How to use this checklist

Work through the sections in order and record a pass or fail for each item. The sections are dependency-ordered: crawl access gates indexation, indexation gates retrieval, and retrieval gates everything editorial. Fix failures top-down.

A failure in section one makes sections four through six irrelevant, which is why so many content-led GEO efforts produce nothing.

Section 1: Crawl access (items 1-8)

Confirm that AI crawlers can physically fetch your pages. This is the first gate and the most common silent failure — a firewall rule that returns 403 to every AI bot will make all other work worthless.

  1. robots.txt is reachable and returns 200.
  2. GPTBot is allowed.
  3. OAI-SearchBot and ChatGPT-User are allowed.
  4. ClaudeBot and Claude-Web are allowed.
  5. PerplexityBot is allowed.
  6. Googlebot and Bingbot are allowed.
  7. No CDN, WAF, or bot-protection rule blocks these agents.
  8. Server logs from the last 30 days show real AI crawler visits.

Section 2: Indexation (items 9-16)

Confirm your pages are actually in the indexes AI engines retrieve from. Bing matters as much as Google here, because ChatGPT's search is grounded in Bing.

  1. The site is verified in Google Search Console.
  2. The site is verified in Bing Webmaster Tools.
  3. sitemap.xml is valid, current, and submitted to both.
  4. A site: query returns your key pages on Google.
  5. A site: query returns your key pages on Bing.
  6. Every page has a self-referencing canonical.
  7. No important page is accidentally noindexed.
  8. No redirect chains or 404s among internal links.

Section 3: Rendering and performance (items 17-24)

Confirm your content exists in the raw HTML and that pages load quickly. Content that appears only after client-side JavaScript may never be seen by some crawlers.

  1. Main content is present in the server-rendered HTML.
  2. Headings and body copy appear with JavaScript disabled.
  3. LCP is under 2.5 seconds.
  4. CLS is under 0.1.
  5. INP is under 200 milliseconds.
  6. Images use modern formats and are lazy-loaded below the fold.
  7. Render-blocking scripts are deferred or removed.
  8. Compression (Brotli or Gzip) is enabled.

Section 4: Structured data (items 25-32)

Confirm your schema is present, valid, and accurate. Markup that describes content not visible on the page is a policy violation rather than an optimization.

  1. Organization schema on the homepage.
  2. WebSite schema on the homepage.
  3. Article or BlogPosting schema on every post.
  4. FAQPage schema wherever real FAQs are visible on the page.
  5. BreadcrumbList schema across the site.
  6. Service or Product schema on commercial pages, with accurate prices.
  7. All markup validates with zero errors.
  8. No Review or aggregateRating markup for reviews that do not exist.

Section 5: Content structure (items 33-42)

Confirm each section can be extracted and quoted on its own. This is where most sites have the largest gap between current state and citable state.

  1. Every H2 is followed by a 40-60 word direct answer.
  2. Headings are phrased as questions users actually ask.
  3. Paragraphs are 2-3 sentences.
  4. Sentences are under roughly 25 words.
  5. Comparisons are presented as tables.
  6. Enumerable items are presented as lists.
  7. Key entities and metrics are bolded.
  8. Images have descriptive alt text.
  9. There are 3-5 internal links per 1,000 words with descriptive anchors.
  10. Each section makes sense read in isolation.

Section 6: Authority and trust (items 43-48)

Confirm a reader — or a model — can tell who produced the content, when, and on what basis. These signals break ties between otherwise equivalent sources.

  1. Every substantive page has a named author or identified editorial team.
  2. Author credentials relevant to the topic are stated.
  3. Publication date is visible on the page.
  4. Last-updated date is visible and matches dateModified.
  5. Every statistic links to a primary source.
  6. No statistic, testimonial, or case study is fabricated.

Section 7: Measurement (items 49-50)

Confirm you can tell whether any of this is working. Without a fixed prompt list and a record of results, you cannot distinguish progress from noise.

  1. A fixed list of 8-15 test prompts is tracked weekly across ChatGPT, Perplexity, Gemini, and Claude.
  2. Analytics segments AI referral traffic separately from organic search.

If you would rather not run this manually, a Terrank GEO audit covers the same ground and returns a prioritized fix list within 48 hours.

Frequently asked questions

How often should I run a GEO audit?

A full audit quarterly is reasonable for most sites, with the crawl access and indexation sections checked monthly. Those two sections fail silently and are worth monitoring more often than the rest.

What should I fix first?

Work in the order of the checklist. Crawl access and indexation gate everything else, so a failure there makes content and authority work ineffective until it is resolved.