The GEO checklist
If you already know what generative engine optimisation is and want the action list, this is it. Each item links to the deeper explainer if you need the reasoning behind it - this page is the checklist to work through.
Structure for readers, not for extraction
Google's own John Mueller and Danny Sullivan have both said publicly that chopping content into disconnected fragments purely to chase AI citations is bad practice - their systems already understand a page covering multiple topics and surface the relevant part without needing it pre-diced. Ahrefs has called chunk optimisation a dead end for the same reason: no publisher controls how any given AI system slices a page, so stripping content down to serve a guess trades a page that works for real readers and search traffic for one optimised against something you can't verify or control. Nothing below is about breaking a good page into isolated snippets - it's clear headings, direct answers, and content that reads as one useful, coherent page to a human, which was good writing before AI search existed and still is.
Confirm AI crawlers can actually reach your site
Before anything else works, GPTBot, ClaudeBot, PerplexityBot, and Google-Extended need to be allowed in robots.txt. Blocking them, deliberately or by an old blanket rule, removes you from consideration before any other signal matters. Decide this deliberately rather than by accident - there are legitimate reasons to block some crawlers, but it should be a choice, not a leftover default.
Add or check your llms.txt file
llms.txt is a plain-language summary of your site placed at the root, similar in spirit to robots.txt, that helps AI systems understand what you offer without parsing your full HTML. It is not a ranking guarantee, but it is a low-effort signal most competitors still have not added.
Get your structured data in order
Organization, LocalBusiness, FAQPage, and Article schema give AI systems structured facts to cite instead of forcing them to infer from prose. FAQ schema in particular maps closely to how people phrase questions to AI assistants, which makes it one of the higher-leverage additions on this list.
Write clear, direct answers within full articles
AI answer engines and human readers both benefit from a clear question as a heading followed by a direct, complete answer in the first sentence or two, with supporting detail after - that's a readability choice, not a fragmentation tactic. The test is simple: if a paragraph would confuse a stranger who landed on it with none of the surrounding page for context, it's been chopped too fine. Write the full, useful answer inside a real article; let good headings do the organising.
Build genuine citation and entity signals
sameAs links to your verified profiles (LinkedIn, Wikidata, industry directories), an author or Organization entity that ties consistently across your site, and mentions on third-party sites all help an AI system corroborate who you are before it cites you. This is the slowest item on the list to build and the hardest to fake - it rewards being genuinely findable elsewhere, not just on your own site.
Check what AI systems currently say about you
Ask ChatGPT, Claude, and Perplexity directly what they know about your business. If the answer is wrong, thin, or names a competitor instead, that gap tells you which of the items above to prioritise first, rather than working the checklist in an arbitrary order.
Frequently asked questions
Should I break my content into small chunks to get cited by AI?
No - this is now a fairly settled point among people who'd know. Google's John Mueller and Danny Sullivan have both said publicly that Google's systems already parse a full page and surface the relevant section, so pre-chunking doesn't help and can actively hurt the page for the human readers and search traffic it also has to serve. Ahrefs has called the practice a dead end for the same underlying reason: you can't control how any given AI system slices your content, so you'd be optimising for a guess.
Is GEO the same as SEO?
No, though they overlap. Traditional SEO optimises for Google's crawler and ranking algorithm. GEO optimises for how AI systems retrieve, summarise, and cite information. Roughly 60% of the underlying work is shared (crawlability, structured data, clear content), and the rest - llms.txt, AI-specific schema, genuinely well-organised content - is specific to GEO.
How long does it take to see GEO results?
Slower than traditional SEO in most cases, because AI systems retrain and re-crawl on their own schedules rather than a predictable index cycle. Technical fixes (crawler access, llms.txt, schema) can take effect within weeks once re-crawled; entity and citation signals build over months as third-party mentions accumulate.
Can I check my own GEO readiness?
Yes - a scan that reads llms.txt presence, structured data, crawler access, and content organisation in one pass is faster than checking each manually. AuditHQ's AI visibility suite runs this as part of a free scan.