How to Use Claude to Run SEO Audit: Workflows That Get Results

How to use Claude to run seo audit depends on which version of Claude you are working with and what kind of audit output you need. Claude.ai handles analysis, content review, and structured reporting when you paste page data, crawl exports, or Search Console CSV files directly into the chat.

Claude Code: Anthropic’s command-line tool: connects to live websites, runs Python scripts against APIs, and generates automated audit reports that repeat on demand. Claude Projects stores your audit framework, site context, and recurring instructions so every audit conversation starts with the same baseline rather than re-explaining the site from scratch. This guide covers all three workflows with copy-paste prompts, what each method produces, and when to use which approach.

What Claude Can and Cannot Do in an SEO Audit

Before running any workflow, the distinction between Claude’s capabilities and its limitations determines whether the output is useful or misleading.

What Claude does well: Analysing structured data you provide: crawl exports, GSC performance CSVs, Screaming Frog output, page HTML, server log summaries. Claude processes large datasets without losing context, identifies patterns across hundreds of rows, prioritises issues by impact, and writes findings in plain language. It also reasons through complex SEO decisions: page consolidation, keyword cannibalisation resolution, content gap analysis: where the answer depends on weighing multiple factors simultaneously.

What Claude cannot do without tools: Claude.ai has no live internet access by default. It cannot crawl a URL, check live page status codes, pull real-time ranking data, or verify whether a page is indexed. It works with data you bring to it, not data it fetches itself. Claude Code with web tools or MCP integrations removes this constraint: but requires setup.

The practical implication: The highest-value use of how to use Claude to run seo audit is not replacing crawl tools: it is replacing the analyst who interprets crawl data. Screaming Frog finds the issues. Claude tells you which ones to fix first and why.

Workflow 1: Technical SEO Audit Using Claude.ai Chat

This workflow requires no setup beyond a Claude.ai account. Export crawl data from Screaming Frog, Sitebulb, or a similar tool as a CSV, paste the relevant columns into Claude, and run the analysis prompt. This is how to use Claude to run seo audit without any coding or API access.

Step 1: Prepare your data export

Export the following columns from your crawl tool into a single CSV: URL, status code, title tag, meta description, H1, word count, canonical, indexability, internal links in, page speed score. Paste the full CSV content into Claude: Claude can handle several hundred rows in a single message.

Step 2: Run the triage prompt

You are a senior SEO auditor. I am pasting a Screaming Frog crawl export below.
Analyse the data and produce a prioritised audit report structured as:
1. Critical issues (block rankings or indexing immediately)
2. High priority issues (significant ranking impact)
3. Medium priority issues (improve performance over time)
4. Quick wins (low effort, measurable uplift)
For each issue: name the issue, list the affected URLs (max 5 examples), explain the SEO impact, and write one specific recommended fix.
Flag any patterns: for example, if H1 tags follow a repeated template across category pages, call that out as a structural issue rather than listing each URL individually.
Here is the crawl data:

[paste CSV content]

Step 3: Run the on-page content audit prompt

For individual page audits, paste the full page HTML or a plain-text copy of the page content and run this prompt:

Audit this page for on-page SEO. Analyse:
- Title tag: length, keyword placement, click appeal
- Meta description: length, intent match, missing or duplicate
- H1 and heading structure: hierarchy, keyword alignment, readability
- Content: thin content risk, keyword coverage, E-E-A-T signals, internal links
- Structured data: what schema is present, what is missing for this page type
Target keyword for this page: [insert keyword]
Page type: [product / category / blog / service / landing page]
Provide specific rewrites for the title tag, meta description, and H1. Flag any compliance concerns for YMYL content if applicable.
Page content:

[paste page HTML or plain text]

Step 4: Run the Search Console analysis prompt

Export 90 days of GSC performance data: queries, pages, clicks, impressions, CTR, position. Paste it into Claude and run:

