AI systems do not cite the same kinds of pages for every query. When a user wants to understand a topic, explanatory articles dominate. When a user wants help choosing among products or services, listicles and other decision-oriented formats become far more prominent.
That difference is one of the most useful findings in current AI citation research because it changes the content question from “What format does AI prefer?” to a better one: “What format fits the job the buyer is asking AI to perform?”
Informational prompts favor content that explains. Commercial prompts favor content that helps choose.
The format shift is substantial enough that prompt intent was more predictive of citation type in the Wix/Peec study than either industry or AI model.
What the Wix/Peec study measured
Wix Studio's AI Search Lab and Peec analyzed 75,000 AI answers containing 1,056,727 citations retrieved from ChatGPT, Google AI Mode, and Perplexity. The research segmented prompts into four intent groups: informational, commercial, navigational/local, and transactional. It also examined five industries: eCommerce, health and wellness, home repair, professional services, and SaaS.
The headline finding was not simply that listicles or articles were frequently cited. It was that query intent was the strongest predictor of which content type appeared as a source. In other words, the same AI system can favor dramatically different source formats depending on what the user is trying to accomplish.
Citations analyzed across 75,000 AI answers in the Wix/Peec dataset.
Informational-intent citations going to articles.
Commercial-intent citations going to listicles.
Informational AI queries cite explanation-heavy content
Informational prompts ask AI to help the user understand something: what a concept means, why something happens, how a process works, or what trends are shaping a topic. The source material needs to supply context and explanation rather than simply identify vendors.
That is exactly what the citation mix shows. Articles accounted for 45.48% of informational citations, more than twice the share of listicles at 21.68%. How-to guides contributed another 9.21%.
By contrast, conversion-oriented pages were rarely used. Product pages represented only 3.45% of informational citations, category pages 1.74%, and homepages just 0.42%.
Explain the subject
The user wants definitions, causes, frameworks, examples, or instructions. Articles and how-to content naturally provide the depth needed to construct that kind of answer.
Immediate conversion
A product page can describe an offer, but it is usually a poor source for a broad educational question. The user has not necessarily reached the vendor-selection stage yet.
Commercial AI queries cite decision-oriented content
Commercial prompts ask a different question. The user may already understand the category and now wants help deciding what to consider: the best options, the right provider for a use case, which alternative fits a budget, or which company is strongest for a particular requirement.
Here the source mix flips. Listicles jump to 40.86% of commercial-intent citations, while articles fall to 6.15%. Category pages rise to 12.42%, discussions to 11.44%, product pages to 7.14%, and comparison pages to 4.08%.
The reason is intuitive. Commercial prompts create a choice problem. Sources that name multiple options, organize them around criteria, expose differences, and provide recommendation context can supply the raw material needed for an AI system to construct a shortlist or buying answer.
Informational: “How does enterprise SEO work?”
Commercial: “What are the best enterprise SEO agencies for a multi-location brand?”
The first query needs explanation. The second needs a consideration set, qualification criteria, and market context. It should not be surprising that the most useful source formats change with the question.
Commercial vs. informational: the citation mix side by side
| Content type | Informational | Commercial | What the shift suggests |
|---|---|---|---|
| Article | 45.48% | 6.15% | Strong for understanding; much weaker once the task becomes vendor or product selection. |
| Listicle | 21.68% | 40.86% | Useful in both, but especially strong when AI needs to construct a shortlist. |
| How-to guide | 9.21% | 3.89% | Procedural depth aligns more naturally with learning than comparison. |
| Discussion | 4.42% | 11.44% | Peer opinions and experience become more useful during evaluation. |
| Category page | 1.74% | 12.42% | Category structure becomes more relevant when buyers are browsing choices. |
| Product page | 3.45% | 7.14% | Specific offer details matter more as the query moves toward purchase. |
| Comparison page | 2.55% | 4.08% | Explicit trade-offs are more relevant in a commercial decision context. |
The numbers should not be read as a universal ranking of content quality. They describe what was cited in this dataset under the study's taxonomy. A 6.15% citation share for articles in commercial prompts does not mean commercial articles are useless, just as a 40.86% share for listicles does not mean every commercial query should be targeted with a “Top 10” page.
The pattern continues as intent moves closer to action
The same dataset becomes even more useful when commercial and informational intent are placed inside the full journey. Navigational/local and transactional prompts shift again toward pages that help users locate or act on a known option.
For navigational/local intent, product pages accounted for 21.95% of citations, category pages 18.31%, homepages 13.56%, and profiles 12.89%. For transactional intent, product pages led at 24.88%, followed by listicles at 16.87% and category pages at 14.97%.
