We Built an AI Visibility Scanner, Ran It on Ourselves,
and Scored 75/100

The moment of truth

We spent months building GetCiteFlow — an enterprise AI brand service with a free scanner that checks websites and tells you exactly why AI search engines like ChatGPT, Claude, Gemini, Perplexity, DeepSeek, and Doubao aren't citing your content. We built landing pages targeting "AI visibility" and "how to get cited by ChatGPT." We wrote blog posts about GEO. We felt like experts.

Then one Friday afternoon, almost as a joke, someone on the team said: "Have we actually run our own scanner on our own site?"

Silence.

We typed getciteflow.ai into our own scanner and hit enter. Thirty seconds later, the result appeared:

Seventy-five. We were telling other companies how to optimize for AI visibility, and our own site was missing the fundamentals. The irony was painful.

What the scan revealed (the ugly truth)

The 6-dimension breakdown surfaced four critical gaps:

1. No FAQ Schema. We had a FAQ section on the homepage answering "What is GEO?" and "How does GetCiteFlow work?" — but it was plain HTML. Without JSON-LD structured data, LLMs couldn't reliably extract our Q&A pairs for citation. FAQ Coverage: 0%
2. Missing llms.txt. AI crawlers like GPTBot and ClaudeBot had no structured index of our site. They were guessing which pages mattered. We had blog posts, landing pages, comparison pages — but no map for machines.
3. Weak entity clarity. Our brand wasn't consistently defined across pages. Some pages said "GetCiteFlow is a brand visibility service," others just said "we help with AI visibility." Search engines and LLMs both rely on clear, consistent entity definitions to build a mental model of what a site is about.
4. Content not structured for AI extraction. Our blog posts were well-written for humans — but lacked the clear heading hierarchy, comparison tables, and numbered lists that LLMs use to extract and cite information. Beautiful prose doesn't help if an AI can't parse it.

What we fixed (in order of impact)

We prioritized changes by effort-to-impact ratio. Here's what we did, in the order we did it:

1. Built a comprehensive llms.txt — 2 hours

This was the quickest, highest-impact win. We listed every important page — blog posts, landing pages, comparison pages, case studies — with clear descriptions in Markdown format. Now when GPTBot or ClaudeBot crawls our site, it knows exactly what's there instead of guessing. Result: AI crawlers now index all key pages on first visit.

2. Added FAQ Schema to the homepage — 1 hour

We converted our existing FAQ section into JSON-LD structured data — seven questions covering "What is GEO," "How does GetCiteFlow work," "Is GEO different from SEO," and common user concerns. FAQ Schema is one of the most commonly cited structured data types by LLMs because it directly answers questions. Result: FAQ Coverage went from 0% to 85%.

3. Defined our entity clearly across all pages — 3 hours

We added Organization and SoftwareApplication JSON-LD schema to every page, consistently using the same brand name, description, and URL. This tells LLMs: "GetCiteFlow is an enterprise brand visibility service, not a generic SEO tool." Entity clarity compounds — the more consistently you define yourself, the more confidently LLMs cite you. Result: Entity Clarity score improved significantly.

4. Restructured blog content for AI readability — 1 day

We audited every blog post and ensured: clear H2/H3 hierarchy, comparison tables where applicable, numbered lists for actionable advice, and a summary section that LLMs could extract as a direct snippet. We didn't change the content — we just made it machine-readable. Result: Content Structure score improved.

The results: 75 → 92

After implementing all four changes, we re-scanned getciteflow.ai:

But the score is just a number. The real test was asking ChatGPT and Claude questions like "what is GEO?" and "how do I get my site cited by AI?" — and seeing GetCiteFlow appear in the answers.

More importantly, we now had a credible story to tell. When potential customers ask "does this actually work?" — we can show them our own before-and-after. Not a hypothetical. Not a fabricated case study. Our own site, our own service, real results.

Key takeaways for your own site

1. Run your own scanner on yourself first. If you're selling any kind of optimization and your own site scores poorly, fix that before pitching anyone else. Credibility is everything.
2. llms.txt is the highest-impact, lowest-effort GEO change. It took 2 hours and immediately gave AI crawlers a structured map of our site. If you do nothing else, do this.
3. FAQ Schema matters more than you think. LLMs frequently cite FAQ content in responses because it directly answers user questions. Without structured data, your Q&A is invisible to AI.
4. Entity clarity compounds over time. Consistently defining your brand across all pages helps LLMs build a reliable mental model of what you do. Inconsistency confuses both search engines and AI.
5. Content structure is for machines, not just humans. Clear headings, tables, and lists make your content extractable by AI — which is how citations happen. Beautiful writing that AI can't parse is wasted effort.