CoinGecko dropped web traffic. This ranking kept it.
In May 2026 CoinGecko rewrote its Trust Score to weight order-book depth in place of web traffic. That swapped the one signal an exchange cannot type about itself for one it can. This site did the opposite.
Every exchange ranking is a bet on which signals are honest. Volume is not — a matching engine can print any trades it wants, and several aggregators have been quietly filtering for years. Order-book depth is better — a fabricated book collapses on any real order — but it comes from the same server, and quoting orders you never intend to fill costs the venue nothing on its own market. Proof of reserves is once-a-year, if it happens at all.
Web traffic is different. Tranco counts DNS resolvers in the wild. There is nothing an exchange can publish, buy, delete or attest that reaches it. A domain that Tranco cannot measure is a domain nobody visits, and no engagement bot in the world has changed a Tranco figure because Tranco is not measuring engagement. It is measuring the internet asking for the address.
CoinGecko's original Trust Score, introduced in 2019, weighted web traffic alongside liquidity, API coverage and incidents. It was not perfect — the composite was opaque, the numbers moved without notice — but it kept the anchor. A venue could not out-book its way past a domain nobody was resolving.
The May 2026 update replaced the traffic component with book depth. The stated reason was that traffic had become gameable; the effect was to make the composite far more gameable. A new venue can stand up a $50M book on its first day of business and read as top-tier under the new formula. Under the old one it still had to make people actually visit the site — an obstacle its book alone could not clear.
Why we kept it
The site's composite score puts web rank at ×20 — the heaviest weight in the formula — and ×10 each on years survived, liquidity size across the venue's ten largest markets, reported volume discounted by how far it outruns the venue's own traffic, that reliability ratio itself, liquidity as a share of volume, proof of reserves, KYC, and the aggregator's own grade. Web rank and liquidity size are scored continuously on log scales, so a venue at Tranco #6,000 outscores one at #21,000 rather than sitting in the same tier, and $90M of standing liquidity outscores $30M rather than sharing a band ceiling. The grade is one signal in nine, halved to the 0-5 scale, and it also stays a contradiction at 6/10 or lower. The full model is written up separately.
Two contradictions catch venues that fake around either. A venue claiming a top-ten volume position while its domain draws a fraction of the traffic every peer draws goes to the tail. And every venue carries a reliability ratio — its claimed volume against its own traffic, with 1× set by the ten best venues on the score — that trims it from five times the norm and flags it from twenty-five. Neither of those wires needs a book to fire; both are asking the same question the old Trust Score was asking. What kind of business has these numbers.
What that means for the ranking
MEXC currently sits at #7 on the CEX leaf. Under CoinGecko's new Trust Score it would sit lower, because its measured book is thinner than several venues below it here. Under the old Trust Score it would sit close to where it does now, because its web rank — Tranco #6,637, better than every venue in positions 8 through 20 on this list — carried the weight that book depth carries today. The ordering that reflects how much of a real business exists behind the ticker is the ordering that reader wants, and it is the ordering that survives when book depth turns out to be theatre.
This is not a claim that web rank is a proof of anything. A venue can be widely visited and still take its users' money; a venue can be small and honest. The claim is that within an integrity-first ranking, the one signal the venue cannot control is the one that ought to anchor the composite. CoinGecko had that right in 2019. The change was unfortunate.
The site's full methodology documents every weight, every threshold, and every wash-suspect wire. The Score column on the CEX leaf can be turned on in profile settings — with a number, a letter grade, or a filled bar — and the tooltip on every row breaks down where its number came from. The reasons are stated so a reader can disagree with them directly.
Weights and thresholds are current as of the build of 13 September 2026 and move only when a stated editorial reason requires. The argument does not.