Your Website Is Optimized for LLMs, But LLMs Aren't Using It
An LLM brand audit extracted over 100 facts about one institution. None of them came from the institution's own website.
We recently completed an LLM brand audit for a well-established educational program using TondroAI Discover, and the headline finding stopped us in our tracks.
The institution's website was well-built, cleanly structured, semantically sound, and fully accessible to AI crawlers. The LLMs captured all human-visible content with zero warnings and zero content gaps. By every technical measure, this site was ready for AI consumption.
Then we looked at what the AI models actually say about the program.
Over 100 Facts, Zero From the Institution
We extracted over 100 distinct factual claims about the program from leading LLMs, the kinds of statements a prospective student, parent, or school counselor would encounter when asking AI for information about tuition figures, eligibility requirements, curriculum details, etc.
Not a single one of those facts traced back to the institution's own website.
Every claim the LLMs made, every answer they gave, was assembled from third-party sources like rankings sites, college advice blogs, forums, aggregator platforms, and competitor pages. The institution's own voice, the story they had carefully crafted on their website, was completely absent from the AI-generated version of their brand.
How Does This Happen to a Well-Built Website?
So why weren't the LLMs using the website's content?
The source provenance data from the audit tells the story. Out of thousands of citations across all AI responses, only 7.7% pointed to the institution's own domain. The remaining 92.3% came from third-party sources that the institution has no editorial control over.
The content existed on the website, but the website wasn't winning the citation game. Third-party sources were being treated as more authoritative, more frequently referenced, and more deeply embedded in the models' training and retrieval pipelines. The institution had the answers, but the LLMs were getting those answers from everyone else.
The Information Isn't Wrong, But It Isn't Yours
To be clear, the LLMs weren't fabricating information. Most of the facts they surfaced were roughly accurate. But "roughly accurate from someone else's perspective" is a dangerous place for a brand to live, because you can't control the framing, the emphasis, or the nuance.
In this audit, we found cases where different LLM providers gave contradictory answers to the same question, one citing an acceptance rate below 10%, another citing 35-50%. The institution's own site could have settled the question definitively, but because it wasn't part of the conversation, the conflicting information just sat there, confusing anyone who happened to ask.
And that's the real risk: it's not that AI is making things up about you, it's that AI is telling your story using everyone's words but yours, and when those words conflict or miss the mark, you're not in the room to correct them.
This Isn't an Edge Case
We see this pattern across certain types of institutions where the website is solid, the SEO is fine, the social media presence is active. But nobody has asked the fundamental question: when a prospective student asks ChatGPT or Perplexity or Gemini about our program, where is the answer actually coming from?
And what we definitively see is that unless you are a top brand in the education space, in most cases, the answer is: not from you.
The information pipeline has shifted, and the institutions that recognize this early have a real opportunity to get ahead of it. But you can't craft a strategy for a problem you haven't measured yet.
What To Do About It
Armed with this kind of data, the path forward becomes clear, and it's a two-step process.
Step 1: Audit what AI is actually saying about you. You need a baseline. Not assumptions, not gut feelings, but a systematic extraction of the facts LLMs are presenting about your institution, where those facts are coming from, and where your own content is and isn't showing up. You can't manage a narrative you can't see, and you can't fix citation gaps you haven't mapped.
Step 2: Craft a strategy to close the gap. Once you know what the LLMs are saying and where they're getting it, you can take targeted action: restructuring content to win citations, creating pages that directly answer the questions AI users are asking, addressing factual inconsistencies that have propagated into model training data, and building a content architecture that treats AI visibility as a first-class priority alongside traditional search and social.
The audit gives you the map. The strategy gives you the route. Neither works without the other.
Start With Step 1
We're offering complimentary LLM brand scans for higher education institutions, no commitment, no sales pitch, just a clear-eyed look at how AI currently represents your brand and where the gaps are.
TondroAI Discover is an AI-powered brand intelligence platform built specifically for higher education.