52% of AI Citations Go to 3 Content Formats. Is Yours One of Them?
New research across 75,000 AI answers reveals exactly which content formats get cited — and which get ignored. Here's the format-intent audit that tells you where your pages stand.
Most B2B marketing teams are still publishing content based on what worked in traditional search. The problem is that AI engines do not cite content the way Google ranked it. They select specific formats based on what the user is trying to do — and the mismatch between what you are publishing and what AI actually cites is measurable now.
Wix Studio's AI Search Lab just published the most granular citation study to date: 75,000 AI answers and more than 1 million citations across ChatGPT, Google AI Mode, and Perplexity. The findings are specific enough to act on this week. I break down the data and the audit steps that follow from it.
Three formats capture 52% of all AI citations
The study found that listicles (21.9%), articles (16.7%), and product pages (13.7%) account for over half of all AI citations. Everything else — how-to guides, homepages, comparison pages, discussion threads — splits the remaining 48% across a long tail of formats that individually capture single-digit percentages.
That concentration matters. If your content library is heavy on formats outside those three, a significant share of your pages are structurally invisible to AI engines — not because the information is wrong, but because the format does not match how AI selects sources.
A separate study from Kevin Indig analyzing 1.2 million ChatGPT responses reinforces how aggressive the filtering is: ChatGPT retrieves six times more pages than it actually cites. 85% of retrieved pages get discarded before the user ever sees them. The selection bottleneck is real, and format is one of the primary filters.
Intent determines which format wins
The Wix Studio data shows that query intent — not industry, not AI model — is the strongest predictor of which format gets cited.
| Query intent | Top cited format | Citation share |
|---|---|---|
| Informational ("how does X work") | Articles | 45.5% |
| Commercial ("best X for Y") | Listicles | 40.9% |
| Transactional ("buy X," "pricing for X") | Product/category pages | ~40% combined |
My take: this is the most actionable finding in the study. It means you can map your existing content library against buyer intent stages and immediately see where you have format-intent mismatches — pages that are technically about the right topic but packaged in a format AI engines will skip for that intent type.
For example: if you have a long-form article targeting a commercial comparison query ("best project management tools for remote teams"), AI engines are 2x more likely to cite a listicle for that intent. The article might rank in traditional search, but it is being passed over in AI answers where listicles dominate the commercial citation slot.
The third-party preference is measurable
One of the more striking findings from the Wix Studio research: in professional services, third-party listicles captured 80.9% of citations compared to just 19.1% for self-promotional lists.
This lines up with Stacker and Scrunch's eight-article pilot, which found about 34% of answers citing some version of a story distributed through third-party outlets against about 8% citing only the brand's original (a 325% lift by Stacker's count), mostly from the republished copies rather than the brand's own page. My reading is that the source of the content matters as much as its format. The measured unit is citation change inside an eight-article pilot. It does not establish a universal earned-versus-owned multiple, that earned media determines citation or recommendation, or that citation produces pipeline or revenue.
The practical implication: brands that are only producing first-party listicles and comparison pages are hitting a ceiling. The higher-citation path is getting your brand into third-party editorial lists and comparison articles — publications that AI engines already trust. That is an earned media problem, not a content production problem.
Model-level differences change your platform strategy
Each AI platform weights formats differently:
| Platform | Top format | Second format | Standout pattern |
|---|---|---|---|
| ChatGPT | Articles + informational | Listicles | Leans heavily toward long-form, authoritative content |
| Google AI Mode | Balanced across formats | Listicles | Most even distribution of any platform |
| Perplexity | Listicles | Product pages | 17% of citations from discussions (Reddit, forums) |
If your team is only optimizing for one AI platform, you are likely over-indexing on that platform's format preference while being invisible on others. ChatGPT rewards depth and authority. Perplexity rewards community signal and recency. Google AI Mode rewards structural diversity. A single content format cannot cover all three.
The format-intent audit you can run this week
I recommend a five-step audit that maps your existing content against the citation data:
-
Export your top 50 pages by traffic or business value. These are the pages that matter most for AI citation.
