The Content Formats AI Search Actually Cites — 6 Formats Ranked by Absorption Rate
Analysis of 18,000+ pages reveals which content formats AI search engines absorb into their answers — and which ones they ignore. Code and statistics dominate. FAQ pages don't work.
For: SME marketing & growth managers | Read time: ~7 minutes
There's a difference between being found by an AI search engine and actually being used in its answer.
Researchers analyzed what happens after an engine retrieves a page. Across 18,000+ pages and 21,000 citation records, they measured which content the engine actually absorbed — meaning the information made it into the final answer — versus what got pulled, glanced at, and discarded.
The finding: ChatGPT pulls an average of 6.88 sources per query, but only about a quarter of them actually end up in the response. The rest get retrieved, skimmed, and ignored.
So what determines whether your content sticks?
The Six Formats, Ranked by Absorption
The researchers classified every page by content format and measured how much each format increased (or decreased) the chance that an engine would actually use it:
| Rank | Format | Absorption Boost | Why |
|---|---|---|---|
| 1 | Code & Configuration | +77% | Copy-paste ready; engines can use it directly |
| 2 | Numbers & Statistics | +62% | Concrete, verifiable, easy to extract |
| 3 | Definitions & Concepts | +57% | "What is X" — engines need these constantly |
| 4 | Comparisons & Pros/Cons | +55% | "A vs B" is one of the most common query types |
| 5 | Step-by-Step Guides | +41% | How-to content with discrete, independent steps |
| 6 | Quotes & Opinions | baseline | Cited less than data-driven formats |
One surprise from the data: pure Q&A pages — the FAQ format — actually showed a small negative effect on absorption. More on that below.
1. Code and Configuration (+77%)
The strongest signal by a wide margin. If your content includes something someone can copy and use immediately — a code snippet, a configuration file, an API call, an automation recipe — engines are dramatically more likely to incorporate it.
This isn't limited to developer tools. A Zapier template, a Notion setup guide, a spreadsheet formula, a CRM workflow — anything that's "here's exactly how to set this up" in a copyable format counts.
2. Numbers and Statistics (+62%)
We've said this before, but it bears repeating from a different angle: specific numbers are the most extractable form of content. An engine can pull "34% faster" or "92% retention rate" and cite it as a standalone fact, without needing the surrounding paragraph for context.
Every page on your site should answer one question: what's the number here? "We serve many customers" → "We serve 500+ enterprise customers with 92% annual retention." The first version is filler. The second version is a citation waiting to happen.
3. Definitions and Concepts (+57%)
A large portion of AI-generated answers are explanatory — "what is X," "how does Y work," "what's the difference between A and B." When an engine needs to build that explanation, it looks for a clear, authoritative definition it can quote.
If your content has a section that says "X refers to..." or "A is..." or "the term means..." — that section is doing heavy lifting for your AI visibility. Include definitions even for concepts you assume your readers already know. The "obvious" definition is often exactly what the engine is searching for.
4. Comparisons and Pros/Cons (+55%)
When someone asks an AI engine "which is better, A or B" — one of the largest categories of AI queries — the engine needs structured comparison data. A table is worth more than paragraphs of prose here, because the engine can extract individual data points directly from the cells.
Structure matters. Compare systematically: by price, by features, by use case, by scale. Include "choose A when..." and "choose B when..." decision logic. The engine will pull from it.
5. Step-by-Step Guides (+41%)
How-to content performs well, but only when the steps are discrete and independent. The engine needs to be able to pull Step 3 — and only Step 3 — and have it make sense on its own.
Use numbered steps. Keep each one short and actionable. Avoid embedding critical context inside a step description where it can't be extracted independently.
6. Quotes and Opinions
Quotes do get cited, just at lower rates than data-driven formats. The problem is structural: an engine can't easily "verify" someone's opinion the way it can verify a statistic. If your main goal is brand exposure — getting your company name in the answer — quotes work fine. If your goal is maximizing citation frequency, lead with data.
Why FAQ Pages Don't Help
You'd think FAQ pages would be perfect for AI search — after all, engines are answering questions. But the data shows pure Q&A pages have a small negative effect on absorption.
FAQ answers tend to be too short to carry context, they rarely include the kind of evidence engines look for (statistics, citations, source links), and the format doesn't give the engine much to build an answer around. A well-argued paragraph with supporting data will outperform a Q&A pair every time.
Keep your FAQ pages for human visitors. Don't count on them for AI visibility.
How to Apply This
| Your Page Type | Focus on These |
|---|---|
| Product pages | Numbers, comparisons |
| Technical docs | Code/config, step-by-step |
| Industry analysis | Numbers, comparisons |
| Educational content | Definitions, step-by-step |
| Case studies | Numbers (results data) |
| About page | Numbers (company data) |
One last thing: don't use just one format. The most-cited content in the study combined two or three — a definition plus a few data points plus a comparison table, for example. Each format gives the engine a different reason to cite you.
Based on: Analysis of 18,151 pages, 21,143 citation records, and 72 extracted features across ChatGPT, Google AI Overview, and Perplexity — measuring evidence genre absorption rates and format-specific citation probability.