What is on a generative engine optimization checklist?

A GEO checklist covers three layers in order. Access: allow AI crawlers in robots.txt, serve content in the HTML rather than via JavaScript, and fix indexability errors. Quotability: answer each question completely in the first two sentences under its heading, add matching structured data, and keep your entity details consistent. Compounding work: page speed, an llms.txt file, and mentions on sites other than your own.

What is the first thing to fix for GEO?

Check robots.txt for rules blocking GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot or Google-Extended. A single Disallow line removes you from that assistant completely, and many sites carry one by accident from a plugin default or a copied file. It is a one-line fix with a larger effect than anything else on the list.

How do I make my content quotable by AI?

Put the question in a heading and answer it fully in the first two sentences underneath, then add detail afterwards. Test it by copying a single paragraph into a blank document: if it still reads as a complete and true answer with no surrounding context, an assistant can quote it. Paragraphs that open with phrases like "as mentioned above" cannot be quoted alone and usually are not.

Does structured data help with AI search?

Yes, because it states facts a parser would otherwise have to infer from prose — your organization name, an article's author and date, a product's price. Add only the schema types that match content a visitor can actually see. Note that Google retired FAQ rich results in May 2026, so FAQPage markup no longer produces a search accordion, though answer engines still parse it.

How long does GEO take to work?

Access fixes such as unblocking a crawler take effect as soon as that crawler next visits, which can be days. Writing and structured data changes depend on recrawl and reindex cycles, so allow a few weeks. Earning mentions on other sites is the slowest item and works on a timescale of months, which is why it sits last on the checklist rather than first.

Do I need to publish llms.txt?

No. It is a proposed standard that Google has confirmed Search does not use, and no major AI provider has committed to reading it. Publishing it takes about ten minutes and may pay off later, so it is reasonable cheap insurance. It is not a substitute for crawler access, server-side rendering or structured data.

Generative engine optimization: the practical checklist

The hands-on companion to the GEO explainer: nine concrete changes, in priority order, with how to verify each one.

By Max9 min read

This is the hands-on version: nine changes that make your site readable and quotable by AI assistants, in the order worth doing them, each with a way to check it actually landed. If you want the background first — what generative engine optimization is and why it appeared — read what GEO optimization is for the short version, or our 2026 guide to Generative Engine Optimization for the fuller story of why it appeared, then come back. This page assumes you are ready to change something.

1. Unblock the AI crawlers you want

Open yoursite.com/robots.txt and look for rules naming GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot or Google-Extended. A Disallow: / under any of them removes you from that assistant entirely.

Decide this on purpose. Allowing them means your content can appear in AI answers and can be used for training. Blocking them protects the content and costs you the citations. Most sites that block are doing it by accident, inherited from a plugin default or a copied file.

How to verify: load your robots.txt in a browser and read it. It is a short text file, and the rules are plain English.

2. Serve your content in the HTML

Most AI crawlers do not execute JavaScript. If your text arrives after load, they see an empty page where your article should be.

Use server-side rendering or static generation for any page you want quoted. This is a build-configuration change on every modern framework, not a rewrite.

How to verify: right-click, View Page Source, and search for a sentence from the middle of your content. If it is not there, no crawler that skips JavaScript can see it either.

3. Fix your indexability basics

A stray noindex, a canonical pointing at the wrong URL, or a 404 in your main navigation hurts AI visibility for the same reason it hurts search: the page stops being a candidate. GEO does not replace this work, it inherits it.

How to verify: run a free website check and read the critical findings. That is precisely what the critical tier is for.

4. Answer the question in the first two sentences

This is the single highest-leverage writing change. Put the question in a heading, then answer it completely underneath, before any context or caveats. Detail goes after the answer, not before it.

The test is whether a passage still makes sense with everything around it removed. A paragraph that begins “this also means” fails, and an assistant that cannot quote a passage in isolation generally does not quote it.

How to verify: copy any single paragraph into a blank document. If it reads as a complete, true answer on its own, it passes.

5. Add the structured data your pages warrant

Schema markup states facts instead of implying them: your organization name, an article’s author and date, a product’s price, a page’s question-and-answer pairs. Add the types that genuinely match your content and no others — marking up an FAQ a visitor cannot see is a violation, not a shortcut.

