There is a piece of advice circulating in AEO and GEO circles that sounds reasonable, gets repeated often, and is not doing most brands much good. It goes something like this: AI systems love structured, extractable content; question-and-answer format is structured and extractable; therefore building out frequently asked questions sections is a meaningful AEO strategy. Logical. Incomplete. Expensive if you act on it without understanding what it is missing.

The logic is not wrong. It is just missing most of what actually matters. Question-and-answer format is a surface property. What AI systems actually reward is quality, depth, consistency, and trust: the properties that make content worth citing rather than just easy to pull out. Brands optimising for the format while neglecting the substance are producing a lot of content and earning very little authority from it.

Where the questions-and-answers-as-AEO idea came from

The idea that questions and answers are a primary AEO lever did not emerge from nothing. It has roots in real, observable behaviour. Google’s featured snippets have historically favoured content formatted as direct question-and-answer pairs. AI systems do extract and surface Q&A-formatted content. Structured data genuinely helps search engines parse content. If you follow each of these threads, building out question-and-answer content for AI visibility feels like a logical extension.

The advice also spread at exactly the moment brands were scrambling to understand how AI overview systems and answer engines like Perplexity work. When a new system arrives and the rules are unclear, the consultants and content teams reach for the tools that worked in adjacent situations. Question-and-answer content was a known quantity in a newly unknown context. So it got carried over. The harder question did not: does the format produce citation-worthy content, or just content that looks citation-friendly?

The problem is that producing content in Q&A format and producing content that AI systems trust enough to cite are two different things. The format is easy to replicate. The trust signals are not. They come from something that takes time: genuine depth, consistent publishing, and content that has been fact-checked rather than assembled. A hundred question-and-answer pairs can be produced in an afternoon. The authority that makes any of them worth citing takes months to build. The format is the easy part. It was always the easy part.

What AI retrieval systems actually look for

AI retrieval and answer generation systems do not have a schema preference. They do not look at a page, see frequently asked questions markup, and promote it in their consideration set. What they look for is content that can be trusted to answer a question accurately and specifically, with enough depth and evidence that citing it will not embarrass the system or mislead the person reading the answer. The format of the content is a distant secondary consideration to whether the content is actually reliable.

The signals that drive that trust are built at the level of the site and the body content, not at the level of a format choice. Depth matters: content that covers a topic thoroughly, from multiple angles, with evidence and specificity, outranks content that covers it superficially in question-and-answer packaging. Consistency matters: a site that publishes reliably over time signals ongoing expertise rather than a one-time content push. Accuracy matters: fact-checked content that does not hallucinate, misattribute, or quietly get things wrong builds the kind of source credibility that retrieval systems look for. Topical authority matters: a site that has published consistently and deeply across a subject area is a more credible citation source than one with a thin archive and neatly formatted questions. Information gain matters: content that adds something to the existing knowledge on a topic is more citable than content that restates what is already everywhere. None of these are format properties. All of them require something formats cannot provide: time, expertise, and the discipline to keep showing up.

None of these properties are delivered by formatting choices. There is also a mechanical point worth being clear about: LLMs do not parse JSON-LD the way a traditional search crawler does. They tokenise it as raw text. Most of the semantic meaning in the markup is lost before the model processes anything. Schema does contribute, through Google’s Knowledge Graph pipeline and as a broader quality signal, and it is worth implementing properly. But content quality comes first in the evaluation hierarchy. Schema is the label on the tin. The content is what is inside it. A page with well-implemented schema and thin body content is a page whose structure is legible and whose value is low.

The analogy is a CV. Well-structured, clearly formatted, easy to read. It still gets rejected if the experience is not there. The formatting made the nothing easier to find.

The quality problem with questions-and-answers content produced at scale

Most Q&A content produced for AEO purposes has a specific failure mode: it replicates what is already known rather than adding to it. The format is adopted. The quality is not. A frequently asked questions section asking “what is merino wool?” and answering with a standard definition has zero information gain regardless of how it is structured. The answer exists everywhere. The AI system already has it. There is no reason to cite this page over the hundreds of others containing the same information.

This is the core of why question-and-answer-led AEO strategies underperform. The format signals extractability. It does not signal quality, depth, or trustworthiness. A direct answer to a common question is only citation-worthy if the answer is better than every other answer in the retrieval index: more accurate, more specific, more evidenced. For most Q&A content produced at speed for AEO purposes, it is not. It is the same answer in a different container. The container did not help.

The content that gets cited is content where the source matters. A definition of merino wool could come from anywhere. An explanation of why a specific brand chose merino over alternative materials, based on specific testing and years of product development, can only come from that brand. That is citation-worthy. Not because of the format it is presented in. Because the depth and specificity of the knowledge is irreplaceable. No amount of question-and-answer formatting manufactures that. Only genuine expertise, rigorously expressed, does.

Q&A format has a structural problem as a vehicle for building AI authority. Short answers to isolated questions are not designed to demonstrate depth, hold a sustained argument, or show the thorough engagement with a topic that signals genuine expertise. A page with twenty question-and-answer pairs has demonstrated that twenty questions can be answered briefly. AI retrieval systems are evaluating trustworthiness and depth. Twenty brief answers is not the same as twenty pieces of evidence that this source knows what it is talking about.

Schema matters, as part of a broader quality stack

Structured data is a genuine signal and worth implementing thoroughly. The point here is not that schema does not matter. It does. The point is that schema without quality content underneath it is a well-labelled empty box.

