AI-Visibility Checklist

Audit

Before you change anything, get an honest read on where you stand today.

• Define your prompt universe

  Write down the 20-50 questions a buyer in your category actually asks an AI assistant. These are the prompts your AEO program is optimizing for. Without them, every other metric is generic.

• Measure your current AI visibility

  Run each prompt against ChatGPT, Claude, Perplexity, and Gemini. Record who gets cited, who gets mentioned, and where you appear (or do not). This is your baseline.

• Identify your top three competitors in AI answers

  They are often different from your top SEO competitors. Note which surfaces they win on — their own domain, Reddit, Quora, podcasts — so you know where to invest.

Structure

Make your site easy for retrieval systems to parse, score, and trust.

• Add schema.org JSON-LD to every public page

  At minimum: Organization on the site, WebPage on every page, Article on long-form content, Product on commercial pages, FAQPage where you have real Q&A. Use JSON-LD, not microdata.

• Use semantic HTML with one h1 per page

  Real `<article>`, `<section>`, `<main>`, `<header>`, `<nav>` elements. One `<h1>` that answers what the page is. `<h2>` and `<h3>` for the major and minor sections. No skipped levels.

• Publish an llms.txt manifest

  Add /llms.txt listing your most important content. Treat it like a curated sitemap for language models. Update it when you ship a major new section.

• Allowlist AI crawlers in robots.txt

  Explicitly permit GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and the others you want to be retrieved by. Default-deny crawlers cannot cite you.

Publish

Create content shaped for how language models extract and cite.

• Lead every section with the direct answer

  First sentence answers the implicit question. Supporting context comes after. Retrieval systems often extract only the lead sentence as the candidate response.

• Build authority on third-party surfaces

  Reddit threads, Quora answers, podcast appearances, conference talks, guest posts. Language models train on these. Presence there compounds independently of your own domain.

• Attribute experts by name and credential

  Author bylines, About pages with real biographies, structured Person schema linking each author to their public profiles. Anonymous content is harder to trust and harder to cite.

Monitor

AEO is continuous. Track the trend, not a snapshot.

• Re-run your prompt universe weekly

  Citations move week to week as models update and competitors ship. A monthly cadence is too slow to catch and respond to drift.

• Track citation velocity, not just citation count

  A rising total can hide a flat trend if competitors are accelerating. Velocity — new citations per week — is the leading indicator. Set a target and review it weekly.

Next step

Find a gap? The how-to guides walk through the fixes one at a time. New to the terminology? The AEO glossary defines every term used here.