Guide

The generative engine optimisation checklist

Twenty items: thirteen on your own site, done in an afternoon, and seven on the pages the engines read, which is where the score is.

Updated September 2026 · by Monish

Short answer

Thirteen checks on your own site, an afternoon's work: AI crawlers allowed, robots.txt, sitemap, llms.txt, JSON-LD with Organization, Product or Service and FAQ blocks, title, description, Open Graph, canonical, HTTPS. Then seven things on other people's pages, which is where being named is decided. Crawler access first; it undoes everything else if wrong.

Your own site: thirteen checks, one afternoon

CheckWhat passes
AI crawlers allowedNo site-wide Disallow for GPTBot, OAI-SearchBot, PerplexityBot, Google-Extended, ClaudeBot or CCBot
robots.txt presentAnd it names the crawlers explicitly
XML sitemapDeclared in robots.txt, with lastmod dates
llms.txtName, one-line summary, the pages worth reading
JSON-LD on the home pageAny structured data at all
Organization schemaName, legal name, logo, contact point, sameAs profiles
Product or Service schemaWhat you sell, declared
FAQ schemaYour buyers' questions, answered in a sentence each
Page titleSays what the company is, not a slogan
Meta descriptionOne factual sentence: what, for whom
Open Graph tagsTitle, description, image
Canonical URLOne address per page
HTTPSEverywhere, with redirects

These are the report's thirteen technical checks; the weights are here. Four of the files can be generated from your pages in one go.

Other people's pages: where the score is

ItemWhat it means
Comparison pages on your site'Us vs the leader', 'Leader alternatives': the pages engines quote for comparison questions
Use-case pages in the buyer's wordsOne per question your buyers actually ask
Review profiles filledG2, Capterra, Trustpilot, Clutch, whichever the engines cite in your category
Directory listings claimedThe roundups and directories cited for your questions
Community answersGenuine replies where your buyers' questions are asked
Editorial mentionsThe publications the engines cite; a campaign, measured in weeks
A rescan on the same questionsMonthly, so movement is measured and not felt

Order them by how fast they land: profiles and listings in days, community answers in weeks, editorial in months. The report ranks the specific ones for your category by impact against effort.

The order to do it in

  1. Measure: which questions name a competitor and not you, and which sources decided it.
  2. Your own site: the thirteen checks, then the comparison and use-case pages.
  3. The sources: profiles, listings, threads, then mentions.
  4. Measure again on the same questions, thirty days on.

Questions about the list

How many of these does the report check?

The thirteen under 'your own site' are the report's technical checks, weighted the same way. The rest are what the report's fixes and a monthly plan are made of.

Which one matters most?

Crawler access. A site-wide block on GPTBot undoes every other item on the list, which is why it carries six times the weight of a canonical tag in the score.

Can I do this without an agency?

The first thirteen, yes, in an afternoon, and the <a href="/tools/aeo-kit-generator">generator</a> writes four of the files. The rest is monthly work on other people's pages, which is what a plan is.

Run the thirteen checks on your site now

The free check runs them all, then asks four engines whether they name you at all.

Check my AI visibility