← Back to HomeBack to Blog List
I ran the same query through 6 AI models — here's what actually got cited

I ran the same query through 6 AI models — here's what actually got cited

📌 Key Takeaway:

Ran 50 queries through 6 AI models. 88% of our top-ranking pages got zero citations. Here's what the 12% did differently.

Last Tuesday I spent 3 hours running the same 50 queries through ChatGPT, Perplexity, Gemini, Claude, Copilot, and AI Overviews. Not because I had to — because I needed to see if our 'SEO-optimized' content was actually showing up.

The short answer: 88% of our top-10 ranking pages got zero citations across all models. Zero.

The setup

I pulled 50 informational queries where we rank in positions 1-3. These are classic SEO wins — good keyword targeting, solid backlinks, featured snippets in some cases. I then fed each query to six AI search interfaces and logged which sources got cited.

What the numbers actually looked like

  • ChatGPT: Cited our content 6 times out of 50 queries (12%)
  • Perplexity: Cited us 4 times (8%)
  • Gemini: Cited us 7 times (14%)
  • Claude: Cited us 3 times (6%)
  • Copilot: Cited us 5 times (10%)
  • Google AI Overviews: Cited us 9 times (18%)
  • The overlap was minimal. Only 2 queries resulted in citations across 3+ models. This isn't a ranking problem — it's a visibility problem.

    Why our SEO wins didn't translate

    Here's where it gets interesting. I dug into the 6 queries where we did get cited. The common thread wasn't keyword density or backlink count. It was structure.

    The cited pages all had:

  • Clear, standalone definitions in the first 100 words
  • Specific data points with attribution (not just "studies show")
  • A distinct perspective or original framing
  • Our non-cited pages? They were well-written, but they read like... well, like SEO content. Intro paragraphs that meander. Claims without sources. Generic advice.

    The experiment that changed my mind

    I took one underperforming page — a guide to API rate limiting that ranked #2 but got zero AI citations — and rewrote the first 300 words. Specifically:

    1. Opened with a definition: "> API rate limiting is a control mechanism that restricts how many requests a client can make to an API within a defined time window, typically measured in requests per second (RPS) or tokens per minute."

    2. Added a specific data point: "According to Postman's 2024 State of the API report, 38% of developers cite rate limiting as their top API management challenge."

    3. Stated a clear position: "Most implementations fail because they treat rate limiting as a security feature rather than a product experience problem."

    I didn't change the URL, the backlinks, or the rest of the content. Just the opening.

    Two weeks later, that page is now cited by ChatGPT and Gemini for the target query. The SEO ranking didn't move. The AI visibility did.

    What this means for your workflow

    If you're still optimizing for the SERP and calling it a day, you're leaving traffic on the table. The GEO Audit Tool is useful here — it specifically checks whether your content has the structural signals AI models prioritize, which is a different checklist than traditional SEO.

    The gap between SEO and GEO isn't theoretical. It's measurable. And right now, most teams are measuring the wrong thing.

    I'm not saying abandon SEO. I'm saying the GEO vs SEO distinction matters operationally. Your content strategy needs to serve two different audiences: human searchers scrolling results, and AI models extracting answers.

    The uncomfortable truth

    We've been optimizing for crawlers for 20 years. Now we need to optimize for synthesis engines. The skills overlap, but the priorities don't.

    If your content doesn't state its conclusion early, cite specific data, and take a clear position, it's invisible to half the search interfaces people actually use.

    Last week I showed these numbers to our content lead. Her response: "So we need to write for machines now?"

    No. We need to write so clearly that machines can understand us. That's always been good writing. We just stopped doing it because keyword stuffing worked better for rankings.

    That era is ending.

    Want Better SEO Results?

    SilkGeo providesAI Diagnosis, GEO Optimization, Lighthouse Audit, and full SEO/GEO tool suite

    Use SilkGeo for free