---
title: "How to Run a First-Pass SEO and GEO Website Audit with Claude Skills"
url: https://www.darwinapps.com/blog/how-to-run-a-first-pass-seo-and-geo-website-audit-with-claude-skills/
type: article
---

![The image features an illustration of a magnifying glass with a flower design on it. The magnifying glass is positioned over a pink background and appears to be open, allowing light to pass through and illuminate the flower design. The overall composition of the image creates a visually appealing contrast between the black and white elements, such as the magnifying glass and the flower design.](https://cdn.sanity.io/images/qd0fa73p/production/d515b6410b4240d375854939bfd8e0ef008e64ed-2984x1679.png?w=1492&q=85&auto=format)

# How to Run a First-Pass SEO and GEO Website Audit with Claude Skills

- [#SEO](https://www.darwinapps.com/blog/category/seo/)

**Quick Answer:**Claude Skills for SEO and GEO run a first-pass website audit that surfaces technical SEO issues, schema gaps, Core Web Vitals signals, crawlability problems and AI-search readiness gaps in a single pass. The primary Skill we currently use is [AgriciDaniel/claude-seo](https://github.com/AgriciDaniel/claude-seo) for the full audit. Three supporting Skills cover narrower parts of the workflow: [zubair-trabzada/geo-seo-claude](https://github.com/zubair-trabzada/geo-seo-claude) for AI-search validation, [lionkiii/claude-seo-skills](https://github.com/lionkiii/claude-seo-skills) for broader technical checks and [Canonry/aeo-audit](https://github.com/Canonry/aeo-audit) for answer-engine readiness. The skills speed up discovery. People still decide what matters and in what order.

## **TL;DR**

- Claude Skills run inside Claude Code and analyze a site crawl to produce structured, categorized findings with severity levels.
- A first-pass audit covers technical SEO, schema and structured data, Core Web Vitals, crawlability, llms.txt and AI-crawler access, content quality and answer-engine readiness.
- Our workflow uses claude-seo for the primary pass. The other Skills support narrower checks: AI-search validation, broader technical review and answer-engine readiness.
- The skills accelerate the discovery phase. Prioritization, dependency mapping and sequencing stay with your specialists.
- A useful audit report ships findings already ordered by severity, tied to business impact and mapped to a build order.

Get a free SEO + GEO audit of your website

We’ll crawl your site, score its search and AI-visibility readiness, and email you a detailed PDF report — free.

SEO work changed shape. For years a technical audit meant crawling a site, reading findings by hand and working through a backlog one issue at a time. Now a site has to rank in classic search and stay legible to AI crawlers, surface in AI Overviews and answer engines, and hold up on a growing set of discovery channels at the same time. That double requirement raises an old question with new weight: how do you audit a large site both thoroughly and fast, with precision intact?

Claude Skills are one answer to the speed half of that question. They run inside Claude Code, read a site crawl and return categorized findings in hours. This article explains how Darwin uses claude-seo for first-pass SEO and GEO audits and highlights three complementary public Claude Skills for more specialized validation tasks.

## **What Has to Be True Before an Audit Becomes a Roadmap**

A skill can list two hundred issues in an afternoon. That list is not yet a plan, and the gap between the two is where most audit value is won or lost. Four things have to happen in order for a raw crawl to become work a team can ship.

First, the problems have to be exposed. A crawl pulls buried technical debt into view: blocked crawlers, broken redirect chains, schema that never validated, pages quietly excluded from indexing. You cannot sequence what you cannot see, so surfacing the full set of signals comes first.

Second, the findings have to connect to something. An isolated issue means little until it is tied to Google Search Console data, to your MarTech stack, and to the business goal the page serves. A schema gap on a high-traffic template and the same gap on a dead archive page are two different problems, and only the connections tell you which is which.

Third, the noise has to resolve into a clear signal. Severity and business impact separate the two or three findings that move rankings and AI visibility from the long tail of minor fixes. Clarity here is the difference between a roadmap and a backlog.

Fourth, the work has to move. Findings become a sequenced set of fixes with owners and dependencies, so the audit turns into shipped changes and not a document that ages in a drive. Each of the sections below sits on one of these four moves.