One caveat worth knowing, because most guides have not caught up: Google retired FAQ rich results in May 2026, so FAQPage markup no longer earns a SERP accordion. It is still valid structured data that answer engines parse, and it costs a few hundred bytes, so it stays worth having — just not for the reason older articles give.

How to verify: paste the URL into Google’s Rich Results Test. It reports which types were detected and any errors.

6. Be consistent about who you are

Assistants assemble an understanding of an entity from every mention of it. If your business name, description and contact details differ between your homepage, your about page and your schema markup, you have given three slightly different entities instead of one confident one.

Pick one form of your name and one one-sentence description of what you do. Use them everywhere, including in your structured data.

How to verify: read your homepage, your about page and your Organization schema side by side. They should agree.

7. Make your pages fast enough not to be dropped

Crawlers work to a budget. Slow responses mean fewer pages fetched per visit, and on a large site that is the difference between your whole catalog being read and a sample of it. Core Web Vitals are the standard measure, and improving them serves human visitors at the same time.

How to verify: our scan reports real Lighthouse measurements with the version and timestamp, so you can reproduce them. What counts as a good score covers how to read the numbers.

8. Publish an llms.txt, with realistic expectations

A Markdown file at your root listing your key pages. It is a proposal, not a standard, and Google has confirmed Search does not use it. Nobody has promised to read it.

Publish it anyway — it takes ten minutes and may matter later. Do not publish it instead of items 1 through 7. The full guide to what llms.txt is covers the format and how to write one, and the AI visibility checker guide puts it in context.

How to verify: it should load at /llms.txt as plain Markdown with a heading and real links.

9. Earn mentions somewhere other than your own site

The uncomfortable item, and the one no tool can do for you. Assistants lean heavily on sources that other sources talk about. Being mentioned in places your buyers already read does more for AI visibility than any markup change.

There is no trick here and anyone selling one is selling something else. It is the same work that has always built authority, and it is slower than the eight items above.

How to verify: search your brand name and see whether anything other than your own properties comes back.

What to do in what order, if you only have an afternoon

Check robots.txt, view source on your most important page, and run one scan. Those three take fifteen minutes and will tell you whether you have a blocking problem or a quality problem. A blocking problem is a one-line fix with a large effect. A quality problem is real work, and worth knowing about before you start.

Run the free check to get the first pass, then work back up this list.

Frequently asked questions

What is on a generative engine optimization checklist?: A GEO checklist covers three layers in order. Access: allow AI crawlers in robots.txt, serve content in the HTML rather than via JavaScript, and fix indexability errors. Quotability: answer each question completely in the first two sentences under its heading, add matching structured data, and keep your entity details consistent. Compounding work: page speed, an llms.txt file, and mentions on sites other than your own.
What is the first thing to fix for GEO?: Check robots.txt for rules blocking GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot or Google-Extended. A single Disallow line removes you from that assistant completely, and many sites carry one by accident from a plugin default or a copied file. It is a one-line fix with a larger effect than anything else on the list.
How do I make my content quotable by AI?: Put the question in a heading and answer it fully in the first two sentences underneath, then add detail afterwards. Test it by copying a single paragraph into a blank document: if it still reads as a complete and true answer with no surrounding context, an assistant can quote it. Paragraphs that open with phrases like "as mentioned above" cannot be quoted alone and usually are not.
Does structured data help with AI search?: Yes, because it states facts a parser would otherwise have to infer from prose — your organization name, an article's author and date, a product's price. Add only the schema types that match content a visitor can actually see. Note that Google retired FAQ rich results in May 2026, so FAQPage markup no longer produces a search accordion, though answer engines still parse it.
How long does GEO take to work?: Access fixes such as unblocking a crawler take effect as soon as that crawler next visits, which can be days. Writing and structured data changes depend on recrawl and reindex cycles, so allow a few weeks. Earning mentions on other sites is the slowest item and works on a timescale of months, which is why it sits last on the checklist rather than first.
Do I need to publish llms.txt?: No. It is a proposed standard that Google has confirmed Search does not use, and no major AI provider has committed to reading it. Publishing it takes about ten minutes and may pay off later, so it is reasonable cheap insurance. It is not a substitute for crawler access, server-side rendering or structured data.