Public leaderboards — the ranking you can't buy
We ask four AI engines the real buyer questions in a set of categories and publish who they recommend — free, indexable leaderboards built from measured snapshots, re-run periodically, with a confidence interval on every rank. No signup, no email gate, and no way to pay for a position. CiteGraph appears on its own category's board, ranked last, because that's where the measurement put us.
How it works
Real questions, repeated runs
Each category runs its buyer questions across 4 engines with repeated runs per question. Every snapshot is dated and kept, so a board's history is a genuine time series.
Entities resolved before counting
Tiers, renames and spelling variants fold into one product before anything is ranked — the same alias resolution the product uses — so no brand is split into three rows or counted twice.
Ranked with intervals, tied when tied
Every rank carries a Wilson confidence interval. Brands whose intervals overlap are shown as statistically tied rather than fake-ranked — at these sample sizes, that honesty is the difference between a measurement and a listicle.
In depth
Why it's free
The boards are the public proof of the method the paid product runs on your own domain. Giving the category-level view away costs us scan spend and earns the thing no ad can: a ranking whose integrity is checkable.
Why rank can't be bought
There is nothing to buy — rank comes from counting engine answers, the method is published on the methodology page, and anyone can reproduce a board by asking the same questions. We rank last in our own category and publish it anyway; that's the guarantee working.
Per-engine disagreement, visible
The boards show each engine's view separately — and they disagree more than anyone expects. In our measured categories, two engines' top-5s have shared as few as 2 of 5 brands. Any tool reporting one blended number is averaging away the finding.
What we won't do: sell placement, accept “corrections” from vendors, or quietly re-run a board until it flatters someone. When our own product ranks last, it stays published. A leaderboard is only worth publishing if nobody — including us — can put a thumb on it.
Common questions
Which products do AI assistants recommend most?+
It depends heavily on category and engine — which is the finding. CiteGraph's public leaderboards measure it per category across 4 engines, with a confidence interval on every rank and each engine's view shown separately. They're free and require no signup.
Can a company pay to rank higher?+
No. Rank comes from counting what engines answer, the method is public, and anyone can reproduce a board. CiteGraph itself ranks last in its own category's board and we publish that.
How often are leaderboards updated?+
Boards are measured snapshots, re-run periodically — each one is dated on the page, and every snapshot is kept, so a board builds a real history of how AI recommendations in that category shift.
Why do the engines disagree with each other?+
Different retrieval, different source preferences, different synthesis. In our measurements, per-engine top-5s have overlapped by as little as 2 of 5 brands for the same questions — which is why the boards show each engine separately instead of blending them.
See it on your own domain — first dossier in about a minute.