![The image is an infographic that illustrates the process of human judgment and how it connects to skill discovery. It features two columns with different colored squares representing skills and judgment. The left column displays a green square labeled "Skill Discovery" while the right column has a purple square labeled "Human Judgment". Each square contains text explaining the connection between these two aspects, providing an organized visual representation of this relationship.](https://cdn.sanity.io/images/qd0fa73p/production/1d8b40c6333623f77bbe966460638553df7360d3-2280x1124.png?w=1140&q=85&auto=format)

## **Why Claude Skills Have Become Part of Modern SEO Audits**

The discovery phase is the slow part of any audit. A site with thousands of pages, hundreds of possible issues and several interdependent technical systems used to take days to review by hand. Claude Skills compress that phase. They read the crawl, work through it methodically and return findings grouped by category and severity, so a team can run a full first pass and see problems in technical SEO, schema, crawlability, AI-crawler access and content quality in one workflow. That combined coverage matters because classic SEO and AI search run on shared ground. Google’s own Search Central guidance is explicit that established SEO practices still apply, since its generative AI features in Search run on the same ranking and quality systems that power classic results.

The aim is a division of labor. The skill handles pattern recognition at scale, which people find tedious and slow. Your specialists handle judgment: weighing each finding against business goals, the existing stack and the competitive position. Speed on one side, discernment on the other.

The mechanics are simple. Your site gets crawled to capture its current state. The skill analyzes that crawl and flags technical debt, AI-crawler access problems, schema gaps and readiness gaps for AI-driven search. What comes back is a structured, categorized set of findings ready for review, ordered by severity and impact, with a recommended next step attached to each one.

Get a free SEO + GEO audit of your website

We’ll crawl your site, score its search and AI-visibility readiness, and email you a detailed PDF report — free.

## **What [claude-seo](https://github.com/AgriciDaniel/claude-seo) Checks in One Pass**

When a site goes into claude-seo, the Skill runs a crawl and analysis in one pass, surfacing issues across technical SEO, content, performance, and AI search readiness.

AgriciDaniel/claude-seo is the primary Claude Skill Darwin currently uses for first-pass SEO and GEO audits, providing the team with a structured, prioritized set of findings to review and act on.

### **The Categories a First Pass Covers**

claude-seo examines a site over several categories, each tied to modern SEO and AI-search visibility. Technical SEO covers crawlability problems, redirect chains, broken internal links, duplicate-content signals, sitemap errors and robots.txt rules that block search engines or AI crawlers. Schema and structured data covers missing or malformed markup, incomplete Organization or LocalBusiness data and Rich Result eligibility gaps. Core Web Vitals covers Largest Contentful Paint (LCP), Cumulative Layout Shift (CLS) and mobile-usability signals.

On the AI side, the pass checks [llms.txt](https://www.darwinapps.com/blog/beyond-seo-how-geo-optimization-prepares-your-brand-for-ai-first-search/) configuration and [AI-crawler access](https://www.darwinapps.com/blog/increase-ai-citations-guide/), so you learn whether answer engines can reach and cite your content. It also flags content-quality signals like thin pages, missing H1 tags and topical gaps, and it scores answer-engine readiness: whether pages give direct, extractable answers that AI Overviews can quote.

### **Example Findings from a First Pass**

The findings arrive as concrete, categorized items with a severity level and a recommended fix. Below are simplified examples of the kinds of findings a first-pass audit can surface, with identifying and implementation details removed.

**Critical, crawlability and indexing.**Paginated archive pages each declare a canonical pointing back to page one and share the same title and social metadata. Search engines read many near-identical URLs all claiming to be the same page, and deeper posts may not get credited. The fix is to give each paginated page its own canonical, a distinct title and a correct URL in the metadata.

**Critical, schema and AI-crawler accessibility.**The site-wide Organization markup omits the postal address and telephone that appear in the page footer. This weakens the entity signal that both Google and AI systems use to identify and cite the brand, and can hold back Article rich results. Adding the full address, telephone and a contact point to the Organization data lets the brand resolve as a single, trusted entity.

**Medium, Core Web Vitals and AI-search readiness.**The main hero image is not preloaded and the CDN response is not cached at the edge, so the largest element paints late for both users and crawlers indexing the page. Preloading the hero as an image resource, and confirming the CDN serves cached copies, pulls the LCP element forward and speeds up how quickly content becomes legible to AI crawlers.

Each item names the category, assigns a severity, states the problem in plain terms and ends with a specific next step. That last part earns its keep: a finding that tells you what to do next is a task, while one that only names a problem leaves you homework.

