The same structure serves both, because AI answers are built from search results. A post that ranks is a post that is retrievable, and a post that is easy to extract from is a post that gets quoted. The changes are structural rather than stylistic.

Rule one: answer in the first two sentences

The old shape, setting up context and arriving at the answer near the end, works against you now. Both a scanning reader and an extraction system look at the top of the page.

State the answer, then explain it. If the question is what SEO costs per month in South Africa, the first sentence contains a number, and the next 2,000 words explain what changes it.

Rule two: headings are the questions people search

Write each H2 as the question someone types, then answer it directly underneath. This does three things at once: it matches the query, it creates a clear structure, and it gives extraction systems a clean unit to pull.

Use Google's People Also Ask box and autocomplete to get the actual phrasing rather than your guess at it.

Rule three: every section stands alone

This is the rule most writers break. If a paragraph begins "As we saw above" or "this is why it matters", it cannot be quoted without the surrounding text, so it will not be.

Write each section so it makes sense to someone who lands on it cold. Repeat the subject rather than relying on pronouns. It reads slightly more formally and it is far more quotable.

Rule four: be specific enough to quote

Not quotable	Quotable
SEO takes a while to work	Local rankings usually move in three to six months
Competitive pricing	From R6,000 per month, excluding VAT
Most businesses see good results	A 2% to 5% conversion rate is normal for a South African service business
We have lots of experience	Founded in 2015, 64 clients, 4.9 stars across 200+ reviews

A model cannot cite a vague claim because there is nothing in it. Numbers, dates and named specifics are what gets pulled into an answer.

Rule five: date your facts

"R6,000 per month in 2026" is more useful than "R6,000 per month", because both readers and retrieval systems weight currency. Keep a visible updated date and actually update the figures when they change.

Rule six: mark it up

Add Article schema with the author, publication date and modification date. Add FAQPage schema for the questions section. Add Organization markup identifying the publisher.

This makes your facts machine-readable rather than something to infer. Note that Google retired FAQ rich results for most sites, so the value here is retrieval and comprehension rather than a visible snippet.

Rule seven: name a real author

A named author with genuine credentials, linked to a profile page with their background, outperforms anonymous content. Google's guidelines lean on experience and trustworthiness, and AI systems weight source credibility similarly.

"Written by our team" is a wasted opportunity on every post that carries it.

The shape of a post that works

Title: the question as searched, under 60 characters.

First paragraph: the direct answer, 40 to 60 words. This is the part most likely to be extracted.

Byline: named author, publication and updated dates.

Body: question headings, each section self-contained, tables for anything comparative.

Specifics throughout: figures, dates, real examples.

FAQ section: five or six related questions with short direct answers, matched by FAQPage schema.

Internal links to the relevant service page, so the ranking has somewhere commercial to send people.

What has not changed

It still has to be genuinely good. Structure makes a useful post extractable; it does not make a thin post worth extracting. The posts that get cited are the ones that answer the question more completely than the alternatives, and that part is still work.

Juicy Designs writes to this structure across the site, see content marketing and AI SEO. Related: is SEO still worth it with AI answers and schema markup for AEO.

Frequently asked questions

How do I structure a blog post so it ranks and gets cited by AI?

Answer the question in the first two sentences, use the questions people actually search as your headings, keep every fact specific and dated, and add FAQ and Article schema. AI systems extract short self-contained passages, so each section must make sense on its own.

Does the old introduction-then-conclusion structure still work?

Not for search or for AI extraction. Building context before reaching the answer means the answer sits too deep to be pulled out. State the conclusion first, then support it.

How long should a post be?

Long enough to cover the question properly, which for most commercial topics is 1,500 to 2,500 words. Length is not a ranking factor; completeness is, and completeness usually takes words.

Does schema markup help with AI citations?

It helps machines read your facts without inferring them. Article, FAQPage and Organization markup make the author, dates and answers explicit, which supports both rich results and retrieval.

What stops a post from being cited?

Vagueness. A model cannot quote "competitive pricing" or "industry-leading results". Specific figures, dates and named sources are what gets extracted.