← Free resources

The AI answer visibility checklist

An AI assistant cites a page it can parse, attribute and trust. This checklist is the twelve things we check first on any site, in the order we check them, and what a failure looks like in each case.

Time to use
20 minutes per page
What you end up with
A scored list of what is blocking one page from being cited, ordered by how much each fix moves it.

Before you start: the number that sets expectations

92.36% of Google AI Overview citations come from pages already ranking in the top ten organic results. Getting cited inside an AI answer sits on top of classic search performance — it does not replace it. If a page ranks nowhere, fix that first; the checks below will not rescue it.

The reason to bother: for the query ai company lebanon, the AI Overview cites six sources while the organic page carries more than twenty results (our own DataForSEO SERP pull, Lebanon, September 2026). Six slots against twenty is the whole argument.

The twelve checks

  1. Answer in sentence one. Does the first sentence answer the question the page targets, without preamble? Assistants extract early sentences. A scene-setting opener wastes the slot.
  2. One H1, clean nesting. One h1, then h2 then h3 with no skipped levels. Heading structure is how a parser finds the section that answers a question.
  3. Real tables for comparisons. Actual <table> markup with <thead> and <tbody>. A comparison written as prose, or faked with divs, does not get lifted.
  4. FAQ schema that matches visible text. FAQPage structured data whose questions and answers appear on the page. Schema that contradicts the page is worse than none.
  5. Entity consistency. Your organisation name, address and identity match across your site, your structured data and every profile you control. Models confirm an entity by agreement between sources.
  6. sameAs completeness. Your Organization schema lists every social and directory profile you own. A missing profile is a profile the model cannot connect to you.
  7. Dates on every claim. Each statistic carries its source and period. An undated number is unciteable — and it ages badly.
  8. Image alt text and ImageObject. Real descriptions, not filenames. If image search is a traffic source for you, this is also what protects it.
  9. Canonical honesty. One canonical URL per piece of content, and it points at the page's real location. Self-canonicalising variants split you against yourself.
  10. Sitemap accuracy. Every URL in the sitemap resolves. A sitemap advertising 404s trains a crawler to trust it less.
  11. Crawlability of the answer. The content that answers the question is in the HTML, not rendered only after a click or a script. Open the page with JavaScript disabled and look.
  12. Per-assistant check. Ask the actual question in more than one assistant. ChatGPT and Perplexity overlap on roughly 11% of cited domains — they are separate ecosystems, so a single check tells you about one of them.

How to score it

Mark each check pass or fail, then fix in this order: checks 1–3 change what gets extracted, 4–6 change whether you are recognised as the source, and 7–12 remove the mechanical reasons a page gets skipped. Re-run after each change rather than in a batch, so you know which fix moved it.

Read next