![The image shows a screenshot of a computer screen displaying a list of criteria and recommendations for a website's SEO. The screen is divided into three sections with each section containing different pieces of information about the website's optimization process. There are several options available to improve the website's ranking, such as optimizing images, using structured data, and ensuring proper page loading speed. The image also includes text that provides further details on how to implement these recommendations effectively.](https://cdn.sanity.io/images/qd0fa73p/production/974a6e86d1269e36a43710059ca4686d579d738e-2280x1870.png?w=1140&q=85&auto=format)

**Key takeaway:**The first pass earns its place on scope and speed. In one cycle you get coverage of every category, severity flags that rank issues by impact, findings grouped so bulk fixes are obvious, explicit AI-crawler and llms.txt checks, and a concrete next step per item. From there the work shifts from discovery to deciding what to fix first.

Get a free SEO + GEO audit of your website

We’ll crawl your site, score its search and AI-visibility readiness, and email you a detailed PDF report — free.

## **Extending the Audit with Specialized Claude Skills**

We tested these skills based on their publicly available capabilities and documentation. The workflow described here reflects how our team currently approaches first-pass SEO and GEO audits. AgriciDaniel/claude-seo remains our primary audit Skill. The three Skills below are complementary: each answers a narrower question that the primary pass raises and leaves open.

The first pass tells you what is broken. These skills answer what it means in one specific situation. The craft is knowing when to reach for each one and what decision it feeds.

### **AI-Search Validation with [geo-seo-claude](https://github.com/zubair-trabzada/geo-seo-claude)**

**When to apply:**after the primary scan flags possible GEO or AI-crawler issues.

The [geo-seo-claude](https://github.com/zubair-trabzada/geo-seo-claude) skill, based on its public documentation, goes deeper on [AI-search readiness](https://www.darwinapps.com/blog/generative-engine-optimization-geo-what-marketers-need-to-know-in-2025/). It scores whether your content is structured so AI systems can extract and cite it, checks that your pages and metadata are reachable by crawlers like ClaudeBot, GPTBot and PerplexityBot, and analyzes llms.txt setup. It goes deep on the AI-visibility gaps the primary audit only flags. One honest note: claude-seo itself treats llms.txt as a disclosure signal whose ranking value remains unproven, and follows Google's position that AI Overviews run on the same systems as classic search. So use this skill to validate access and citability, and keep llms.txt in its place as an access signal.

### **Broader Technical Checks with [claude-seo-skills](https://github.com/lionkiii/claude-seo-skills)**

**When to apply:**during roadmap development, when you need wider operational context.

The [claude-seo-skills](https://github.com/lionkiii/claude-seo-skills) set, based on its public documentation, is a broad toolkit of SEO commands for Claude Code. It appears useful for broader SEO operations and follow-up validation once the primary pass is done. Reach for it when you want to cross-check the primary findings against live search data before you commit fixes to the roadmap, so recommendations tie back to solid evidence.

### **Answer-Engine Readiness with [aeo-audit](https://github.com/Canonry/aeo-audit)**

**When to apply:**when AI Overviews or [answer-engine readiness](https://www.darwinapps.com/blog/best-aeo-tools-2025/) is a business priority.

The [aeo-audit](https://github.com/Canonry/aeo-audit) tool scores a site on the factors that decide whether answer engines cite it: structured data, extractability, snippet eligibility and self-contained answer blocks. Where the primary pass surfaces content structure, this one goes deeper on how content behaves in answer-engine contexts. Bring it in when AI Overviews represent real traffic potential, or when clients compete in verticals where answer engines are already reshaping who gets seen.

**Key takeaway:**Together with AgriciDaniel/claude-seo, these supporting Skills move the workflow from initial discovery to validated, context-aware recommendations. GEO strategy gets checked against real crawler access, technical fixes get checked against Search Console reality, and content gets checked for answer-engine fit. People then merge the results into one prioritized sequence.

## **Turning Audit Findings into an Implementation Roadmap**

Raw findings, however thorough, are a starting point. The value shows up when a team reviews, contextualizes and sequences them into a prioritized roadmap that fits real business goals, technical constraints and available people. This is the human layer, and it is what turns AI-generated insight into strategy.

When a skill surfaces dozens or hundreds of issues on a large site, the pull is to fix everything at once. Specialist judgment is what resists that. A critical technical issue on a high-traffic template carries different weight than the same issue on a low-traffic archive page. A schema gap that blocks AI-crawler access matters more in a vertical already showing AI Overviews than in one that is not.

