Why FAQ schema is the single highest-leverage signal for AI citation.

What AI models are actually optimizing for

When a language model generates an answer, it's running a retrieval step first — finding the most relevant passages across a corpus — and then a generation step that composes an answer from those passages. The retrieval step heavily favors self-contained chunks of text: passages that make sense without reading the rest of the page.

A structured FAQ is exactly that. The question is the query. The answer is a self-contained response. The model doesn't need context from the surrounding page. It can lift the pair directly.

Why FAQ wins over prose

Compare two ways of saying the same thing on a service page.

Prose version

Our pricing is straightforward — we charge a flat monthly rate of 2,400 SEK for small businesses, which includes up to 50 receipts per month and quarterly VAT filings. Larger clients with more complex needs can reach out for a quote.

FAQ version

Q: How much does small-business bookkeeping cost?
A: Our small-business bookkeeping starts at 2,400 SEK per month. That includes up to 50 receipts monthly and quarterly VAT filings.

Q: What if I need more than 50 receipts per month?
A: Larger clients can contact us for a custom quote based on volume.

Same information. Both are factual. But when a user asks ChatGPT "how much does bookkeeping cost in Stockholm," the FAQ version is vastly easier to retrieve and cite. The question is pre-matched to the query. The answer is self-contained.

How to add FAQ schema in WordPress

WordPress has had a built-in core/details block since version 6.3. It renders as a collapsible Q&A. Dennis GEO auto-detects these blocks and emits FAQPage schema for them. You don't write any JSON-LD by hand.

Workflow:

1. Edit a page. Add a Details block where you'd normally write a subheading.
2. Put the question in the summary, the answer in the body.
3. Repeat for 5–6 questions per page.
4. Publish. Schema is generated automatically.

Content patterns that get cited

The question and answer themselves matter as much as the schema. Patterns that work:

• Phrase questions as users ask them. "How much does X cost?" beats "What is our pricing?"
• First sentence directly answers. Don't make the reader (or model) wait.
• Include specifics. Numbers, timelines, locations, brand names — these are the signals AI prefers to cite.
• Keep answers under 80 words. Longer answers get summarized; shorter ones get quoted verbatim.
• No filler. Skip "Great question!" and "It depends." Start with the answer.

How many to add, and where

As a rule of thumb: 5–6 questions per service page, 3–4 per product page, 8–10 on a dedicated FAQ page. Don't put them on blog posts — FAQ schema on editorial content can trigger manual actions from Google.

The questions should come from your actual sales conversations. Look at your email inbox, your chat logs, your support tickets. The questions you answer on a weekly basis are the questions AI users are asking too.