There has been a persistent assumption in AEO circles that AI search engines favour editorial content, forum discussions, and community platforms over commercial pages. A citation study by Ten Speed challenges that assumption with some precision. According to their research, product pages account for 24% of AI citations in B2B buying contexts. Reddit and YouTube combined account for just 4%. That is a ratio of roughly six to one in favour of owned, commercial content.
This is not a marginal difference. It points to something structural about how AI answer engines handle queries with genuine purchase intent - and it has direct implications for how B2B teams should be thinking about their product page architecture, not just their editorial programme.
What the Citation Gap Actually Tells Us
Forum content - Reddit especially - has been widely cited as a preferred source for AI models because it contains unfiltered user opinion and experience. That is true for some query types, particularly exploratory or opinion-driven questions. But B2B buyers querying AI search engines are often much further along in a decision process. They are asking about specific capabilities, integrations, pricing structures, or use cases. For those queries, a well-structured product page with clear, extractable information is a more useful source than a Reddit thread from 18 months ago.
The Ten Speed data reflects this. When B2B-relevant queries are the frame of reference, product pages win the citation battle. That does not mean community content is irrelevant - it still has a role, particularly for brand sentiment and top-of-funnel awareness queries. But if your AEO strategy has been skewed towards generating Reddit presence and forum mentions at the expense of your own product content, the allocation probably needs revisiting.
What Makes a Product Page Citation-Ready
AI answer engines extract information differently from how Google's traditional crawler indexes a page for ranking. For a product page to earn a citation in Google AI Overviews, ChatGPT, or Perplexity, the content needs to be structured so that specific claims, features, and answers are immediately extractable without the model having to interpret ambiguous or buried copy. That means clear headings, explicit feature descriptions written in plain language, and answers to the questions your buyers actually ask - not the questions your internal team thinks they ask.
Schema markup plays a supporting role here. Product schema, FAQ schema nested within a product context, and clear entity signals - company name, product name, category - help AI crawlers understand what a page is about and what claims it is making. But schema alone is not the mechanism. The underlying prose needs to be extractable. A product page that buries its core value proposition behind a hero banner video and a carousel of client logos is not going to be cited, regardless of how well-structured its JSON-LD is.
Crawl access also matters. If your product pages are behind login walls, soft-blocked via robots.txt directives that affect AI bots, or slow to render because they depend heavily on client-side JavaScript, the citation opportunity disappears before it starts. Confirming that Googlebot, GPTBot, PerplexityBot, and ClaudeBot can access and render your key product pages is a basic prerequisite that many B2B teams have not yet checked.
The Fact-Check Problem With AI-Generated Product Claims
The Ten Speed study was apparently put through a rigorous fact-check process - six pointed verification questions before the findings were accepted. That detail matters, because it speaks to a wider issue with AI-cited content: AI models can misquote, paraphrase inaccurately, or conflate information from multiple sources. If your product page makes claims that are imprecise, contain internal contradictions, or are expressed in ways that make accurate paraphrase difficult, the AI citation you earn may not accurately represent your product.
This is an underappreciated quality-control problem for B2B marketers. The goal is not just to be cited - it is to be cited accurately. Product pages need to be written with sufficient precision that an AI model pulling a sentence or a feature description into a generated response is pulling something that is correct and complete. Vague superlatives and marketing generalities are not just unhelpful to human readers; they actively increase the risk that an AI misrepresents what your product does.
Measuring Whether Product Page Citations Are Driving Pipeline
Knowing that product pages earn citations is useful directionally. Knowing whether those citations are generating qualified traffic and pipeline is what makes the investment case. AI-referred traffic is still inconsistently tagged in standard GA4 setups. Sessions from ChatGPT, Perplexity, and Gemini can arrive as direct traffic, as organic, or under referral sources that are not immediately recognisable. The referral domains to look for include chat.openai.com, perplexity.ai, and gemini.google.com - but the list grows as AI surfaces proliferate.
Attribution through to CRM requires linking those referral sessions to lead captures and then tracking those leads through to opportunity and revenue. Without that thread, you are optimising for citations without knowing whether they convert. In B2B, where product pages are often the entry point for high-intent queries, that connection matters. A prospect who arrives on a product page via an AI citation, fills in a demo request, and converts to a paying customer is a different kind of lead to one who arrives from a branded search. Understanding which AI surfaces are generating which types of lead - and at what quality - is the measurement problem that most B2B teams have not yet solved.
Where This Sits in a Broader AEO Programme
Product pages earning 24% of citations does not mean editorial content is irrelevant. The remaining 76% is distributed across other source types, and a well-constructed AEO programme needs coverage across the citation mix - thought leadership, third-party editorial mentions, data-driven content, and yes, community presence where it is genuinely warranted. The point is not to abandon forum strategy; it is to stop underinvesting in product content on the assumption that AI search prefers informal sources.
For B2B teams running AEO programmes, the immediate priority is an audit of existing product pages against the criteria that drive AI citation: extractable claims, precise language, correct schema signals, verified crawler access, and honest answers to the questions buyers are actually putting into AI search engines. That is not a content refresh in the traditional sense. It is a structured rewrite with a specific extraction use case in mind. The citation data suggests it is worth doing.