The work tends to move in stages. Review the findings and assign severity by scope and likely impact. Weigh business impact: which issues touch rankings, AI-search visibility or the experience of your target audience. Map dependencies: does a fix need development coordination, or wait on a data migration already in flight. Then sequence, so early wins respect the stack's readiness and the longer-term data plan, so they avoid creating debt downstream.

The sequence guards against a common reflex. As SEO consultant [Aleyda Solis](https://www.aleydasolis.com/en/ai-search/ai-search-optimization-checklist/) notes, *“Optimization should start from observed presence, not from a reflex to create more content.”* The roadmap decides what the evidence justifies, in what order, against real business goals.

### **Decision Criteria for Prioritization**

When specialists rank findings from an AI-assisted audit, a handful of criteria do most of the work:

- **Scope and severity:**how many pages or users are affected, and whether it breaks function or merely improves it.
- **Search-visibility impact:**whether the finding touches rankings, crawlability or AI-search indexing.
- **Business alignment:**whether fixing it advances the SEO or GEO goal that matters this quarter.
- **Dependency mapping:**whether the change waits on data-structure work, MarTech integrations or the development roadmap.
- **Technical feasibility:**whether current resources can ship it, or it needs outside help or real infrastructure work.
- **Quick win versus strategic play:**whether it resolves in days or calls for longer-term investment.

![The image is a flowchart that illustrates the process of prioritizing finding and implementing solutions to problems. It shows six steps, each with its own priority level, guiding users through the decision-making process for choosing the best solution. The flowchart is color-coded, making it easy to follow along as it progresses from one step to another.](https://cdn.sanity.io/images/qd0fa73p/production/7551f669a89750fea8eb6ffabb400bad7c56b464-2280x2772.png?w=1140&q=85&auto=format)

### **From Findings to Priorities**

A short example shows the human layer at work. Say an audit finds that most product pages carry incomplete Product schema, missing the offers, aggregate-rating and review properties. The finding is simple. The decision needs context.

Suppose the team recently stood up a new product-information-management system and is migrating legacy product data into it in stages, while watching AI-search results for product queries in a vertical where competitors are starting to appear in AI Overviews. The call: high priority, sequenced in two phases. Phase one this sprint covers the products already in the new system, validated against Search Console with AI-search impact monitored. Phase two next quarter migrates the rest and applies schema systematically. That respects the migration timeline, delivers early proof and lines schema work up with the longer data-governance goal.

The implementation note writes itself: generate schema automatically from product attributes so maintenance stays low, and add a data-quality check in the publishing workflow to catch incomplete markup before it ships. This is the human layer acknowledging the technical finding while grounding it in the organization's reality.

**Key questions specialists ask when prioritizing:**does this support our SEO or GEO goal this quarter, and if not, does it unblock something that does? Who owns it and what is the effort? What is the dependency chain? How does it fit the data strategy, saving rework later or adding debt? What is the validation plan? And is there a quicker path to similar value, since a 70 percent fix that ships in days can beat a perfect one that takes months?

Get a free SEO + GEO audit of your website

We’ll crawl your site, score its search and AI-visibility readiness, and email you a detailed PDF report — free.

## **What a Good First-Pass Audit Should Deliver**

When an AI-assisted audit report lands, you need a fast way to tell a real deliverable from a data dump. A strong first pass hands you findings already shaped for implementation. Five things separate the two.

**Prioritized findings with clear severity.**Every issue carries a rating, from critical to low, based on impact to search visibility, experience or crawlability. That is what lets you tell a broken robots.txt file apart from one missing alt attribute, and stops the debate about what matters before it starts.

**Implementation order tied to dependencies.**Recommendations should arrive in sequence, so you clear indexing blockers first, then optimize content, and resolve schema errors before you build on top of them. A good report maps the dependencies so work happens in the right order and rework stays low.

**Business context per recommendation.**A generic “improve Core Web Vitals” does not drive action. A useful finding explains why the issue matters here, tying mobile page speed on key templates to a measurable conversion effect, for instance, so the recommendation becomes a decision.

**Concrete next actions and owners.**Each finding should say what happens next and who does it, from “allow AI crawlers in robots.txt” to “add FAQ schema on support pages.” Vague advice leaves teams guessing.

**AI-search readiness signals.**The report should flag what classic audits skip: misconfigured llms.txt, citability problems, AI-crawler restrictions and answer-engine gaps. These are now part of being found.

