Claude Sonnet 5: How I Run a Whole-Site AEO Audit for $8

I just ran a full AEO audit on a 200-page website in a single pass. It took 12 minutes and cost me less than eight dollars. This wasn’t possible last week. What changed is that this work used to require Claude’s top-tier Opus model, chunked into multiple sessions, at a cost closer to $40.

Anthropic just released Claude Sonnet 5, and it’s now the default model in Claude Code. It has a one-million-token context window and near-Opus performance for agentic tasks, all at half the price. This post is the how-to for turning that new power into a service you can sell today.

---

What is Claude Sonnet 5?

Claude Sonnet 5 is a large language model from Anthropic, released in June 2026, that serves as the new default for Claude Pro and Claude Code. It offers performance close to the premium Opus model for complex coding and agentic tasks but at a significantly lower cost. Its primary feature is a one-million-token context window, which allows it to process and reason over entire codebases or websites in a single conversation.

---

The AEO Audit Prompt I use — copy it right now

This is the exact configuration and prompt I use in Claude Code to run a full-site audit. The first line sets Sonnet 5 as the default model for your session. The prompt then instructs Claude to act as an agent, crawl a sitemap, and analyze every single page for key AEO signals and schema markup, then output a prioritized fix list.

# Step 1: Set Sonnet 5 as your default model in Claude Code
/config default_model claude-sonnet-5

# Step 2: Run the audit prompt
Task: Perform a comprehensive AEO and Schema.org audit on my website.

Context:
My website's sitemap is located at: [PASTE YOUR SITEMAP.XML URL HERE]
The primary audience is [describe your audience, e.g., 'design agency owners'].
The main goal of the site is [describe your goal, e.g., 'to book 1-on-1 working sessions'].

Execution Steps:
1.  Fetch and parse the sitemap.xml to get a list of all URLs.
2.  For each URL, retrieve the full page content.
3.  Analyze each page for the following AEO signals:
    - Presence and quality of Schema.org markup (Article, FAQPage, etc.).
    - Use of declarative, answer-first paragraphs under H2 headings.
    - A clear, single topic per page.
    - Recency signals (e.g., "Last updated" dates).
    - Internal linking to pillar pages like `/blog/aeo-pack/`.
4.  Consolidate your findings into a single report.
5.  Structure the report as a markdown table with columns: [URL], [Current Schema], [AEO Score (1-10)], [Top 3 Recommended Fixes].
6.  Prioritize the list by the pages with the lowest AEO score and highest potential impact first.

Before Sonnet 5, a prompt this large would fail or require splitting the site into 5-10 separate runs. Now, it just works.

+----------------------+      +----------------------+      +-----------------+
| Entire sitemap.xml   | --> | Claude Sonnet 5      | --> | Prioritized     |
| (200+ pages)         |      | (1M Token Context)   |      | AEO Fix List    |
+----------------------+      +----------------------+      +-----------------+

---

Here’s exactly how I’d do this

My process for running a full-site audit with Claude Sonnet 5 is a simple, four-step workflow inside Claude Code. The key is providing the agent with a clear entry point (the sitemap) and a structured set of instructions for analysis and output. This turns a multi-day manual process into a 15-minute automated one.

1. Find the sitemap. I locate the website’s `sitemap.xml` file. For most WordPress sites, it’s at `domain.com/sitemap_index.xml`. For other platforms, it’s usually `domain.com/sitemap.xml`. This is the map Claude will use to find every page.
2. Configure Claude Code. I open a new session in Claude Code and run the command `/config default_model claude-sonnet-5`. This ensures I’m using the new, cheaper model with the large context window for the task.
3. Customize and paste the prompt. I take the asset from this post and fill in the three bracketed context fields: the sitemap URL, the target audience, and the site’s primary goal. This context helps Claude make smarter recommendations.
4. Run the agent and analyze the output. I hit enter and let the agent work. It will show its plan, then execute it, fetching and analyzing each page. In about 10-15 minutes, it delivers a clean, prioritized markdown table of fixes that I can hand directly to a client or developer.

---

What this changes for designer-run agency work

The shift from chunked, expensive audits to single-session, cheap ones fundamentally changes the economics of AEO services. What was once a high-ticket, multi-day project is now a fast, scalable, and even productized offering. The value moves from manual labor to the strategic interpretation of the AI’s output.

