Most FAQ pages get ignored by ChatGPT, Perplexity, and Google AI Overview not because the questions are wrong, but because the answers are written for humans skimming a page, not for a model chunking that page into scored passages. Large language models use retrieval-augmented generation (RAG) to break content into standalone passages and rank each one for relevance independently. A vague FAQ answer that only makes sense next to the three sentences above it fails that test, even if the question itself is exactly what buyers are asking. Fix the structure, and the same content becomes citable. That structural fix alone has been measured to lift citation rates by 17.3% in a March 2026 University of Tokyo study.

TL;DR

• LLMs don't read pages linearly. They chunk content into passages and score each one independently, so an FAQ answer must be self-contained to survive that process.

• Structured formats like FAQs and comparison tables get cited 2.3 to 2.4 times more often than long-form narrative prose.

• Different models have different tastes: ChatGPT favors long-form authoritative articles, Perplexity leans on Reddit, LinkedIn, and G2, and Google AI Mode sits in between with a slight preference for educational content.

• FAQ schema markup helps humans and search engines parse a page, and it correlates with higher citation rates, but the underlying prose still has to do the work of being self-contained and clear.

• Answer engine optimization (AEO) and generative engine optimization (GEO) require rewriting the content, not just adding structured data on top of it.

About the Author: This article was produced by Simaia, an agentic marketing team that runs AI search audits and builds LLM-optimized content for B2B companies across APAC, including a healthcare SaaS client that went from 0% to 45% AI search visibility in 2.5 months.

Why Do LLMs Skip FAQ Pages That Look Well-Written?

An FAQ page can read perfectly well to a person and still be invisible to an LLM, because the model isn't reading the page the way a person does. Retrieval-augmented generation works by splitting a page into passages, embedding each one, and scoring it against the user's query in isolation [integrisdesign.com]. That means an answer like "It depends on your plan tier, see above" is functionally useless to the model, because "above" doesn't exist once the passage has been pulled out of context.

This is the mismatch at the center of the problem. Most FAQ pages are written in a conversational, referential style: short answers that lean on the question for context, or point back to earlier paragraphs. That's fine for a person scrolling the page top to bottom. It's a dead end for a system that never sees "top to bottom" at all.

Think of it like handing someone a single index card from a stack of trivia questions. If the card says "The answer is 12%, same as the previous example," it's worthless without the rest of the deck. If the card says "Employer payroll tax in this scenario is 12% of gross salary," it stands on its own. LLMs are graders working one card at a time. Every FAQ answer needs to survive being read completely out of context.

What's the Actual Difference Between a Citable Answer and an Ignored One?

The gap is structural, not topical [contently.com]. A citable answer states the fact, defines the term, and gives the number or claim in the first sentence, before any hedging or scene-setting. An ignored answer buries the useful part in the second or third sentence, after a rhetorical question or a soft lead-in like "great question."

Here's the practical contrast:

The second column works because it doesn't rely on anything outside the sentence itself. That's the whole test: could this answer be lifted out of the page, dropped into a chat response, and still make complete sense? If not, an LLM will quietly skip it in favor of a competitor's answer that passes that test.

Does FAQ Schema Markup Actually Get You Cited?

The picture here is more nuanced than a simple yes or no. FAQ schema markup is JSON-LD, structured data meant to help search engines and browsers parse a page's layout, and LLMs largely tokenize that markup as raw text rather than parsing it as structured data the way a traditional search crawler would [ziptie.dev]. At the same time, multiple industry analyses report a real correlation between schema use and citation rates: Presence AI found a 2.4x citation increase for pages using FAQ schema, and other benchmarks have found pages with schema markup are more likely to appear in AI-generated citations. The likely explanation is a dual pathway: schema doesn't get parsed directly by the LLM the way it does by Google's crawler, but it feeds Google's Knowledge Graph and organic ranking signals, which in turn feed AI Overview citations, while the visible, well-structured Q&A text on the page is what gets extracted directly by LLMs. Adding schema to a badly written answer still doesn't fix the answer.

That doesn't mean schema is worthless. It still has value for traditional search features and for organizing a page's information architecture in a way that supports crawlability [frase.io][theaffordableseoservices.com]. But treating schema alone as the fix for a low citation rate is solving the wrong layer of the problem. The prose is the product. Schema is packaging that reinforces it.

A related but distinct question is whether structure elsewhere on the page matters more than the markup itself. It does. Clear heading hierarchies, tight paragraph breaks, and one-idea-per-block writing all align with how RAG systems extract passages [integrisdesign.com][nowspeed.com]. A page can have zero FAQ schema and still get cited constantly if every section is a self-contained, well-labeled block of fact. That's the deeper principle behind generative engine optimization: structure the meaning, not just the markup.

Why Do Different AI Models Cite Completely Different Sources?

Building on the point above, structure alone doesn't guarantee a citation because each model has its own sourcing habits, and this is where a lot of AEO strategy goes wrong by treating "AI search" as one target instead of several. ChatGPT shows a strong preference for long-form authoritative articles and informational content. Perplexity, by contrast, uniquely favors discussion forums and community platforms such as Reddit, LinkedIn, and G2. Google AI Mode sits in the middle, with a more balanced distribution across formats and a slight lean toward educational articles.