Analyse this Google Search Console data for [domain].
Identify and prioritise:
1. CTR underperformers: pages ranking position 1-15 with CTR below position benchmarks (estimate benchmarks as: P1 = 28%, P2 = 15%, P3 = 11%, P4-10 = 5-8%, P11-15 = 2-4%)
2. Impression-high, click-low pages: high visibility with no click benefit: likely title/description misalignment
3. Keyword cannibalisation signals: the same or closely related query driving traffic to two or more different pages
4. Traffic decline signals: queries with strong impression volume but declining position over the period
For CTR underperformers, write revised title tag suggestions that maintain keyword intent but improve click appeal. Keep titles under 60 characters.
GSC data:

[paste CSV export]

This four-step Claude.ai workflow covers the full scope of a standard technical and on-page audit without any API access or code. The output is a structured markdown report that can be pasted directly into a client deliverable.

Workflow 2: Automated Audit Using Claude Code

Claude Code is Anthropic’s command-line tool that connects Claude to a real browser, live APIs, and your local file system. This is how to use Claude to run seo audit at scale: the same analysis runs on any URL on demand, and outputs save to a markdown or HTML report file automatically.

Prerequisites: Claude Code installed, a terminal (Mac/Linux) or WSL (Windows), and optionally a Puppeteer MCP server for live browser access.

Step 1: Install Claude Code

npm install -g @anthropic-ai/claude-code
claude

Step 2: Connect Puppeteer MCP for live URL crawling

With Puppeteer MCP running, Claude Code can visit a URL, render the page as a browser would, extract meta tags, check load times, take screenshots, and pull structured data: all within a single audit prompt.

Step 3: Run the full-site audit prompt in Claude Code

You are an SEO auditor. Using Puppeteer, visit [URL] and conduct a complete SEO audit.
Check the following in sequence and record your findings as you go:
1. Meta tags: title tag, meta description, robots meta: flag missing, duplicate, or noindex
2. Heading structure: H1 presence and uniqueness, heading hierarchy on the page
3. Structured data: detect any schema markup present, validate against Google's supported types
4. Core Web Vitals: record LCP, CLS, and INP from PageSpeed data if accessible
5. Mobile rendering: take a screenshot at 375px width and note any layout issues
6. Crawlability: check robots.txt and XML sitemap availability
7. Internal links: count internal links on the page and list any broken link URLs
8. Page speed: record load time and flag any render-blocking resources visible in the source
When complete, write all findings to a file called seo-audit-[domain]-[date].md with:
- An executive summary (3-5 sentences)
- Critical issues list
- All findings by category
- Prioritised fix list with effort estimates (low/medium/high)

Step 4: Build a repeatable audit skill

Save this as a slash command in ~/.claude/skills/seo-audit.md so the full audit runs with a single /seo-audit [URL] command on any future site:

# SEO Audit Skill
Run a comprehensive SEO audit for the provided URL using Puppeteer MCP.
## Checks to run in sequence:
1. Meta tags and title analysis
2. Heading hierarchy
3. Structured data detection and validation
4. Core Web Vitals via PageSpeed API
5. Mobile screenshot and layout review
6. Robots.txt and sitemap check
7. Internal link count and broken link detection
8. Load time and render-blocking resource check
## Output format:
Save results to seo-audit-[domain]-[date].md with executive summary, critical issues, full findings by category, and prioritised fix list.

Claude Code with a saved audit skill reduces a full technical audit to a single command. The same output format runs consistently across every client site.

Workflow 3: Ongoing Audit Management Using Claude Projects

Claude Projects stores persistent context: your audit methodology, client site information, brand guidelines, and recurring instructions: so every audit conversation in that project starts with the correct baseline. This is how to use Claude to run seo audit for agency clients or sites you audit on a monthly basis.

