The GEO Checklist: 25 Things to Fix Before AI Engines Cite You

Each item is a pass or fail check. For why each one matters, see our complete guide to generative engine optimization.

Generative engine optimization checklist items 1 to 6: access and indexation

Access and indexation checks confirm that AI engines can fetch your pages and find them in the indexes they retrieve from. These are the highest-impact fixes on the list, because a blocked page cannot be cited at all.

1. robots.txt allows AI search crawlers. Check OAI-SearchBot, PerplexityBot, Claude-SearchBot, Googlebot and Bingbot. As of October 2026, OpenAI's crawler documentation says sites that block OAI-SearchBot will not appear in ChatGPT search answers. Our AI crawler checker tests this in seconds.
2. You decided separately on training crawlers. GPTBot and ClaudeBot collect training data; OAI-SearchBot and Claude-SearchBot serve search. Blocking training does not require blocking search.
3. Your firewall and CDN don't challenge AI agents. Bot protection can return challenge pages to crawlers your robots.txt allows. Check your server logs for the user agents above.
4. Key pages are indexed in Google. Use Search Console's page indexing report and fix "crawled, currently not indexed" on commercial pages.
5. Key pages are indexed in Bing. Microsoft Copilot grounds answers in Bing. Verify Bing Webmaster Tools and submit your sitemap there too.
6. Content exists in the raw HTML. If prices, features or answers only appear after JavaScript runs, some AI crawlers will miss them. View the source, not the rendered page.

Answers and structure (items 7 to 12)

Answer and structure checks make each page easy to quote. AI engines lift sentences and blocks; your job is to write ones that stand alone.

7. Each key page opens with a direct answer. The first 40 to 70 words answer the page's main question completely.
8. Every H2 opens with its answer. The first sentence under a heading answers the heading's implied question.
9. Definitions are full sentences. "Generative engine optimization (GEO) is..." beats a fragment or a pun.
10. Comparisons use tables. Any "X vs Y" or feature comparison lives in a table, not paragraphs.
11. Processes use numbered steps. Each step is a complete clause that makes sense alone.
12. Pages have a short FAQ with real questions. Use questions buyers actually ask, each answered in 40 to 80 self-contained words.

Run your top pages through our answer-ready checker to see which of these are missing.

Evidence and originality (items 13 to 18)

Evidence and originality checks give engines something specific to repeat. The GEO research paper found that adding statistics, quotations and source citations raised visibility in generated answers by up to 40%, while keyword stuffing did worse than nothing.

13. Vague claims are replaced with numbers. "Most customers" becomes a sourced percentage, or the claim goes.
14. Every statistic links to its original source and names it in the sentence.
15. Expert quotes are attributed. Named people with real roles, quoted accurately.
16. Claims about fast-changing topics are dated. "As of October 2026" ages better than an undated claim.
17. At least some pages carry original information. Your own data, test results, pricing or first-hand experience. Google's guide to optimizing for generative AI features calls unique, non-commodity content the factor most likely to matter long term.
18. Pricing and key facts are on the page, not behind a form. Engines can't cite what they can't read.

Entity consistency (items 19 to 21)

Entity consistency checks make sure every source describes your company the same way. Contradictions make engines hedge or skip you.

19. One canonical company description. The same name, category and one-line description on your site, LinkedIn, directories and review profiles.
20. Organization structured data is present and accurate. Include name, URL, logo and sameAs links to your official profiles, following schema.org's Organization type.
21. About, pricing and product pages agree with each other. Old plan names and retired features get updated or removed everywhere.

Off-site mentions and measurement (items 22 to 25)

Off-site and measurement checks cover what others say about you and whether you can see progress.

22. You appear in the third-party comparisons for your category. Search the "best X" and "X vs Y" pages that AI answers cite, and pursue genuine inclusion.
23. Reviews are current on the platforms your buyers use. Recent, real reviews, never bought or gated.
24. A fixed prompt set is tracked monthly. 30 to 100 buyer questions across the engines that matter, recording mentions, citations and descriptions.
25. First-party AI reports are switched on. Google Search Console's Generative AI performance report and Bing Webmaster Tools' AI Performance report, plus an analytics segment for AI referrals.

How to set up items 24 and 25 properly is covered in how to measure GEO results. For rewriting pages to pass items 7 to 18, see the GEO content strategy guide.

Our view: an llms.txt file is not on this list on purpose. As of October 2026, Google says Google Search ignores it, and no engine has shown it affects citations. Add one if you like; don't expect it to fix anything above.

Frequently asked questions

What should a generative engine optimization checklist include?

A generative engine optimization checklist should cover five areas: AI crawler access and indexation, direct-answer page structure, specific evidence and original information, consistent entity facts across the web, and off-site mentions plus measurement. Access comes first, because AI engines like ChatGPT and Perplexity cannot cite pages their crawlers are blocked from fetching.

How long does it take to work through the GEO checklist?

Items 1 to 6 usually take a few days for a technical person. Rewriting key pages for items 7 to 18 takes a few weeks, depending on how many pages matter commercially. Entity cleanup is a few days. Off-site mentions are ongoing work measured in months. Measurement setup takes an afternoon once your prompt list is written.

Which GEO checklist item matters most?

Crawler access matters most, because it is binary. If robots.txt or a firewall blocks OAI-SearchBot, PerplexityBot or Claude-SearchBot, nothing else on the checklist helps in those engines. After access, original information and specific evidence tend to have the biggest effect, based on Google's guidance and the published GEO research.

Do I need structured data for AI engines to cite me?

Structured data is not required. As of October 2026, Google says there is no special schema for its AI features. Accurate Organization, Product and FAQ markup still helps engines identify your entity and makes pages eligible for rich results. Treat it as part of entity consistency, and make sure the markup matches what is visible on the page.

How often should I rerun the GEO checklist?

Rerun the access and indexation items monthly, because a site release, CDN change or robots.txt edit can block crawlers without anyone noticing. Review content and entity items quarterly, and whenever you change pricing, products or positioning. Prompt tracking from item 24 runs continuously, and its results tell you which other items to revisit.

If you want this checklist worked through for you, AEO GEO Labs runs GEO, AEO and SEO programs for B2B and SaaS teams. See our GEO and AEO services.