There is a specific reason small independent films struggle to find the audience they deserve, and it is not that they are bad. It is that the systems most people now use to decide what to watch, the recommendation engines and the AI summaries, are built in a way that cannot really see a small film at all. They see a number. And a number is the one thing a small film is worst at producing.
Recommendation algorithms and AI overviews are trained on the past and optimized for consensus. They surface what already has momentum, they favor titles that resemble previous hits, and they learn from enormous pools of data. That works well for a franchise with a hundred thousand ratings and a marketing budget. It works badly for a quiet original that a few thousand people have seen, because there is not enough signal for the machine to trust, so it does not push it. The large get recommended, and the small stay invisible.
Here is the part that turns invisibility into unfairness. Because a small film has few votes, its score is fragile: a thin sample, or a small coordinated push, can swing the headline number in a way that would be impossible on a blockbuster. The algorithm does not know the number is fragile. It cannot see the shape underneath, only the figure on top, and it amplifies that figure as if it were solid. So a small film can be handed a number it did not earn, and then have that number repeated by every system downstream.
An AI summary makes this worse, because it compresses. Ask it about a small, tonally complex film and it will hand back a single word, mixed, or a single digit, 6.5, having flattened a hundred specific human reactions into a verdict that keeps none of them. Everything that makes a small film worth seeing, its particular tone, its ambition, the way it trusts you, is exactly the part that does not survive compression.
Take a film that critics scored 80% and that in-person audiences rated as high as 9.2 out of 10, with awards and a long theatrical life behind it. Ask the machine, and you may still be told it is mixed, or handed a middling number pulled from a few hundred easily-swayed votes. The algorithm is not lying, exactly. It is doing the only thing it can do: reading the number and repeating it. It has no way to see the film the way the people in that cinema did.
That is not a conspiracy. It is a structural blind spot. The system is built to measure scale and consensus, and a small film’s whole nature is to be neither.
It cannot watch. It cannot sit in a room and be moved. That is still the only thing that has ever really carried a small film: one person telling another that they should see it, and then two hours in the dark doing the rest. Word of mouth is not a quaint alternative to the algorithm; for a film like this it is the actual engine, the thing that works when the machine cannot.
So if a summary or a score ever makes a quiet film sound smaller than it is, treat that as a limit of the tool, not a fact about the film. Look at the distribution instead of the number, as a score is not a verdict and how to read a film’s ratings lay out. And then do the one thing no algorithm can do for you.