Dimension	Old way (Opus, chunked)	New way (Sonnet 5, single session)
Scope	Audit 10-15 key pages, manually selected.	Audit the entire site (200+ pages) in one go.
Cost	~$40+ in API credits for a medium site.	Under $10 for the same site.
Workflow	Split sitemap, run multiple Claude sessions, merge reports by hand.	One prompt, one session, one unified report.
Turnaround	2-3 days of combined human and machine time.	Under 30 minutes from start to finish.

This change means I can now include a full AEO audit as a standard part of every website proposal instead of an expensive add-on. It’s a massive value-add that costs me almost nothing to produce.

---

My $0.02 — How I’d roll this out

I wouldn’t just use this new tool; I’d build a specific, sellable service around it. This is a perfect opportunity to productize expertise. This is the three-day plan I’d use to go from learning the technique to selling it as a fixed-price package.

Day 1 — Audit my own house. Before I sell this to anyone, I’d run the full AEO audit on my own site, mikekwal.com. This does two things: it debugs my process and gives me a real, tangible asset to show potential clients. I’d document the before-and-after, turning my own site into the case study.

Day 2 — Create the ‘AEO Snapshot’ package. I’d package this audit as a $499 one-time service. The deliverable is the prioritized fix list generated by Claude, plus a 30-minute screen-share with me to walk through the top three priorities. I’d build a simple landing page for it, using the report from Day 1 as the primary example.

Day 3 — Market it to my existing network. I’d send a simple email to my past clients and my current newsletter list: “Want to see how AI engines grade your website? I’m offering a new $499 AEO Snapshot.” The low price point makes it an easy impulse buy and a great entry point for larger projects. When you can deliver the work in 15 minutes, the margins are incredible.

---

FAQ

Is Claude Sonnet 5 better than Opus?
For tasks like coding and running agentic workflows, Sonnet 5’s performance is near Opus 4.8 levels at a much lower cost. For tasks that require maximum accuracy and nuanced reasoning on extremely complex topics, Opus 4.8 is still superior. For the AEO audit, Sonnet 5 is the better tool because of its cost-effectiveness.

What is a one-million-token context window?
A one-million-token context window means the AI can hold about 750,000 words—or roughly 1,500 pages of text—in its working memory at once. This allows it to analyze an entire medium-sized website, a full codebase, or a long legal document without losing track of the details from beginning to end.

Do I need a paid Claude plan for this?
Claude Sonnet 5 is the default model on both the free and Pro tiers of the consumer-facing Claude.ai. However, to run agentic workflows like this one, you need to use Claude Code, which is part of the paid Claude Pro or Team plans. The cost of running the audit is separate from the subscription fee.

How much does an audit like this actually cost?
The cost is based on token usage. With Sonnet 5’s promotional pricing ($2 per million input tokens), analyzing a 200-page site with an average of 1,500 words per page is about 300,000 words or ~400,000 tokens. The total cost for input and output tokens would be under $10.

Can I use this for things other than AEO audits?
Absolutely. The same principle applies to many other tasks. I use it to perform brand voice consistency checks across an entire blog, to review a whole codebase for security vulnerabilities, or to summarize a large batch of customer feedback documents into key themes. The large context window is a general-purpose tool.

What is Claude Code?
Claude Code is Anthropic’s integrated development environment (IDE) for building with AI. It’s a workspace where you can write, edit, and run code with Claude as a direct collaborator. It has a built-in terminal and filesystem, allowing the AI to execute commands and manage files, which is what makes these agentic audits possible.

Does this replace SEO tools like Ahrefs or SEMrush?
No, it complements them. Tools like Ahrefs are best for off-page SEO analytics like backlink profiles and keyword ranking. This Claude workflow is for on-page technical AEO—analyzing the actual content and structure of your site to make it more citable by AI engines. I use both as part of my full Claude Toolkit.

---

Want help applying this?

Four ways to go deeper:

• Build with Builders. Join the Talk-to-Build community to learn to build AI-native websites, cinematic AI video, and agent-driven workflows you can sell.
• 1-on-1 working session. Book a screen-share with me — bring a real problem, leave with a working piece of it.
• Done-for-you. MK-Way builds AEO-ready websites, apps, and AI agent workflows.
• Quick question. DM me on Instagram or LinkedIn. I read every message.

---

Part of the AI Pulse series. If you commented “CLAUDE” on one of my videos — this is the breakdown.

Last updated: 2026-07-09.