# How to get cited by ChatGPT, Claude, and Perplexity

When an AI assistant answers a question, it pulls from a handful of sources and names a few. Being one of those names is the new front page of Google. The good news: which sources get cited is far more predictable — and more technical — than most marketers assume.

## 1. Be machine-readable, not just human-readable

Models extract structure, not vibes. The single highest-leverage move is **valid [schema.org](https://schema.org) JSON-LD** — Organization, Product/Service, FAQPage — so a model can parse exactly who you are and what you offer without guessing from prose. This is the one lever with repeatedly observed citation lift.

## 2. Write atomic, quotable claims

Assistants lift short, self-contained, verifiable sentences. Compare:

- ❌ "We deliver world-class, synergistic solutions tailored to your journey."
- ✅ "Acme reduces onboarding time from 14 days to 3."

The second is a claim a model can quote. Structure pages as one idea per heading, with short standalone sentences and an FAQ that answers real questions directly.

## 3. Ground your identity

Models hedge on entities they can't confirm. Tie your profiles together with a consistent `sameAs` graph (LinkedIn, Wikidata, official pages) and use one canonical description everywhere. When models agree you exist and what you do, they cite you with confidence.

## 4. Let the crawlers in — and be fast

If your content only renders after JavaScript, much of it never reaches the model. Serve real HTML, allow AI crawlers ([GPTBot](https://platform.openai.com/docs/bots), [ClaudeBot](https://support.anthropic.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler), PerplexityBot, Google-Extended) in robots.txt, publish a sitemap, and keep latency low.

## 5. Stay fresh and sourced

Recency and evidence both matter. Emit `dateModified`, keep content current, and back quantitative claims with a linked source — unsupported superlatives lower a model's trust and get filtered out.

## Measure it

You can't improve what you can't see. Run your site through the [Legible Readiness Index](/) to score these signals across eight dimensions and get a prioritized list of exactly what to fix first.