Structuring Pages for Retrieval

To structure a page for retrieval, organise it as a set of clearly labelled, self-contained sections, each answering one question, with tables for comparisons, numbered lists for steps and a short summary. AI engines retrieve and quote passages, so a page built from clean, standalone answers gives them more to use, and makes it far easier for readers to scan.

In this lesson

• How retrieval sees a page
• Question H2s and self-contained sections
• When to use tables, lists and summaries
• Technical basics that keep structure visible
• Exercise and quiz

How does retrieval "see" a page?

Retrieval-augmented systems find relevant passages from documents and pass them to a model that writes the answer (Lewis et al.). In practice, a page is split into pieces, and the pieces most relevant to a question are used. If a section only makes sense after reading three earlier sections, it's weaker as a standalone passage.

Google's guidance on AI features doesn't require special formatting. It says the same SEO best practices apply (Google). Good structure is simply what makes content easy for both people and systems to understand.

Why use question-style H2s?

People ask AI assistants questions, often long and specific ones. A question H2 such as "How much does a sitemap cost to maintain?" does three things:

1. It tells readers exactly what the section answers.
2. It matches the way the question is likely to be asked.
3. It forces the writer to answer, not wander.

Not every heading has to be a question. Labels like "Checklist" or "Sources" are fine.

What makes a section self-contained?

A useful test: copy one section into an empty document. If a stranger can understand it, it's self-contained.

When should you use tables, lists and summaries?

Keep tables simple: clear column headers, one fact per cell, and no merged cells that make the meaning ambiguous.

Which technical basics keep the structure visible?

• Use real HTML headings (<h2>, <h3>) in order, not styled <div> text.
• Render key content in HTML. Content that only appears after JavaScript runs may be missed by some crawlers. Google can render JavaScript (Google), but not every AI crawler documents that it does.
• Avoid hiding answers in images or tabs that aren't in the page HTML.
• Keep one topic per page so sections don't compete with each other.

How long should a page be?

As long as it takes to answer the main question and the important follow-ups, and no longer. Google's helpful content guidance warns against writing to a particular word count because you've heard search engines prefer it (Google). A focused 700-word page with clear sections usually beats a padded 3,000-word one. If a section grows large enough to answer a distinct question on its own, consider giving it its own page and linking to it.

A quick example

Illustrative. A "Pricing" page lists three plans as paragraphs of marketing copy. The restructured version:

• Opening answer: "Plans start at €29 per month for up to 5 users; all plans include a 14-day trial."
• H2: "Which plan is right for my team?" with a comparison table
• H2: "What's included in every plan?" as a bulleted list
• H2: "Can I change plans later?" with a two-sentence answer
• H2: "Do you offer discounts for nonprofits?" with a two-sentence answer

Every section now answers a question a buyer might ask an AI assistant. (The prices are illustrative.)

Page-structure checklist

• [ ] Answer-first opening
• [ ] H2s phrased as the questions readers ask
• [ ] Each section answers in its first 1–2 sentences
• [ ] Sections make sense on their own
• [ ] Comparisons in tables; steps in numbered lists
• [ ] A short summary or key takeaways
• [ ] Real HTML headings; key content visible without JavaScript
• [ ] Sources linked where claims are made

Exercise

Restructure one page: question H2s, a table, a summary.

1. Pick a page with long paragraphs and few headings.
2. List the 5–8 questions it answers, or should answer.
3. Turn each into an H2 and move the relevant content under it.
4. Open each section with a direct answer.
5. Convert one comparison into a table.
6. Add a three-bullet summary near the top.

Quiz

1. Why should sections be self-contained?

a) Retrieval systems often use individual passages b) It makes the page longer c) It's required by HTML Answer: a

1. Which format suits comparing plans on the same criteria?

a) A paragraph b) A table c) An image Answer: b

1. What does Google say about special formatting for AI features?

a) It requires a special schema b) The same SEO best practices apply c) Tables are mandatory Answer: b

1. Which opening makes a section weaker as a standalone passage?

a) "IndexNow submits URLs to search engines…" b) "It also does this…" c) "A sitemap lists URLs…" Answer: b

1. Why render key content in HTML?

a) Some crawlers may not run JavaScript b) It's faster to write c) Images are banned Answer: a

Next: Lesson 3: Comparison and best-of pages

Sources

• Google, AI features and your website
• Google, JavaScript SEO basics
• Google, Creating helpful, reliable, people-first content
• Lewis et al., Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks