Question-formatted H2s and answer blocks
The single highest-leverage content-structure change you can ship. Princeton found this pattern alone drives up to 40% citation visibility lift without changing your word count.
The pattern that drives 40% citation lift
Princeton's GEO research analyzed 10,000 queries across 9 AI sources and isolated the content patterns that drive AI citations. The highest-leverage single pattern: question-formatted H2 headings followed by 120-180 word answer blocks immediately below them.
This pattern alone delivers up to 40% citation visibility lift without changing word count. You're not writing more content — you're restructuring existing content to match how AI engines fragment and retrieve information.
The reason: AI engines don't read pages top-to-bottom. They break content into passages, evaluate each passage for citation-worthiness, and emit the most extractable ones in their responses. Question H2s + tight answer blocks are pre-formatted passages. The AI doesn't have to guess where the answer starts and ends — you've labeled it.
Anatomy of a citable answer block
A perfectly extractable Q&A passage has four elements:
- An H2 phrased as a direct question — exactly the way a user would ask an AI. Not "Pricing models" but "How much does X cost?" Not "Implementation guide" but "How do I implement X?"
- The answer in the very first sentence after the H2 — concrete, direct, no preamble. The first sentence is what AI engines pull when they need a one-line answer.
- Supporting context in 120-180 words total — examples, numbers, comparisons. Enough to be informative; short enough that the AI can quote the whole block.
- At least one statistic or named reference — fact density is what tips an extractable passage into a cited passage.
A working example
Bad version (vague heading + buried answer):
## Pricing models
There are many ways to think about pricing in the GEO services industry.
Different operators have tried various approaches based on what their
clients want and what fits their workflow. In our experience working
with hundreds of operators, we've found that retainer-based models tend
to work well for established agencies, while project-based models work
better for solo consultants...
This won't get cited. The heading isn't a query. The first sentence is filler. The actual content gets to the point in paragraph 3.
Good version (same content, citation-ready):
## How do GEO operators typically price their services?
GEO operators most commonly price using monthly retainers ($2,500-$8,000
for SMB clients, $10,000-$25,000 for mid-market), project-based work
($3,000-$15,000 for one-time audits or implementations), or
performance-based contracts tied to citation lift. A 2025 survey of 200+
GEO operators by Reffed Academy found 68% use retainers as their primary
model, 24% use projects, and 8% use performance-based pricing. Retainers
work best for ongoing optimization; project pricing fits scoped
one-time engagements.
Heading is a real query. First sentence answers it concretely. Statistic + source attribution. ~85 words in this example (would be 120-180 in full lesson context). Every element here makes the AI more likely to extract this passage as a cited answer.
Why 120-180 words specifically
The length range matters more than most writers realize. Under 100 words, the passage feels truncated and the AI looks elsewhere for fuller context. Over 200 words, the AI starts paraphrasing instead of quoting, which dilutes attribution.
120-180 is the sweet spot — enough content to satisfy the query, short enough to quote verbatim. Once you've written 50 of these blocks, the rhythm becomes natural. You start writing for the citation, not for the page scroll.
The five question types AI engines retrieve from
Not every H2 question converts equally. Five question types dominate AI retrieval:
1. Definitional ("What is X?")
The most-cited question type. AI engines retrieve definition blocks constantly because most user queries start with "what is..." or "tell me about..." Make sure every key concept on your site has a "What is [concept]?" H2 with a tight definition underneath.
2. Comparative ("X vs Y" / "What's the difference between X and Y?")
High-intent. Buyers ask AI to compare options before deciding. If your category has 3 main alternatives, write the comparison H2 for each pairing. Tables under these H2s get 35% additional citation lift per Princeton.
3. Procedural ("How do I X?" / "How to X")
Action-oriented. AI engines retrieve procedural blocks for tutorial queries. Pair with HowTo schema for maximum extractability. Numbered steps, each step 1-2 sentences.
4. Quantitative ("How much does X cost?" / "How long does X take?")
Specific. Quantitative questions almost always have concrete answers, which AI engines prefer. Always lead with a number range, then add context. "$2,500 to $8,000 per month" beats "It depends on the scope."
5. Evaluative ("Is X worth it?" / "Why does X matter?")
Decision-stage. Buyers near the bottom of the funnel ask these. Answer honestly, including the "no" case (when is X not worth it?). AI engines reward balanced evaluation more than one-sided pitches.
The dedicated FAQ block
In addition to question-H2s throughout the article, every pillar page should have a dedicated FAQ block at the bottom — 5-10 specific questions, each with 60-120 word answers. Mark this block with FAQPage schema (covered in Lesson 1.1).
Why a dedicated block when you already have question H2s: AI engines treat FAQ schema as a strong signal that this content is purpose-built for Q&A retrieval. The schema unlocks structured extraction even when the page's H2s aren't all questions. Belt-and-suspenders approach.
Converting existing pages to the pattern
The fastest way to apply this pattern across an existing site:
- Inventory your H2s. Pull every H2 across your top 20 pages into a spreadsheet.
- Rewrite as questions. Each H2 becomes a question someone might ask AI. Vague headings ("Features," "Benefits," "How it works") become specific questions ("What features does X have?", "Who benefits from X?", "How does X work?").
- Tighten the first paragraph. Move the actual answer to sentence 1. Cut preamble.
- Pad to 120-180 words. If a block is under 100 words, add a concrete example, a statistic, or a comparison. If over 200, split into two questions.
- Add one statistic per block. Princeton found statistics alone drive +22% citation lift. Every block should have at least one specific number with source attribution.
One full page restructure takes about 90 minutes once you've done a few. The first three pages are the slowest because you're learning the rhythm.
Anti-patterns to avoid
- Question H2 with no answer first. Defeats the entire point. If your H2 is "How do I X?" and the first paragraph starts with "Before we get to that, let's first understand..." — cut everything before the actual answer.
- Generic questions with no specificity. "What are the benefits?" is too vague. "What are the cost benefits of switching from X to Y?" is specific enough to be retrieved.
- Fake FAQ blocks. Don't write FAQ entries no real user would ask just to pad schema. AI engines and Google both penalize purely-promotional FAQs.
- Long preamble before the answer. If a passage takes more than one sentence to start answering, the AI looks elsewhere. Cut ruthlessly.
Implementation: restructuring your top 5 pages this week
- Day 1. Identify your top 5 commercial pages (homepage, top product pages, top blog posts by traffic or strategic importance).
- Day 2-3. Restructure page 1 and page 2. Question H2s, answer-first paragraphs, 120-180 word blocks, statistic per block.
- Day 4-5. Restructure pages 3 and 4.
- Day 6. Restructure page 5. Add a dedicated FAQ block (5-10 Q&A pairs) to each page.
- Day 7. Add
FAQPageschema (from Lesson 1.1) to each FAQ block. Re-validate via Google Rich Results Test.
What comes next
Module 1 is the technical foundation: schema, entity, crawler access, content structure. Module 2 builds on that foundation — content production for citation. The five content patterns that drive 30-40% citation lift, including comparison tables, original research, and expert quotation strategy.