That creates a practical continuum. Education favors explanatory content. Evaluation favors decision content. Action favors product, category, homepage, and other conversion-adjacent pages. A complete search and GEO strategy therefore needs more than one content type because the buyer is asking the system to do different jobs at different moments.
Learn → Compare → Locate → Act
Articles and how-to guides are strongest toward the learning end. Listicles, discussions, and comparison-oriented content gain importance during evaluation. Product, category, homepage, and profile pages become more prominent as the buyer approaches a specific action.
Four mistakes marketers can make with this data
First, do not conclude that articles are obsolete. They dominate informational citations and remain essential for expertise, education, category understanding, and first-party depth.
Second, do not conclude that every commercial query needs a self-published listicle. The study identifies the formats AI systems cited; it does not say a brand should publish itself as the winner of its own ranking. In many commercial categories, independent third-party recommendation content can provide context a brand cannot credibly create about itself.
Third, do not convert citation shares into a content-budget formula. A 40.86% commercial citation share for listicles is not evidence that 40.86% of your publishing budget should go to listicles. Your mix should reflect your prompt portfolio, competitive gaps, buyer journey, and what evidence already exists.
Fourth, do not optimize only for the content type. A weak listicle is still weak. AI systems and buyers need useful criteria, accurate facts, meaningful distinctions, credible publishers, and enough context to support a decision. Format creates the structure; evidence creates the value.
The real lesson: optimize for the task, not the model
One of the most strategically important conclusions from the research is that intent differences outweighed model differences. ChatGPT, Google AI Mode, and Perplexity did not behave identically, but the user's task was still the stronger organizing variable.
That changes how marketers should think about GEO. Instead of creating one “AI-optimized” content template and deploying it everywhere, build an intent architecture:
Map informational questions
Identify the definitions, problems, processes, trends, and “how does this work?” questions buyers ask before they are ready to compare vendors.
Build explanatory owned content
Create strong articles, guides, original research, and supporting pages that establish first-party expertise and give AI systems clear material for educational answers.
Map commercial decision prompts
Identify “best,” “top,” “recommended,” “for [use case],” “vs.,” “alternatives,” and other prompts where buyers are deciding who belongs on the shortlist.
Build recommendation-oriented coverage
Prioritize credible Best Of lists, comparisons, reviews, buyer guides, expert roundups, directories, and other third-party decision content that fits those specific commercial prompts.
Measure by intent cluster
Track whether the brand is cited, mentioned, or recommended across informational and commercial prompt groups separately. A blended visibility score can hide very different strengths and weaknesses.
Why an informational-content strategy can still lose the recommendation
A brand can have excellent educational content and still be absent from commercial AI answers. That is not necessarily a contradiction. The brand may be a strong source for explaining the category while lacking enough independent evidence that it belongs in the consideration set.
This is one reason traditional content strategies can look healthy in organic search while producing a weaker recommendation footprint. A large library of articles answers “what,” “why,” and “how,” but the market may rely on different sources when the buyer asks “which company should I choose?”
The reverse can also happen. A brand may appear across Best Of lists and commercial recommendation pages yet lack enough first-party depth to support factual understanding, validation, and conversion once the buyer investigates further.
AI visibility requires different content for different jobs in the buyer journey.
Articles and how-to content are especially important when the user is learning. Listicles, category pages, discussions, comparisons, and other decision-oriented sources become more important when the user is choosing. The strongest strategy does not pick one side; it connects the two.
Turn citation research into an intent coverage plan
Start by grouping your priority AI prompts by intent rather than by keyword alone. Then compare the pages and publishers AI systems actually cite for those prompts. The goal is not to mimic every source. It is to understand what type of evidence is missing from your brand's visibility footprint.
If informational visibility is weak, strengthen the owned knowledge layer: original explanations, research, service education, FAQs, technical depth, and expert content. If commercial visibility is weak, look outward: which third-party listicles, comparisons, reviews, guides, directories, and decision frameworks consistently shape the consideration set?
This is where Mention Engine fits. It helps brands build third-party recommendation coverage around specific content types and buyer questions rather than treating every publisher opportunity as an interchangeable placement. The objective is not more mentions for their own sake; it is stronger visibility in the commercial decision environments where AI systems and buyers are evaluating options.
Do not use one content strategy for every AI query.
If your brand educates the market but disappears when AI starts comparing vendors, the gap is not another generic blog post. Mention Engine helps build third-party recommendation content around the commercial prompts, content types, and publisher environments that shape consideration.
Explore Mention Engine