How to Get Cited by ChatGPT

Neil Yan • May 18, 2026 • 7 min read

We spent three months running experiments on what makes a source citable by LLMs. The results were not always intuitive. Some of the most authoritative, well-written content we tested was ignored entirely, while pages with thinner content but better structural alignment were cited consistently. Here is what actually works.

Entity Clarity Is Table Stakes

Every LLM needs to resolve what your brand is before it can cite it. If your site uses vague language — "our platform," "the solution," "next-gen technology" — the model has nothing concrete to anchor to. Define what you are in plain terms, repeatedly. Your homepage should state your category within the first two paragraphs. Your about page should do the same. Every page should reinforce the entity mapping.

A simple test: ask ChatGPT "What is [your company]?" and see if the answer matches how you describe yourself. If it is wrong or vague, the model has not learned your entity. Fixing this requires consistent, category-specific language across your entire web presence — not just your marketing pages, but your documentation, your integrations pages, and any third-party sites that reference you.

Comparison Pages Drive Citations

In our experiments, comparison content (e.g., "Product X vs. Product Y") was the single most reliably cited format. LLMs love comparisons because they provide clear, structured relationships between entities. When the model needs to answer "what is the best tool for X," it pulls from content that explicitly ranks or contrasts options.

If you are a smaller brand, do not avoid comparisons with larger competitors. Being compared to an established player is one of the fastest ways to establish a shared entity cluster. The model may remember the comparison even if it does not remember the individual details of your product page. Write comparison pages that are factual and thorough, not promotional. Models penalize biased content.

Comparison Page Best Practices

1. Use real data points, not subjective ratings. APY, response time, pricing, feature counts — concrete numbers that the model can extract and repeat.
2. Keep the same row labels across all comparison pages. Consistent structure makes it easy for the model to compare entities programmatically.
3. Be fair to competitors. Models penalize obviously biased comparisons. Acknowledge where competitors excel and where your product falls short.
4. Include a summary verdict. A single sentence like "Best for enterprise teams that need X" gives the model a concise citation to use in recommendation answers.

FAQ Pages with Schema Markup

FAQ pages with Schema.org QA markup are disproportionately cited. The structure gives the model an easy extraction path — it can pull the question-answer pair directly without having to parse narrative text. We saw roughly 2x citation rates for FAQ pages with markup versus identical FAQ pages without it.

The format also works well for voice queries and featured snippets, but the LLM citation benefit is the stronger signal. Focus each FAQ on a single question with a direct, self-contained answer. Avoid cross-referencing other answers. The model should be able to extract any individual Q&A pair independently.

Monitor Your Citation Baseline

Before you invest heavily in GEO, establish your current baseline. Run your top 20 category queries across ChatGPT, Perplexity, and Gemini. Note whether you appear, and if so, in what context. Repeat monthly. Most teams see movement within 60-90 days if they consistently publish structured comparison and FAQ content. If you see no movement after three months, the issue is likely entity ambiguity or a training data gap, which requires a different approach — more external mentions on high-authority sites that the model trusts.

Monthly Monitoring Process

1. Prepare your query set. Maintain a fixed list of 20 category-defining questions that represent your most important citation opportunities.
2. Run across three platforms. Query ChatGPT (with search), Perplexity, and Gemini. Record results in a consistent format.
3. Score each mention. Note whether your brand appears, in what position (first, second, third+), and whether the sentiment is positive, neutral, or negative.
4. Track competitor presence. Who appears where you do not? Their content structure reveals entity gaps in your own strategy.
5. Pivot based on data. If citations are not moving after 90 days, focus on third-party mentions and external consensus signals before investing more in on-site content.

Related: How Generative Engines Choose What to Cite · Search Rankings Don't Translate to AI Citations · Why ChatGPT Ignores Your Site