Do comparison pages still win AI answers?
We sampled 50,000 buying-intent prompts across five platforms. Comparison pages are cited 3.2× more often than feature pages — but the gap narrowed every month this year.
RK
Rui Kimura
Head of Research
Jul 28, 2026 · 9 min read
Every answer engine has a revealed preference for certain page shapes. When a buyer asks "Nike vs Adidas" — or even "best running shoes" — the model reaches for pages that already did the comparative work. The question is whether that advantage is structural or a temporary artifact of what the web happens to contain.
Citation rate by page type50K PROMPTS · JUL 2026
Share of buying-intent answers citing at least one page of each type. Types are not exclusive.
Comparison pages hold the lead on every platform we sample. But the interesting number is the trend: in January the multiple was 4.6×; by July it was 3.2×. As models get better at synthesizing comparisons themselves from primary sources, the premium on pre-built comparisons is eroding — slowly, and unevenly by category.
3.2×more citations for comparison pages than feature pages — down from 4.6× in January. The gap is closing at roughly 0.2× per month.
What changed in 2026
Two things. First, platforms now cite primary documentation more when it is crawlable — the docs line in our chart doubled since ClaudeBot and GPTBot access became table stakes. Second, community threads increasingly beat thin comparison content: a detailed practitioner answer outranks a template "X vs Y" page on four of five platforms.
KEY TAKEAWAYS
→Ship comparison pages for prompts you're losing — they still convert citations fastest.
→Unblock crawlers from docs before writing anything new; it's the cheapest citation win.
→Watch the multiple in your own category — Answr tracks page-type mix per prompt set.