This has a direct operational consequence: a company optimizing only its own blog for AEO vs SEO is covering maybe a third of the surface area that actually drives AI referral traffic. A press release pitched to outlets an LLM already trusts, a well-argued Reddit reply, or a LinkedIn post with a clear stance can outperform an on-site FAQ page for citation purposes, depending on which model a buyer is using. This is the kind of gap an AI search audit is built to expose. Simaia runs audits across ChatGPT, Gemini, Claude, Perplexity, and Google AI Overview, testing roughly 50 prompts per client to see exactly which platform is winning citations in a given category and which one the client hasn't touched at all.

How Much Does Fixing FAQ Structure Actually Move the Needle?

Stepping back from the mechanics, the measured impact of structural fixes is large enough that this isn't a marginal tactic. Structured formats like FAQs and comparison tables receive 2.3 to 2.4 times more citations from major LLMs than unstructured, long-form narrative content. Presence AI's benchmark found a similar 2.4x lift specifically tied to FAQ schema use, and a separate analysis of 6.8 million AI citations by Digital Applied found a +0.71 correlation between structural readiness and citation rates. The University of Tokyo's March 2026 study isolated structure as a variable and found a 17.3% lift from structural optimization alone, independent of topic authority or backlinks.

Those numbers matter because they show the ceiling isn't about writing more content, it's about writing the content that already exists in a form models can actually lift. Simaia saw this directly with a global textile manufacturing client: publishing 90 LLM-optimized blog posts in the first month drove AI bot visits up 3.5x year over year and turned inbound leads from roughly one every two months into five per month within two months. None of that required new topics, just content built for how retrieval actually works.

Frequently Asked Questions

What is the difference between AEO and SEO?
SEO optimizes for search engine ranking algorithms and click-through behavior on a results page. Answer engine optimization (AEO) optimizes for being extracted and cited inside an AI-generated answer, which depends on passage-level clarity rather than page-level ranking signals.

Does adding FAQ schema markup improve AI search visibility?
It helps, though not by getting parsed directly by most LLMs, since they largely treat JSON-LD as plain text rather than structured data [ziptie.dev]. Multiple benchmarks, including a 2.4x citation increase found by Presence AI, show a real correlation between FAQ schema use and citation rates, likely because schema strengthens Google's Knowledge Graph and organic ranking signals that feed AI Overview citations. The underlying prose still has to be self-contained and clearly written for LLMs to extract and cite it directly.

Why does Perplexity cite different sources than ChatGPT?
Perplexity has a documented preference for discussion forums and community platforms like Reddit, LinkedIn, and G2, while ChatGPT favors long-form authoritative articles. A single piece of content optimized for one model won't necessarily perform on the other.

How long should an FAQ answer be to get cited?
There's no fixed word count, but the answer must be understandable with zero outside context. If it references "above," "below," or "as mentioned," it needs rewriting.

What is generative engine optimization?
Generative engine optimization (GEO) is the practice of structuring and writing content so AI systems can extract, trust, and cite it in generated answers, as opposed to traditional SEO's focus on ranking positions.

Do I need an AI search consultant or agency to fix this myself?
It's possible to fix in-house with the right structural discipline, but most teams underestimate how model-specific the work is. An AI search audit that tests actual prompts across ChatGPT, Gemini, Claude, Perplexity, and Google AI Overview shows where a company is already appearing and where it's losing ground to competitors, which is hard to replicate without running that audit directly.

How is AI referral traffic different from organic search traffic?
AI referral traffic arrives after a user reads a generated answer that cites or names a company, meaning the visitor often arrives with more context and higher intent than a cold organic click. Capturing and identifying those visitors matters as much as earning the citation in the first place.

About Simaia

Simaia operates as the marketing team for B2B companies across APAC that want to be found by buyers using ChatGPT, Gemini, Claude, Perplexity, and Google AI Overview. It combines the strategic layer (AI search audits, competitor gap analysis, trusted-source mapping) with full execution (on-site content built for LLM extraction, press releases, LinkedIn posts, Reddit replies) and a lead identification layer that surfaces the company name, contact, and role behind every AI-referred visitor. For companies losing citations and pipeline to competitors that show up in AI answers, Simaia runs the entire process end to end rather than handing over another dashboard to manage.

If your FAQ pages, blog posts, or product content aren't showing up when buyers ask ChatGPT or Perplexity about your category, get in touch with Simaia at https://www.simaia.co/ to find out exactly where you stand.

References

1. How to Write FAQs That LLMs Actually Cite (2026 Best Practices) - Contently (contently.com)

2. FAQ Schema for AI Answers: Does It Actually Get You Cited by AI? – ZipTie.dev (ziptie.dev)

3. Are FAQ Schemas Important for AI Search, GEO & AEO? | Frase (frase.io)

4. How to Answer Questions for LLM Citations & FAQ Schema (integrisdesign.com)

5. How To Create FAQs That AI LLMs Will Actually Use | Nowspeed (nowspeed.com)

6. How to Add FAQ Schema for AI Overviews That Works (theaffordableseoservices.com)