Four numbers.
Every receipt.

Every score you have ever read was built by someone with a reason to move it — a studio buying placement, a campaign run from a bot farm, a site whose traffic depends on outrage. You were rarely told which, and never told how the number was assembled.

The deeper problem is not that the critics are wrong. It is opacity. You cannot see how a score was calculated. You cannot tell whether a review came from a person, a bot, or a paid operative. You cannot tell whether a number reflects broad agreement or a vocal minority. And you are handed one figure that pretends to answer four different questions.

We are not going to fix reviews by finding better critics. We are going to fix them by showing where every number came from and letting you decide what it is worth.

Every title displays four numbers. We do not publish a fifth number that combines them, and we will not.

When the four disagree, that is not a defect we failed to resolve. It is the most useful thing on the screen, and the product is built to point at it.

A counted review costs one dollar a month. Not because an opinion is for sale — because fraud at scale needs free accounts. A dollar plus evidence you saw the film is a wall a bot farm will not climb to move one score.

We check that you watched it. Review slots open after the credits roll, not before. A season needs six episodes behind you. For a cinema we never see the transaction, so you show us the ticket instead; while we are checking it, it counts for nothing, and if we cannot accept it we tell you why rather than letting it sit.

The free tier is not a trial and it is not crippled. You see all four numbers, always. You simply cannot move one. If you cancel, your existing reviews stay visible and stop counting; nothing is deleted and nothing is rewritten.

Not every counted review counts equally, and you should know exactly why. Weight comes from one thing: whether your ratings hold up as a title settles. Not tenure, not volume, not followers, not how much you paid.

Every account starts at the same baseline and stays there until accuracy is demonstrated. A poor record earns nothing; it is not punished. And only a review that counted can earn anything — a review outside the pool stays readable on the title and earns its author nothing. Credit never flows from a number your review did not move.

There is no quality score attached to any person or any review. We do not publish one, we do not show one privately, and we do not keep a leaderboard. A number like that becomes a thing to optimise, which is the engagement metric we deliberately removed.

Reviews that say nothing about the film do not count toward Member Consensus. The test is narrow and we will state it plainly: does the review contain at least one claim that would be false, or simply inapplicable, if it were pasted onto a different film? “Loved it, 10/10” survives that paste. “The third act drops the sister entirely” does not. That is the line, and it is the whole line.

Four things are excluded and only four: a review with no claim about the title; a review about something other than the film; text that is not the author's; and text that cannot be read as language. The author is told which one, immediately, in a sentence that says what to do about it. The review stays on the page — de-published from the pool, never hidden and never deleted.

What we will never exclude for, because each is a legitimate review: being short, being negative, being ungrammatical or misspelled, being written in another language, being unpopular or alone, declining to spoil, or being felt rather than argued. Two hard prohibitions in the implementation: length is never an input, and no vocabulary or keyword list is ever an input. If naming a craft term helped you pass, the filter would be a vocabulary test and it would be gamed in a week.

This applies to the member pool only. General Audience is genuinely unfiltered, including reviews that say nothing — that is what unfiltered means. It exists as the control group, and cleaning it up would erase the clearest evidence we have: the gap between a brigaded General Audience number and a steady Member Consensus.

We will not claim this catches sophisticated fraud. A patient adversary can write one true sentence about a film. The dollar, the viewing requirement, and anomaly detection are the wall; this is only the thing that keeps the pool from filling with rows that carry no information.

An upheld spoiler report changes exactly one thing: the review's text sits behind a tap. It does not remove the review, change its weight, exclude it from Member Consensus, or touch the author's standing. There is no strike, no ladder, no warning language.

Two facts are recorded separately and shown separately: what the author declared, and what a moderator decided. A veil never rewrites an author's own declaration. Members who flag their own spoilers unprompted are doing the thing we want most, and they will never be filed alongside members who were reported.

Every decision that goes against you carries one tap: I think this counts. That opens an appeal. Only appealed decisions enter the queue, so its depth is an honest measure of how often we are wrong.

We answer within 72 hours, and within 24 hours for a film in its opening week — because an appeal about a film still in cinemas is worth less every day it waits. The deadline is fixed when you file and never recalculated. If we later shorten our own standard, appeals already open keep the promise they were given; an announcement changes what we owe next, not what we already owe.

If we miss a deadline, you are told that we missed it. Not quietly, and not only if you ask.

When a number moves for any reason other than new reviews arriving — a manipulation campaign removed, a methodology change, a correction — we publish an audit note on the title with the date, the reason, and what the number was before. Permanently. If a methodology change affects existing scores, both the old and the new figure are shown through the transition.

Scores are ours. Almost nothing else is, and we name the sources on the surfaces where they appear.

Theatrical showtimes are supplied by MovieGlu and credited on the showtimes screen. Title metadata — casts, crews, synopses, release dates, artwork — comes from TMDB. Streaming availability is surfaced through TMDB's providers data, which originates with JustWatch. Where a partner requires specific credit, that credit appears on the screen the data is on, not buried in a legal page.

Today we do not sell tickets and cannot see seat availability — booking hands off to the cinema or its own ticketing partner. If that ever changes, this section changes with it, and we will state plainly what we earn on a ticket before you buy one. If a theatre is not listed, we do not have a feed for it, and we will say that rather than guess at times.

We collect payment details through Stripe and never store raw card data. We keep your reviews and ratings. We keep viewing history only where you have opted in for Taste Match.

We do not sell user data to studios, streamers, or advertisers. We do not train commercial models on review content without explicit opt-in. We do not share individual ratings with third parties. You can delete your account at any time; reviews that formed part of a historical score remain in the archived dataset, anonymised.

Proposed changes are published with 30 days of comment. Material changes to scoring methodology carry 60 days' notice and a summary of the expected effect on existing scores. We never alter past scores without public disclosure. This charter is also read aloud, in full, inside the app — by a synthetic voice that is disclosed as synthetic every time it speaks.