An independent ranking of AI subscriptions

There is no shortage of “ChatGPT vs Claude” articles. Almost all of them are written once, never updated, and quietly reordered by whoever pays the most. Prices change, usage limits change, models ship weekly — and a comparison written six months ago is worse than useless, because it looks current.

Which AI is built as a continuously updated dataset rather than an article. Every plan, price, limit and benchmark is a dated record with a source attached. When something changes, the ranking changes, and the site says what changed and why.

How it pays for itself

Sponsored placements and affiliate links. Neither can affect a ranking, and this is enforced structurally rather than by policy: the scoring engine’s input has no field for a commercial relationship, so there is nothing for money to attach to. Sponsored units display the product’s true rank next to the label.

The full explanation, including what we are not confident about, is on the methodology page.

Corrections

Data will sometimes be wrong. Provider pricing pages are inconsistent, usage limits are described in incomparable units, and some providers publish almost nothing machine-readable. Where a figure is uncertain we mark it as unverified rather than presenting it with false confidence. Corrections are welcome.