 per month. These are keywords where Google already considers your page relevant enough to show, but something is holding it back from the top three positions where the real click volume lives. Sort this filtered list by impressions descending. The top of this list is your optimization priority queue.

For each query on the list, you need to answer four questions. Does the page’s title tag contain this query or a close semantic match? Does the page have a dedicated section, ideally with an H2, that directly addresses this query? Is the page receiving internal links from other relevant pages on your site? And does the content actually satisfy the search intent behind this query, or does it mention the topic only in passing?

This is where Claude becomes useful again. Feed it the query, the current page content, and the top three organic results for that query. Ask it to identify what the ranking pages cover that yours does not, and to suggest specific additions or restructuring that would make your page more complete. The output is usually a set of concrete edits: add a section covering X, rewrite the introduction to match the informational intent more clearly, strengthen the H2 to include the query phrase naturally.

Run Google Lighthouse against each target page as well to ensure there are no technical performance issues dragging the page down. A page with strong content relevance but a three-second load time is leaving rankings on the table for a fixable reason. Our technical SEO service addresses these performance bottlenecks alongside the content improvements.

Working through ten of these quick-win optimizations per week is realistic for a single person. Over the course of a quarter, that adds up to 120 pages improved with targeted, data-backed changes. The cumulative ranking impact of this approach consistently outperforms publishing new content, because you are improving pages that Google already trusts enough to show.

Using Microsoft Clarity to Find Where Content Fails

Rankings data tells you how Google perceives your pages. Microsoft Clarity tells you how readers perceive them. The gap between these two things is where most SEO content underperforms.

Clarity is free and installs with a single script tag. Once it is running, it records session replays and generates heatmaps showing exactly how visitors interact with each page. For SEO purposes, three Clarity features matter most: scroll depth heatmaps, dead click tracking, and the engagement metrics by page section.

Start with your highest-traffic pages. Open the scroll depth heatmap for each one. You are looking for sharp drop-offs, points where a significant percentage of readers stop scrolling. Every page has some natural attrition as you move down, but a sudden cliff, say from 60 percent of readers to 25 percent, indicates a specific section where you are losing people. Go to that section of the page and read it critically. Is it a wall of text without subheadings? Is it covering something the reader did not come for? Is the information there but buried in jargon?

Dead clicks, locations where users click but nothing happens, reveal a different kind of problem. They often appear on text that looks like a link but is not, on images that users expect to expand, or on section headers that users think are collapsible. Each dead click is a moment of user frustration. Fixing them is usually trivial: make the text an actual link, add an image lightbox, or restructure the section so users do not expect interactivity that is not there.

The real power of Clarity for SEO comes from combining its behavioral data with your Search Console data. Take a page that ranks at position 7 for a high-value query and has a Clarity scroll depth showing that 70 percent of readers leave before reaching the middle of the article. That page has a content engagement problem, not a relevance problem. Google thinks it deserves to rank, but user behavior signals are holding it back. Rewrite the section where readers disengage, tighten the opening paragraphs to deliver value faster, and break up long sections with clearer subheadings. When you re-check a few weeks later, the ranking improvement from better engagement signals is often measurable.

This type of analysis feeds directly into content strategy decisions. Pages where engagement is strong across the full length tell you what your audience values. Pages where readers bail at the halfway mark tell you where your content does not meet expectations. Over time, these patterns inform how you structure every new page. Our content strategy service uses this behavioral data alongside ranking and traffic metrics to build editorial plans grounded in what actually works for your specific audience.

Bringing the Workflows Together

These five workflows are not isolated techniques. They form a cycle. The content audit identifies which pages need work. The Search Console data tells you which of those pages have the highest upside. Clarity data tells you specifically what is wrong with the user experience on those pages. Claude helps you draft the improvements. Gemini verifies that your updated content is factually current. Claude Code handles the repetitive execution tasks like generating new meta tags for every page you touched.

The order matters. Starting with bulk meta generation before you have done a content audit means you might be writing polished titles for pages that should be consolidated or redirected. Starting with Clarity analysis before you have Search Console data means you might optimize engagement on pages that have no ranking potential regardless. The audit comes first because it defines the scope of work. Everything else follows from that prioritized list.

One important thing to note about AI in these workflows: it is not making the strategic decisions. You are deciding which keywords to target, which pages to keep versus consolidate, and whether a piece of content should be informational or commercial. AI is compressing the execution time for each of those decisions. The SEO practitioner who understands search intent, competitive dynamics, and content quality is still the one driving the outcomes. AI just means they can drive faster.

The biggest mistake we see teams make is using AI to generate content without first doing the analytical work. Writing new pages with Claude before auditing existing pages is how you end up with more cannibalization, not less. Writing meta descriptions before validating target keywords is how you optimize for the wrong queries. The analytical workflows covered in this guide should precede any generation work, and the generation itself should be reviewed by someone who understands the strategy behind each change.

If you want to run these workflows but lack the internal bandwidth or expertise to manage the full cycle, our AIO optimization service handles the end-to-end process: audit, prioritization, execution, and measurement. We use the same tools described here, applied with the context of having run these workflows across hundreds of sites.

For teams running this internally, start with the Search Console quick-win analysis. It requires the least setup, produces results within weeks, and builds the internal confidence to invest in the deeper audit and engagement workflows. The returns compound as you layer each workflow on top of the last. If you want a benchmark before starting, run your site through our SEO score calculator to see where you stand and request a full audit when you are ready to go deeper.

Frequently Asked Questions