Skip to content
← Blog
Data6 min

Which content formats do AI engines cite most?

In a measurement of one software category, comparison-format pages won 46% of the citations that beat the tracked brand, and blog posts made up 79% of cited content by type. The brand in question already published comparisons, yet held only 6% of citations, which points at angle rather than format as the thing to fix.

Sam ThorntonData, Volexi
Published 31 July 2026

What was measured

We ran a prompt set for a privacy-focused analytics brand against six engines, captured every answer verbatim, and classified all 286 citations that named a competitor rather than the brand. Classification covers four dimensions: source class, content type, content format and content angle.

One category, one brand, one point in time. Treat the numbers as a worked example of what this analysis produces rather than an industry benchmark. Any vendor presenting a single category as a general law is overreaching.

46%
of competing citations were comparison-format pages
Volexi measurement, privacy analytics category, six engines · measured 30 July 2026

Citation share by content format

Share of the 286 competing citations, by the format of the page cited.

FormatShare of citationsWhat it looks like
Comparison46%Head-to-head or multi-tool evaluation with a table
How-to18%Step-by-step instructions for a specific task
Roundup15%Ranked list of options in a category
Definitive guide12%Long-form reference covering a whole topic
Other9%Reviews, listicles, checklists, case studies
Checked 30 July 2026

Does publishing the winning format guarantee citations?

No, and this measurement is the clearest illustration of why. The brand already published comparison pages, the dominant format, yet held only 6% of citations in the category.

When the format matches and the citations do not follow, the gap is usually angle. In this category, 31% of winning citations were written for a named audience segment rather than a general reader, while the brand's own comparisons addressed everybody.

That distinction changes the recommended work entirely. The obvious conclusion from a format table alone is write more comparisons. The correct conclusion here was rewrite the existing comparisons for specific audiences, which is cheaper and faster.

Citation share by content type

The same 286 citations classified by what kind of page they sit on, which is a different question from format.

Content typeShare of citations
Blog post79%
Forum thread8%
Product page8%
Listing or directory3%
Other3%
Checked 30 July 2026

The part that surprised us

Blog posts at 79% is higher than the received wisdom that AI answers lean on forums and reviews. In this category the engines were overwhelmingly quoting ordinary editorial articles, including several from small sites with no particular authority.

That is encouraging if you publish, and a caution against assuming domain authority decides retrieval. Several cited pages came from sites a rank tracker would consider unremarkable.

Common questions

Do these percentages hold in other categories?

Unknown, and we will not pretend otherwise from one category. The method transfers; the numbers should be measured per category rather than borrowed.

Why classify format and angle separately?

Because they fail independently. Matching the format and missing the angle produces exactly the outcome above: the right kind of page, cited by nobody.

Sam Thornton
Data, Volexi

Writes up what the stored answers show, one category at a time.

volexi
MONITOR · ACT · MEASURE
The AI search console for marketing teams and the agencies that serve them.
© 2026 JOLTCLICK LIMITED T/A VOLEXIGDPR-READY · EU DATA RESIDENCY