What it means to rank in ChatGPT

The first question every marketing leader asks us is some version of: how do I rank in ChatGPT? It is the right instinct and the wrong vocabulary. LLMs do not rank. There is no SERP, no position three, no featured snippet to win. What there is, instead, is a citation graph, and that graph behaves in ways the old SEO playbook does not predict.

How an LLM picks who to mention

When a model generates an answer, it is drawing on two things: what it learned during training, and what it pulls in at inference time via retrieval. Your job is to influence both.

For training, models ingest enormous text corpora, web pages, books, code, forums, news. Brands that show up consistently, in context, across high-trust sources become part of the model's baseline knowledge. For retrieval, models query the live web or curated indexes. Here, your structured content, fresh publishing, and citation-friendly formatting matter most.

The four things that move the needle

• Mentions, not just links. LLMs read prose. A brand named in a sentence counts even without a hyperlink.
• Third-party coverage. Independent sources, reviews, comparisons, listicles, carry more weight than your own site.
• Structured, scannable content. Clear H2s, definitions, numbered lists. Models lift these almost verbatim.
• Recency. Retrieval-based answers strongly prefer content from the last 90 days.

If a smart human researcher would cite you, an LLM probably will too. Optimize for the researcher, not the crawler.

What does not work

Keyword stuffing, thin AI-generated pages, link farms, schema-as-magic, none of it moves LLM visibility. The systems are too good at filtering low-quality signal. What works is the slow, unglamorous work of being genuinely cited by sources that other people trust.

Measuring it

You cannot improve what you cannot see. Inside Serpier we track citation frequency, share of voice, and prompt coverage across the major models, daily, per market. That gives a real number to chase, instead of a vibe.