**Key takeaway:**A useful audit does more than list problems. It tells you what to fix, why it matters, how to sequence it and who owns each step. If the report leaves those questions open, it has not delivered enough context. Look for findings already filtered for impact, ordered for efficiency and connected to your goals. That is what turns discovery into action.

Want an audit that hands you priorities and a build order?

## **Where This Fits into Darwin's Work**

When a first-pass audit finishes, the hard part starts. The report lists two hundred findings, and nothing in it flags what touches revenue, what depends on a migration already in flight, or what order keeps early fixes from creating debt later. The skill surfaced the problems fast. Turning them into a sequence a developer, an analyst and a CFO all sign off on takes human judgment.

Darwin works with B2B SaaS marketing and revenue teams to close that gap. That means tying each finding back to the [MarTech stack](https://www.darwinapps.com/integrations-automations/) it lives in, checking it against real [marketing data readiness](https://www.darwinapps.com/blog/marketing-data-readiness-for-ai-agents-what-to-fix-before-automating-analytics-workflows/), and building the [prioritized roadmap](https://www.darwinapps.com/ai-readiness-enablement/) that turns a raw audit into shipped changes with owners and a build order.

A first-pass audit earns its value once someone decides what the findings mean. It gives a team a shared starting point: what surfaced, why it matters, what depends on other work and what can move first. If your SEO and AI-search work keeps stalling in the handoff between audit and roadmap, that is the part worth building first.

### **Pairing Skill Speed with Human Judgment**

The strongest audits pair speed with judgment. Claude Skills bring the speed, crawling thousands of pages and surfacing technical SEO issues, schema problems, Core Web Vitals signals and AI-search gaps faster than manual work alone. Speed alone misses the goal: a short list of business-aligned priorities that move results.

That is where people come in. A team reviews the findings, weighs business context, maps dependencies against the MarTech stack and sequences the work in an order that fits the organization. A flagged issue becomes a roadmap item only when it lines up with capability, timeline and strategy. AI acceleration plus human judgment is what turns a long list into a plan you can ship.

Get a free SEO + GEO audit of your website

We’ll crawl your site, score its search and AI-visibility readiness, and email you a detailed PDF report — free.

## FAQs

**Q2. What is the difference between SEO and GEO in these audits?**

SEO covers ranking and crawlability in classic search. GEO covers visibility in AI-driven search: whether AI crawlers can reach your content and whether it is structured to be cited by answer engines. Google's own position is that AI Overviews run on the same ranking systems as classic search, so the two overlap more than they compete.

**Q2. How does the EU AI Act affect SaaS companies using third-party AI APIs?**

SaaS companies using AI APIs are usually deployers under the EU AI Act, with duties beyond data residency. They need to document AI use, run risk assessments, add logging and traceability and stay transparent with users. Most B2B SaaS falls under limited risk, which centers on transparency.

**Q3. What does llms.txt do?**

llms.txt is an emerging file that discloses what content is available for AI systems to read. It works as a disclosure and access signal. Current evidence stops short of treating it as a direct ranking or citation lever. Audit it for access, and hold ranking claims to the evidence.

**Q4. How long does a first-pass audit take?**

The discovery phase runs in hours, where a manual review of a large site would take days. Prioritization and roadmap building add human time on top, which is why a full audit turns around in a matter of days.

**Q5. What should a SaaS team prioritize first when implementing AI privacy controls?**

Start where risk is highest. Teams processing sensitive data at scale should prioritize encryption and isolation, while teams leaning on third-party AI should audit vendor compliance first. For most, the practical order is to inventory AI features, classify risk and add logging and transparency.

### Not sure whether your site is ready for AI search?

A first-pass audit shows you exactly where classic SEO and AI-search readiness stand, and what to fix first.

![Andrei Kazhala](https://cdn.sanity.io/images/qd0fa73p/production/9da342d59bcb82c7e0785c35bf83128a41315c5e-840x840.jpg?w=420&q=85&auto=format)

###### You might also like

## [15 Best SEO Automation Tools Actually Worth Using in 2026](https://www.darwinapps.com/blog/15-best-seo-automation-tools-2026/)

## [How to Rank in Perplexity AI: Complete Guide (2026)](https://www.darwinapps.com/blog/how-to-rank-in-perplexity-ai-complete-guide-2026/)

## [Best Rank Tracking Software for 2025: Complete Guide](https://www.darwinapps.com/blog/best-rank-tracking-software-2025/)