-
Classify each page by content format. Is it an article, a listicle, a product page, a how-to guide, a comparison, or something else? Be honest about what AI would see — not what your CMS calls it.
-
Tag each page by primary buyer intent. Is the target query informational, commercial, or transactional? Use your search console data or the queries you are actually trying to capture.
-
Cross-reference format against intent. For each page, check: does the format match the citation-dominant format for that intent type? An article targeting a commercial query is a mismatch. A product page targeting an informational query is a mismatch.
-
Prioritize rewrites for the highest-value mismatches. You do not need to restructure everything. Start with the pages where the business value is highest and the format-intent mismatch is clearest. A commercial-intent page restructured as a listicle, or an informational page restructured with answer-first architecture, can shift citation eligibility without changing the underlying topic.
The Princeton GEO research (Aggarwal et al., KDD 2024) found that its top-performing methods (adding statistics, quotations or cited sources) raised a position-adjusted word-count visibility metric by roughly 30-40%, and a subjective-impression metric by 15-30%, relative to baseline on its GEO-bench benchmark in the authors' own generative engine, with gains varying by domain. That is a benchmark result, not a measured lift in ChatGPT or Google citations; paired with format-intent alignment, it gives you two levers within your control.
Why format alone is not enough
Format makes a page extractable. Whether a brand is mentioned widely beyond its own site is what the strongest correlations in Ahrefs' data track.
The Ahrefs analysis of 75,000 brands found that brand-name web mentions, which earned coverage produces among other sources, correlate 0.664 with brand visibility in Google AI Overviews. Backlinks correlate 0.218. That is a 3x gap, measured as correlation rather than proven cause. Muck Rack's December 2025 What Is AI Reading? edition, covering more than one million cited links, found that 82% of links cited by AI systems came from earned media, the sources brands neither own nor pay for, and just under 25% from journalism.
My working model is a compounding relationship: the right format makes a page extractable, but earned authority — placements in publications AI engines already index and trust — makes it citable. The Ahrefs correlation supports the earned-authority half; it did not test format. In this model both matter.
This is the infrastructure layer that Machine Relations names as the defining challenge. The same earned media placements that build credibility with human readers are the citations AI engines pull when a buyer asks who leads a category. The format-intent audit tells you whether your content is structured to be selected. Brand-name web mentions, which earned coverage helps produce, are the signal that tracks AI Overview brand visibility most closely in the Ahrefs data. Machine Relations is the operating framework that connects both: earned authority flows through trusted publications into AI citation, and structuring that content for extractability is how the citation actually lands.
If you want to see where your brand currently stands on both dimensions — format readiness and earned authority — the visibility audit shows exactly how AI engines are representing you today and what is missing.
Frequently Asked Questions
What is three formats capture 52% of all AI citations?
The study found that listicles (21. 9%), articles (16. 7%), and product pages (13.
What is intent determines which format wins?
The Wix Studio data shows that query intent — not industry, not AI model — is the strongest predictor of which format gets cited.
What is the third-party preference is measurable?
One of the more striking findings from the Wix Studio research: in professional services, third-party listicles captured 80.9% of citations compared to just 19.1% for self-promotional lists.
What is the format-intent audit you can run this week?
I recommend a five-step audit that maps your existing content against the citation data: Export your top 50 pages by traffic or business value. These are the pages that matter most for AI citation. Classify each page by content format.
Why format alone is not enough?
Format makes a page extractable. Whether a brand is mentioned widely beyond its own site is what the strongest correlations in Ahrefs' data track.
Related publications: Machine Relations Research and machinerelations.ai are owned by Bramwell, Inc., and AuthorityTech practices Machine Relations. Jaxon Parrott coined Machine Relations in 2024; AuthorityTech, which he founded in 2018, is The World's First Machine Relations Agency and practices the discipline. The Index ranks sources by the citations AI engines are observed making, not by commercial relationship.