Schema markup helps retrieval systems understand the structure and meaning of content. For certain types this matters a great deal: product schema surfaces attributes, pricing, and availability that parsers cannot reliably extract from prose. Article schema provides publication metadata. Review schema makes ratings machine-readable. These are cases where the markup carries real information. FAQ schema is narrower: it marks up question-and-answer pairs so parsers can identify them, which is useful when the pairs contain something worth finding.

When those pairs contain generic information that restates the obvious, the markup has made generic content slightly more parseable. The quality problem underneath it is unchanged.

The honest position: schema is one important layer in a quality stack. Sprite implements full JSON-LD schema on every published piece because it is part of doing the technical work properly. But it sits alongside topical authority, consistent publishing, fact-checked accuracy, and brand-specific depth. Not above them. Schema tells retrieval systems how to read the content. The content tells them whether it is worth reading. Both matter. Neither substitutes for the other.

When questions and answers genuinely earn their place

Questions and answers have genuine value. Used correctly. A product page that anticipates the questions a buyer has before purchasing reduces friction and increases conversion. A well-structured FAQ page reduces support volume. Internal search works better with clear questions and answers. These are real, commercially meaningful functions. Just not AI citation functions.

A brand that builds questions-and-answers content to serve its customers is doing something genuinely useful. A brand that produces Q&A content primarily as an AI citation tactic, without the underlying depth and quality that makes content citable, is spending effort on the container rather than what goes in it.

Question-and-answer format can be genuinely powerful when the answer contains something only this brand can say. A question answered with a detailed, specific account of a real brand decision (supply chain, product development, material sourcing) is providing information no competitor and no generic AI generation can replicate. That is worth structuring clearly, worth marking up, and worth publishing. The format and schema make it accessible. The depth and accuracy make it citable. In that order, always.

What actually earns AI citations

The factors that consistently predict AI citation are the same factors that predict long-term organic search performance. Depth: content that covers a topic thoroughly, with specificity and evidence, outranks shallow coverage regardless of format. Consistency: a publishing cadence that signals ongoing engagement builds topical authority that one-time content pushes cannot replicate. Accuracy: fact-checked content that can be trusted builds source credibility retrieval systems actively reward. Brand-specific perspective: information only this source can provide is inherently more citable than information available everywhere. Schema and structural signals make all of it machine-readable. But they are the last layer, not the first.

Of these, the content quality properties are the dominant variables. A site with deep topical authority, fact-checked accuracy, and genuine information gain will be cited regardless of whether its questions and answers are neatly formatted. A site with thin, generic content will not be cited regardless of how well-structured it is. The format and schema did not cause either outcome. The content did.

The data is clearer than the vendor studies suggest. Studies claiming Q&A format or frequently asked questions schema increases citation rates by 3x or more typically come from SEO tool companies with a commercial interest in the finding, and most measure informational query performance rather than commercial intent. The more credible signal: independent analysis of AI Overview citations consistently finds that around 93% of citations link to pages already ranking in the top ten organic results. That is the number. Organic authority, built through quality and consistency, is the dominant variable. Format choices are not a meaningful independent predictor once content quality and ranking position are controlled for. The investment that moves AI visibility is the investment in content quality: depth of coverage, consistency of publishing, brand-specific perspective, and the accuracy that comes from rigorous fact-checking rather than fast generation. This is unglamorous work. It does not have a clear before-and-after moment. It compounds over months and shows up in citation rates that content-format tactics cannot produce.

It is also, by a significant margin, what actually works. The brands appearing in AI-generated answers about their product categories are there because they built something worth citing. Not because they marked it up correctly. The markup is the last five percent. The content is everything else.

There is also a structural reason why FAQ-led AEO strategies are a poor fit for ecommerce brands specifically. Most of the citation data used to argue for frequently asked questions schema comes from informational query performance (the “what is,” “how does,” and “why does” queries where AI Overviews appear frequently and Q&A content performs well). For transactional queries, the queries where ecommerce brands most need visibility, the picture is different. AI Overviews appear on roughly 13% of transactional searches, essentially unchanged over the past year, compared to over 80% of informational searches. The queries your customers use when they are close to buying are the queries AI systems are least likely to generate an overview for. A frequently asked questions optimisation strategy built around AI citation is being optimised for the wrong end of the purchase journey.

The opportunity cost of chasing format over quality

Over-investing in Q&A format as an AEO tactic does not just underdeliver. It consumes time that would compound if spent differently. An hour structuring generic questions-and-answers content is an hour not spent deepening topical coverage, strengthening a commercial cluster, or building the internal link architecture that routes authority to the pages that need it.

This is the real cost. Not inefficiency. Opportunity cost. The brands investing in depth, consistency, and accuracy are compounding citation authority month by month. The brands producing question-and-answer content at volume, without the quality underneath it, are very busy and roughly where they started.

Sprite builds for citation by building for quality. Every piece is grounded in the brand’s actual knowledge corpus, not a generic model’s best guess about the category. Automated fact-checking runs after every section. Brand Reflection evaluates the output against the brand’s established voice before anything publishes. Full JSON-LD schema deploys as standard on every piece. The schema is part of the technical stack because doing the technical work properly matters. But it sits alongside depth, accuracy, consistency, and brand-specific perspective. Not ahead of them. Never ahead of them. Questions and answers appear in Sprite content where they serve the reader. Not as a citation tactic. As a format that happens to fit the question. The citation comes from the content being worth trusting. Everything else is in service of that.

Frequently asked questions