← All use cases

Use case B35 · Business

Have AI audit your website for search and AI answers, then work through the fixes it recommends together

Charlie Barmore ·

The short version

What I do

I point an AI coding agent at my website’s code and ask it for an SEO and AEO audit first, with no changes. AEO (answer-engine optimization) means making the site easy for ChatGPT, Claude, Google’s AI answers and the like to read, quote and credit correctly. The audit comes back as a list: what’s wrong, why it matters, and the fix. Then I go through it with the agent one item at a time. On my site the fixes were: one file that holds the facts about me, so every page and the search data say the same thing; structured data that ties each page back to me; short FAQ answers written so they can be lifted word for word; a robots.txt that welcomes the AI crawlers by name; a sitemap; and an llms.txt summary generated from the pages. I did the same on a second site I worked on.

What it gets me

An audit an agency would charge for, done in an afternoon, and fixes I understand because I approved each one. The site now describes me the same way everywhere, and it stays that way: when I change a page, the summary files rebuild with it.

Prompt

I own [business name], serving [customers and location]. My website’s code is in this folder and the live site is [URL]. These are the confirmed facts about my business: [name, services, credentials and location]. Audit how well search engines and AI assistants can find, read and describe it.

1. Check titles, descriptions, headings, internal links, page speed, structured data, sitemap and robots.txt. Explain technical terms in plain English.
2. Check whether pages answer customers’ questions clearly and whether business facts agree across pages.
3. Return a table ranked by likely importance: page, problem, evidence, why it matters, and the fix you propose.

Make no changes yet. Don’t invent credentials or services, or promise better rankings. Mark anything you couldn’t inspect. After I choose a fix, work through it with me one item at a time.

Do it yourself

These are suggested steps for you, not a copy of my exact setup.

What you need

  • Your website’s code in a folder an AI coding agent can work in, such as Claude Code. If your site is on a builder like Squarespace or Shopify, a chat assistant can audit the live pages instead, and you make the fixes in the builder.
  • Google Search Console set up for your domain (free), so you can see the before and after.

Start simple

  1. Run the audit prompt above. Ask for the report only, no changes.
  2. Read the list and pick the top three items you understand and agree with.
  3. Ask the agent to fix those three on a new branch, then preview the site and check what changed.
  4. Publish, then look the pages up in Google Search Console and ask an AI assistant a question your customers would ask. Note what it says about you.

Take it further

  • Put the facts about your business in one place in the code and have every page, and the structured data, read from it, so they can’t drift apart.
  • Write a short FAQ: the questions customers actually ask, each answered in two or three sentences that name your business and stand on their own.
  • Run the same audit every month, or after a big change, and compare it with the last one.
  • Add your new-page rules (a clear title and description, an answer near the top, alt text on images) to the agent’s notes file, so new pages start out right.

Check your work

  • Google says there are no special files or markup needed to appear in its AI answers: the basics are being indexed, letting crawlers in, and useful pages. Treat llms.txt as a nice extra, not a fix; studies in 2026 found most AI crawlers rarely request it.
  • Read every sentence the agent writes about your business. It will happily invent a credential or a service.
  • Make sure the structured data says only what the visible page says.
  • Check results over weeks, not days, and don’t trust a grade the agent gives itself.

Others did this too

  • Mika Reyes, co-founder at King’s Cross Labs, ran the same audit, fix and re-audit loop on her site with Claude Code and reported search impressions up about 300% after the fixes (her own numbers, one site): https://mikareyes.com/ai/how-to-audit-website-seo-aeo-with-ai
  • Google’s guidance on AI features and your website: https://developers.google.com/search/docs/appearance/ai-features
  • A study of which businesses AI assistants name. AIWorthy, which sells AI-visibility work, asked five assistants (ChatGPT, Claude, Gemini, Perplexity and Google AI Mode) 10 buyer questions about one service, Single Audits by CPA firms in North Carolina’s Research Triangle: 500 answers in total, checked against the federal record of who actually did the work. What the cited pages had in common: “every page cited on both dates named the Triangle, between 11 and 32 times; described the buyer’s own service; and ran to real length.” It also found stale listings cost firms: “In 79 of 250 answers, the buyer was told to call a firm that no longer exists under that name.” The study is vendor research, not a business running the audit, and says it “does not show that any of these produces a recommendation, and it does not show that a recommendation produces a client.” The study, 2026-10-07; Accounting Today’s coverage.

I use this. My own setup. I use it in my own work. Prompt adapted for this library, inspired by the case. Business size: One-person firm. Results are quoted as the source claimed them and never improved. Spot an error or a broken link? Email charlie@cbarmorecpa.com.

More in Marketing

New cases go out every Saturday in Inside AI.