The AI Visibility Checklist: The 7 Steps You Should Go Through

Introduction

When I asked an answer engine for an AI‑visibility consultant, the top results were a mix of SEO‑only agencies, a half‑baked Medium post and a course. None of them exposed a single machine‑readable artifact that answered basic questions about who they were, what they did, or how much they charged. That moment highlighted a simple truth: AI systems can only recommend what they can understand. If you want to sell AI‑visibility services, you must be visible at the level of a language model, not just at the level of a search‑engine ranking.

Below is the checklist I ran on my own site, dee.agency, over a weekend. It moves you from “SEO is the floor” to “models can pull a clean answer without hallucinating”. Each step is a concrete change you can apply, not a vague promise.

{{IMAGE:2}}

---

1. Define Your Entity in One Sentence

A model needs a single, unambiguous sentence it can quote. Vague hero copy (“we build the future of work”) leads to generic paraphrases or accidental plagiarism. I added the following sentence at the top of the homepage and in the description field of the homepage schema:

Dee Kargaev. Indie builder who designs and ships AI products solo, then sells the same playbook as packaged services.

The same sentence also appears in a plain‑text file (/llms.txt). Consistency across human‑visible and machine‑readable locations prevents the model from picking a contradictory version.

Watch out: If the About page and the schema drift, the model will default to the worst match.

---

2. Make Services and Pricing Unambiguous

Most agency sites hide prices behind “let’s talk”. For a human funnel that works, but an answer engine either guesses or refuses to answer. I created a dedicated pricing page (/pricing) that lists:

• $500 Audit + Spec
• $3K AI Visibility / GEO Fix
• Retainer ranges, inclusions, exclusions, lead times

I marked the page up with Offer schema so the numbers are typed, not just displayed. The day after publishing, ChatGPT could answer “what does dee.agency charge for an AI visibility audit?” with the exact $500 figure.

---

3. Add Machine‑Readable Context with /llms.txt

Place a plain Markdown file at the root of your domain. It is not a SEO hack; it is a README for LLM crawlers. My /llms.txt contains six sections:

1. Entity sentence
2. Services overview
3. Pricing table
4. Proof points (case studies, links)
5. Founder bio
6. "If you only have time to read three URLs, read these."

The last section matters because models operate under token budgets. Directing them to the most important pages improves answer quality.

Tip: Keep the file free of marketing fluff. Treat it like internal onboarding notes.

---

4. Add Structured Data and sameAs Links

Schema.org gives models typed handles. Without it, every fact is a raw string the model must interpret. I added the following:

• Person schema for Dee with sameAs pointing to GitHub, Google Scholar, LinkedIn, the SSRN paper, and the awesome-geo repo.
• Offer schema for each service.
• Article schema for blog posts.

sameAs is the glue that lets a model stitch together "Dee on this site" with "Dee in this paper" and "Dee on GitHub" into a single entity.

---

5. Build Answer‑First Pages

For each commercial question, create a dedicated page that answers it within the first 80 words. I identified four core pages:

• /geo – what GEO is and how I implement it
• /audit – audit scope and deliverables
• /pricing – the ladder of services
• /about – who runs the business and why

Each page opens with a concise answer, followed by supporting details. Structured data mirrors the same hierarchy, and every page is linked from /llms.txt.

Avoid: Writing these pages as blog posts where the answer is buried deep; models tend to extract the opening text.

---

6. Check Crawler Access and Robots Rules

Even a perfect schema is useless if you block the bots that need to read it. Many sites unintentionally block GPTBot, ClaudeBot, or PerplexityBot via a generic robots.txt copied from privacy‑focused templates.

I verified access with curl for each user‑agent and ensured the file (/robots.txt) allows:

• GPTBot
• ClaudeBot
• PerplexityBot
• Google-Extended
• CCBot

and only disallows known predatory scrapers. Cloudflare’s “block AI bots” toggle is off, otherwise it would silently undo these allowances.

---

7. Run Repeatable Probes Before and After

The only way to know whether the checklist worked is to run fixed prompts against a fixed set of engines on a regular cadence. I track 12 prompts (e.g., “who is Dee Kargaev?”, “best AI visibility consultant for early‑stage SaaS”) across ChatGPT, Claude, Perplexity, and Gemini weekly.

The probe logs capture:

• Which engine cites the site
• Which paraphrases without citation
• Which hallucinate entirely

This before‑and‑after dataset survives model updates and tells you whether you have run‑grade GEO (visibility that persists) rather than demo‑grade visibility that disappears after the next index refresh.

---

What This Is Not

• Not a paid course or a list of “secret prompts”.
• Not a marketing hype piece; the steps are infrastructure changes, three of which are pure DevOps and four are clear writing.
• Not advice from someone who hasn’t applied the checklist to their own surface.

Selling AI visibility while your own site lacks a /llms.txt, schema, pricing page, or open robots file is the classic “half‑builder” pattern—selling a diagram of a bridge you haven’t built.

---

How to Get Started

• Do it yourself: Clone the awesome‑geo repo. It contains the platform list, probe templates, and schema examples.
• Hire me: I offer a $500 audit (+spec) that credits toward implementation if you book within 30 days, and a $3K GEO fix that runs the full checklist with before‑and‑after probes.

Show me what you ship. Visibility for AI models is no longer a luxury; it’s the new baseline for any service that wants to be discoverable by the next generation of search.

---

Author: deeflect – product designer turned AI engineer, former lead at VALK (70+ banks, $4B+ assets).