Step 1: Set up the project context

Create a new Claude Project and add the following as project instructions:

You are a senior SEO auditor working on [client name / site name].
Site context:
- Domain: [domain]
- CMS: [WordPress / Shopify / custom]
- Primary audience: [describe]
- Target keywords: [list primary focus areas]
- Known issues from previous audits: [list or write "none yet"]
- Reporting format: structured markdown with executive summary first
When I provide crawl data, GSC exports, or page content, analyse against this context. Prioritise issues based on business impact for this specific site, not generic SEO rules. Flag any issue that contradicts a previous decision we have documented here.

Step 2: Upload reference documents

Upload your most recent crawl export, last month’s GSC data, and any previous audit reports to the project. Claude will reference these automatically in every conversation: you can ask “what has changed since last month’s audit?” and Claude will cross-reference the uploaded files without re-uploading.

Step 3: Run the monthly delta audit prompt

At the start of each audit cycle, upload the new GSC and crawl data and run:

Compare this month's data against the files already in this project.
Identify:
1. New issues that did not appear in the previous audit
2. Resolved issues from the previous audit: confirm they are fixed
3. Issues that were flagged previously but have worsened
4. Traffic or ranking changes that require investigation: pages with >20% change in clicks or position
Produce a delta audit report showing: what changed, what improved, what regressed, and the three highest-priority actions for this month.

The Projects workflow turns how to use Claude to run seo audit into a repeatable monthly process. Previous audit context eliminates the setup time that makes one-off audits inefficient at scale.

Prompt Templates for Specific Audit Tasks

These prompts work in Claude.ai chat, Claude Code, or Claude Projects: copy-paste ready for the most common individual audit tasks.

Keyword cannibalisation analysis

Analyse this list of pages and their ranking queries for keyword cannibalisation.
For each cannibalisation conflict identified:
- Name the competing pages
- Identify the shared query
- Assess severity: HIGH (two pages splitting traffic for a primary keyword), MEDIUM (intent overlap but different stages), LOW (synonym overlap only)
- Recommend resolution: consolidate, differentiate, or redirect
Page data:

[paste GSC data or list of page URLs with their top queries]

Internal linking audit

Analyse the internal linking structure from this crawl export.
Identify:
1. Orphaned pages: indexed pages with zero internal links pointing to them
2. Over-linked low-value pages: pages with many internal links but low GSC traffic
3. Under-linked high-value pages: pages with strong GSC performance but few internal links
4. Anchor text patterns: flag over-optimised or generic anchor text at scale
For each priority finding, write the specific internal link addition recommended: source page, target page, anchor text, and placement context.
Crawl data:

[paste internal links export]

Schema markup audit

Review the structured data on this page and produce:
1. A list of all schema types detected
2. Validation status for each: does it match Google's supported schema requirements?
3. Schema types that should be present for this page type but are missing
4. The corrected JSON-LD code for any invalid or missing schema
Page type: [product / article / local business / FAQ / recipe / event]
Page HTML:

[paste page source]

E-E-A-T content assessment

Assess this page against Google's September 2025 Quality Rater Guidelines for E-E-A-T.
Evaluate:
- Experience: does the content demonstrate first-hand experience with the topic?
- Expertise: is the author's qualifications or knowledge clearly demonstrated?
- Authoritativeness: are there signals that this source is recognised in the field?
- Trustworthiness: is there transparency about authorship, date, sources, and purpose?
Flag specific sentences or sections that weaken E-E-A-T and write replacement copy that addresses the weakness without changing the factual content.
Page type and topic: [describe]
Page content:

[paste content]

Which Claude Workflow to Use and When

Common Mistakes When Using Claude for SEO Audits

These are the errors that consistently reduce audit output quality patterns that appear across both first-time and experienced Claude users regardless of which workflow they are running.

