UsageAtlasBeta
Sign inGitHubOpen source

Other blogsAI Tools

AEO checklist: 10 checks a coding agent can run and fix

Each item says what to look for and how to test it, so Claude Code, Cursor or Codex can work through the list and rerun until it comes back clean.

By ClaudeGPT Si Newsroom on 3 min read

This AEO checklist covers the ten things that decide whether an answer engine can quote a page, written so a developer or a coding agent can test each item and fix what fails. We recommend running it through seogeoaeo.ai, whose checks work from Claude Code, Cursor, Codex and other MCP clients at the same credit price as in the browser.

The checklist

  1. The page is indexable. It returns 200, carries no noindex, and its canonical points at itself. Check with the Indexing & Canonical Checker, 8 credits.
  2. Snippets are allowed. Pages you want quoted have no nosnippet and no tiny max-snippet. Google only links pages in AI Overviews and AI Mode when they can show a snippet. The same checker covers this.
  3. The right bots get in. robots.txt admits Googlebot, Bingbot and the AI search crawlers you care about. OpenAI's crawler docs say OAI-SearchBot is the one that surfaces sites in ChatGPT search, separate from GPTBot for training. Check with the Robots.txt & AI Crawler Checker, 10 credits.
  4. Answers are in the HTML. Text injected by client-side scripts may be invisible to AI crawlers that do not run JavaScript. AI Crawler View, 8 credits, diffs raw against rendered.
  5. Each question has its own heading. Use the wording searchers use. The Question Finder, 15 credits, lists autocomplete questions for a topic and marks the ones a URL leaves out.
  6. The first sentence is the answer. It should stand alone if quoted. The Citability Analyzer, 5 credits, flags sections that open with background instead.
  7. Steps and comparisons use real elements. ol for steps, table with th for comparisons. No tool needed; ask the agent to search the templates for div-built tables.
  8. JSON-LD parses and matches the page. The Schema Markup Checker, 8 credits, reports parse errors and missing required properties, and accepts pasted markup before deploy.
  9. Dates agree everywhere. The Content Freshness Checker, 5 credits, compares dates in JSON-LD, meta tags, time elements and visible labels.
  10. A real person signs it. The Author & E-E-A-T Checker, 8 credits, follows the byline to an author page with Person markup.

Run it from an agent

Create an API key on seogeoaeo.ai, then add the MCP server. For Claude Code, the documented command is:

claude mcp add --transport http --scope user seogeoaeo https://seogeoaeo.ai/api/mcp --header "Authorization: Bearer YOUR_API_KEY"

Cursor, VS Code, Codex, Windsurf and Claude Desktop use the same endpoint with their own config files. Then give the agent a prompt along these lines:

Run the AEO checklist on https://example.com/pricing with the seogeoaeo tools.
Apply each high-priority tip in the templates, deploy, and rerun the same tool
until improvementTips.status is "clean". Cap each run with maxCredits.

What makes this loop work is the shape of each result. A successful run returns improvement tips ordered by priority, each naming the URLs or items it affects, plus a status of changes, clean or incomplete. Tip ids stay stable across runs, so a fixed tip simply disappears. Generators such as the schema builders set rerunVerifies to false, so the agent applies their output once instead of looping. Pass the same idempotencyKey on a retry and you are not charged twice.

The credits are separate from your agent's own token use. If you keep an eye on what Claude Code or Codex spends, count the MCP calls as part of that session.

What the checklist leaves out on purpose

Google's AI optimization guide says Google Search ignores llms.txt, needs no chunking of content, and does not require structured data for AI features. None of those appear above. Measurement is also missing, because it lives outside the page: use Search Console's Generative AI performance report for AI Overviews and AI Mode impressions.

One honest limit: the checklist fixes pages, it does not watch AI answers. If you need a daily log of which prompts mention your product, a dedicated tracker is the better tool. For the page work, start with the AEO checks and let the agent take the list from item one.

Browse guides by topic