50% of ChatGPT citations are listicles. So, here's a listicle about how to structure your pages to get cited on ChatGPT. This is a nice companion to Ed Chater's classic blog post AI Search Optimization: 10 Steps to Get Your Brand Recommended.
How did we come up with this data? Will Robinson figured out that ChatGPT Loves Listicles by analyzing more than 40,000 URLs that ChatGPT cited heavily on the Evertune platform in the last 60 days, across nearly 50 product categories. The page components listed below were compiled based on 33,000 of those URLs.
How to structure content for ChatGPT
- Write content that has about 941 words. 941 words is the median length of a ChatGPT-cited page. Not 2,500. Not a sprawling pillar page. AI models don't exclusively reward long-form content — they reward dense, specific content at whatever length the topic requires. If your page is 400 words of thin copy, add depth. If it's 4,000 words of padding, cut it.
- Break your content into about 18 paragraphs. 18 paragraphs is the median. That's roughly one paragraph every 50 words — short, scannable blocks that make it easy for AI models to extract discrete ideas. Think of each paragraph as a self-contained unit of evidence. If a paragraph covers two ideas, split it.
- Keep your average sentence length around 17 words. At 17 words per sentence, you're in the range of pages ChatGPT cites most. AI wants content that’s easy to understand, and 17 words appears to be the sweet spot.
- Use about four H2 headers. Headers are how AI models navigate document hierarchy. The median page ChatGPT cites uses 4 H2s — enough to signal topical structure without over-fragmenting the content. Each H2 should map to a distinct sub-question someone might actually search. If your headers don't reflect real user queries, rewrite them.
- Add about two H3 headers. H3s appear more sparingly — the median is just 2. Use them to add a second layer of hierarchy within a longer section, not as visual decoration.
- Include about 28 internal links. This one surprises marketers. 28 internal links is the median for a ChatGPT-cited page. Pages ChatGPT cites tend to be embedded in larger content ecosystems. If you've covered a topic across multiple pages, link between them. Build the ecosystem intentionally.
- Include about 15 external links. Externally linking to credible sources signals that your content is grounded rather than self-referential. The median page includes 15 external links. Prioritize primary sources, peer-reviewed research, statistical data, and established institutions — the kinds of sources AI models are already trained to trust. Quality likely matters more than quantity here, but that's beyond what this dataset can confirm.
- Include about 10 images. The median cited page includes 10 images. The pattern is consistent across the 33,000 URLs we analyzed. Whether that reflects content depth, production quality, or topical richness is worth debating — but pages with no images are consistent outliers in the cited set.
- Include about 2 lists. Cited pages typically contain 2 embedded lists. Structured lists and comparison tables are among the content formats AI models favor most — they're easy to extract and repurpose in a response. You don't need to turn every page into a roundup. A structured list somewhere in the content goes a long way.
Next steps to get cited by ChatGPT
These are patterns, not rules. The median page length ChatGPT cites is a baseline, not a guarantee. Meeting median thresholds won't ensure citation, but consistently falling short reduces your odds meaningfully.
ChatGPT prefers structured, richly linked, image-supported content at a medium length. Improving thin, unlinked, image-free pages is a good place to start.
For new articles, use Evertune's Content Studio to create data-driven content that follows these guidelines. Then see whether your pages are being cited — and by which models — with Evertune's Content Analytics.
Frequently Asked Questions about citations in Answer Engine Optimization
What is Answer Engine Optimization (AEO)?
Answer Engine Optimization (AEO) is the practice of understanding and shaping how AI models and answer engines — including ChatGPT, Perplexity, and Google AI Overviews — recommend brands in response to user queries, or prompts. AEO is usually used interchangeably with the term Generative Engine Optimization (GEO), and is increasingly a core discipline for enterprise marketing teams.
Does word count affect AI search visibility?
Yes, but not in the way most marketers assume. The median word count for ChatGPT-cited pages is 941 words — not 3,000. AI models favor content that is specific, structured, and appropriately sized for its topic. Length alone does not drive citation.
Does schema markup help AI visibility?
This isn’t an area we researched for this report, but Forrester says that schema markup helps crawlers accurately interpret, contextualize, and index content.
How do I know if ChatGPT is citing my content?
Manual prompt testing gives a rough signal but doesn't scale. Evertune’s AEO platform tracks AI model citations across categories at scale, so you can see precisely where your content appears — and where competitors show up instead.
What types of pages does ChatGPT cite most?
Listicles dominate — approximately 50% of cited URLs are listicles, with ranked lists making up 58% of that group. Beyond format, cited pages share structural characteristics: moderate length, multiple headers, heavy internal and external linking, and embedded images.
How does being cited by AI models impact being included in the answer?
They're related but not the same thing. Being cited means your URL appears as a listed source. Being included in the answer means your content shaped what the AI actually said.
Most AI answer engines run on a framework called Retrieval-Augmented Generation (RAG) — they retrieve external sources, use them to generate a response, and then surface some of those sources as citations. But a page can be retrieved and cited without meaningfully influencing the answer (read more about using topic relevance to see what sources influence the answer). And content can shape an answer without being explicitly linked.
What drives inclusion is whether your content is clearly structured and usable as evidence for a specific claim. Read more about how AI decides what to recommend. Citation is a measurable proxy — a signal that tells you your content is in the conversation at all.
What other areas has Will Robinson researched?
I’m so glad you asked. Here are three great nuggets from Will on Answer Engine Optimization:
Follow Will Robinson to get new research as soon as it’s released.
Don't forget to share this post
