The GEO Checklist I Use (LLMS.txt, On-Page AIO, and What I Skip)

People ask what I actually do to show up in AI answers — LLMS.txt, on-page AIO, GEO, whatever we're calling it this week. Here's the checklist I use in practice, and what I deliberately skip.

If you want the bigger picture on *why* this matters — traffic drops when AI answers appear, the GEO market growth, and how it ties to ChatGPT rolling out ads — I wrote AI Search Is Eating Your Clicks first. This post is the tactical follow-up: what to do, in order.

What I Do

1. One clear topic per page, stated up front

Every page should answer a real question or cover a real topic, and say so early. The first heading (or the first paragraph) should make it obvious what the page is about. No "Welcome to our insights" fluff. AI systems and users both need to get the point in one pass.

2. Headings that describe the section

H2s and H3s should be meaningful. "What we do" is weak. "What we do: AI strategy, web development, and integrations" is better. Use headings as a table of contents. That helps crawlers and LLMs understand structure and pull the right snippets.

3. Entity-rich, intent-aligned content

Name the things you're talking about. Use consistent terms. Answer the intent of the page — if it's a service page, say what the service is, who it's for, and what happens next. If it's a how-to, give steps. Vague or decorative copy is hard to cite.

4. Fast, crawlable, well-linked pages

Technical basics still matter. Slow or broken pages get crawled less. Clean URLs, sensible internal links, and a logical site structure make it easier for any system — Google or an AI crawler — to find and index you. I treat this as non-negotiable before layering on "AI" tricks.

5. E-E-A-T without the theater

Experience, expertise, authoritativeness, trustworthiness. Not keyword stuffing. Author bylines where it makes sense, clear attribution, and content that shows you know the topic. That builds trust for humans and gives AI systems stronger signals about who you are.

6. LLMS.txt (and keep it updated)

I maintain an LLMS.txt at the root of this site. It describes Invisible Window, lists services, key links (home, blog, AI Readiness Scorecard, contact), and what we don't do. I update it when I publish something important or change the offer.

Do I have proof it moves the needle? No. I do it because it's cheap, it will likely matter more later, and it's a single place to say "here's who we are" in a format some crawlers use. I don't rely on it as the main lever.

7. Schema where it pays off

Structured data (JSON-LD) for organization, articles, people, FAQs, and key pages. It helps search and can help AI systems understand your content. I don't over-engineer — I add schema where it clearly describes real content (e.g. article FAQs and takeaways on this site). No fake FAQs or stuffed markup.

8. Internal links that add context

Link to other relevant pages and posts. Use descriptive anchor text. That helps users and crawlers see how your content fits together and what you're an authority on. For example, linking from this post to where AI fits in your business and questions to ask before hiring an AI consultant reinforces that we actually do this work.

What I Skip

"AI visibility" tools with no proof. I don't buy services that promise to get you into ChatGPT or Perplexity with no transparent methodology or measurable results. Until there's a reliable way to measure "cited by AI," most of that is speculation.

Treating LLMS.txt as the main strategy. It's one small layer. If your on-page content is vague or your site is a technical mess, LLMS.txt won't save you. Do it as hygiene, not as the centerpiece.

Expecting instant results. GEO is early. Crawler behavior and citation patterns change. I optimize for clarity and structure because that helps humans and future systems; I don't assume a specific percentage lift in "AI traffic" by next month.

Automation for its own sake. I don't spin up content or metadata just to feed crawlers. The content has to be useful. If it's not, no amount of markup will make you a source worth citing.

How This Fits Together

Order of operations: structure and clarity first, then technical hygiene (speed, crawlability, internal links), then metadata (LLMS.txt, schema), then optional experiments. Same idea whether you're thinking about AI search eating your clicks or getting ready for ChatGPT ads: be easy to find and easy to cite. The checklist is just a way to do that systematically.

---