• Asking Claude to audit a URL without providing data: Claude.ai has no live crawl access. If you type “audit example.com” without pasting page data, Claude will produce a generic checklist, not a site-specific audit. Fix: always paste the actual data: crawl export, page HTML, or GSC CSV: before running any analysis prompt
• Accepting output without verifying against live site: Claude analyses what you give it. If your crawl data is two weeks old or incomplete, the audit reflects that. Fix: always date-stamp your data and note data recency in the prompt context
• Using vague prompts that produce generic advice: “audit my site for SEO” produces a generic checklist. Fix: specify page type, target keyword, CMS, and the specific issue category you are investigating in every prompt
• Running Claude Code audits without server-side rendering check: JavaScript-rendered pages may appear empty to Puppeteer if the site requires user interaction to load content. Fix: add “confirm the page has fully rendered before extracting content: check that the main body text is visible in the DOM” to any Claude Code audit prompt
• Not saving the audit output to a file: Claude Code sessions do not persist. If you do not instruct Claude to write findings to a file, the analysis exists only in the terminal session. Fix: end every Claude Code audit prompt with “write all findings to [filename].md”

FAQs About How to Use Claude to Run SEO Audit

Can Claude audit a website without any tools or crawl data?

Claude.ai cannot crawl a live URL: it analyses data you provide.

Pasting page HTML, crawl exports, or GSC data is required for a site-specific audit. Claude Code with Puppeteer MCP can visit and render live URLs without pre-exported data, but requires installation and configuration.

For most users, the fastest path to a useful audit is exporting data from a crawl tool and pasting it into Claude.ai with a structured analysis prompt.

How accurate are Claude’s SEO audit findings?

Claude’s findings are as accurate as the data you provide.

When given complete, current crawl data and GSC exports, Claude produces analysis that matches or exceeds what a mid-senior SEO analyst produces manually: with the advantage of processing the full dataset rather than sampling.

The risk is hallucination on questions Claude cannot verify from the data, such as competitor rankings or live page status. Frame prompts around data analysis rather than asking Claude to speculate about things it cannot confirm.

What data should I prepare before running a Claude SEO audit?

The minimum useful dataset for a technical audit is a Screaming Frog crawl export (URL, status, title, meta, H1, canonical, word count) and 90 days of Google Search Console performance data (queries, pages, clicks, impressions, CTR, position).

Adding server log data, previous audit reports, and Core Web Vitals data from PageSpeed Insights API significantly improves the analysis. All of these can be uploaded directly to Claude.ai or Claude Projects as files, or pasted as CSV content in chat.

Is Claude Code necessary for SEO audits or is Claude.ai sufficient?

Yes, Claude.ai is sufficient for most SEO audit workflows: data analysis, on-page review, GSC interpretation, content quality assessment, and schema validation all work through the chat interface without any setup.

Claude Code becomes valuable when you need to audit live URLs without pre-exporting data, build repeatable automated audit scripts that run on command, or integrate SEO audit workflows with Google APIs and other data sources. For one-off audits, Claude.ai is faster. For recurring audits at agency scale, Claude Code with saved skills reduces per-audit time significantly.

Can I use Claude to audit multiple pages at once?

Yes, Claude can analyse multi-page crawl exports containing hundreds of rows in a single conversation.

For bulk on-page audits, paste the crawl export as a CSV and instruct Claude to identify patterns across all pages rather than analysing each URL individually. For individual page content analysis, run one page per conversation to maintain analysis depth. Claude Projects allows you to build a session where multiple pages are analysed sequentially within the same context window, maintaining consistency across the audit.

Author Profile

HiAgency

HiAgency is a digital marketing and web development agency specializing in SEO optimization and content management solutions. We help businesses improve their online visibility through technical implementations, user experience enhancements, and strategic content organization.

Latest entries

• Local SEOOctober 7, 2026Bing Places vs Google Business Profile: Which Platform Is Right for Your Business?
• Local SEOOctober 6, 2026Why My Google Business Profile Suspended? Fix It
• Local SEOOctober 5, 2026What Is Google Business Profile? A Beginner’s Guide
• SEOSeptember 28, 2026How to Choose an SEO Agency in Australia: A Practical Guide