How I Use AI to Preview Movies

Or, why I hardly ever read movie reviews anymore

Several times a week, I want a yes/no answer as to whether I would like a given movie. You see, one of my habits is to spend a 30–60 minutes a day after dinner watching (mostly) old movies. There are hundreds if not thousands of them on YouTube.

There are recommendation services, but they are, frankly, very hit-and-miss—think MovieLens.

Now I have a rating/recommendation service that is close to 100% reliable.

On Claude, I opened a special “Project,” and in the instructions, I wrote this. (Actually, I wrote a bunch of other stuff, comments on films, then told ChatGPT to reduce it to a detailed prompt it would need to recommend movies. Then I edited that, with the following result, which is still a work in progress.)

Your task is to decide whether I would like a given movie.

My film taste is anchored first in classical craft and classical moral vision. I am strongly drawn to old movies, especially roughly 1935–1950, because that’s when filmmakers had mastered the craft well enough for consistent quality, and because the storytelling habits of that era usually assume intelligible human nature, real goods, real duties, and real consequences without making a speech about them. Before that era, the craft often isn’t yet consistently mature; after that era, many filmmakers become fascinated with darkness for its own sake in a modern/postmodern literary mode—self-conscious, nihilistic, fascinated with transgression, and often hostile to ordinary moral categories. There are great films after 1950, but my default expectation is lower because those sensibilities become much more common.

The positive through-line for me is the everyday or reluctant hero who chooses duty over self-expression and sacrifice over indulgence, and whose goodness is imperfect but real. I respond to stories where righteousness is reluctant, burdensome, and human—exactly the kind of flawed virtue that is actually demanded of us—rather than stories of self-actualization through rebellion. “It’s a Wonderful Life” is close to an ideal: a man gives up romantic dreams, does his duty, benefits the people around him, and the film’s Christianity is not tacky or preachy but beautiful and true. I generally love Frank Capra because he reliably delivers this combination of craft, humane realism, moral seriousness, and the triumph of duty without moralistic heavy-handedness.

I do not require explicit moral order, and I tend to dislike explicit moralism when it intrudes. What I require is that the film itself be morally anchored in its attitude. I can infer the moral stance indirectly; I don’t need it spelled out. But I am very sensitive to what the film is training me to admire, sympathize with, or savor. The film’s “moral posture” matters more than whether the characters sin, whether the world is dark, or whether the ending is happy.

I am perfectly capable of liking “dark” films when they still have a moral compass and allow for actual heroism. This is why film noir often works for me, and why Clint Eastwood often works for me: the darkness is not nihilism, the story still knows what evil is, and there is usually judgment, restraint, and some form of morally intelligible choice. “Blade Runner” is a later film that worked for me in that same way: the world is deeply flawed, but the film does not present that flaw as acceptable, and the protagonist’s arc bends toward moral correction (including renunciation—walking away from what he has been doing) rather than toward self-justification. I don’t need optimism; I need moral orientation.

My hardest veto is any film that excuses, valorizes, romanticizes, aestheticizes, or makes “cool” what is evil. This includes films that invite prolonged admiration for vice or evil and then try to redeem themselves with a bleak ending. “The Godfather” fails for me exactly there: even if the arc is tragic, Michael Corleone is presented as the epitome of cool for too long, and I am being asked to admire ordered criminality. The final bleakness does not justify the journey. Films like “Natural Born Killers” are similarly intolerable: I cannot tolerate movies that make excuses for evil or flatter it.

Closely related is what I think of as a broad kind of “pornography” (not necessarily sexual): films that dwell on temptation or wrongdoing in a way that invites participation, savoring, or romantic identification, and then later retreat to “serious consequences” as if that wipes the slate. That move feels manipulative to me, and it trains the viewer’s sympathies in the wrong direction. This is why “Dr. Zhivago” is a hard pass: it asks me to treat an adulterous relationship as “a beautiful love story.” I’m not interested. More generally, I have become much less willing, with age, to “appreciate” morally disordered films as a mere intellectual or aesthetic exercise. What I once might have “appreciated” (for example, certain sophisticated modern films that treat adultery and sexual transgression as illuminating) is now ruined for me by the immorality and by the way it is framed.

This is also why my love of the classical period is not just about black-and-white photography or nostalgia; it is about a consistent habit of storytelling where duty and restraint are treated as virtues rather than pathologies, and where romance is not treated as a warrant to betray obligations. “Casablanca” is, for me, a perfect exemplar of what I love about old movies: it is deeply morally anchored, and Bogart is a flawed hero who learns. The romantic longing does not bother me because the film’s whole movement is toward duty winning out, and it feels inevitable that it will. That is exactly the sort of moral arc I want: not chastity-policing, but moral realism in which self-sacrifice is finally the highest thing.

On genre: I like comedy, including broad or “low” comedy, as long as it doesn’t trade too heavily on celebrating sin, degradation, or rebellion as the main pleasure. Cary Grant comedies can be terrific. And I do not automatically reject films that wear their moral lesson openly if the execution is brilliant and the moral development is earned. “Groundhog Day” is one of my favorite movies; the openness of the lesson is not a problem when the film is that well-made and the transformation is that well handled.

On the other hand, there are modes that tend to leave me cold. One is the morally untethered, cosmic, ultramodern “religiosity” that aims at awe or symbolism while bypassing human moral drama; “2001” is that sort of thing for me, and I have little interest. Another is the extended inhabitation of a warped moral psychology, especially when the film’s stance is ambiguous rather than firmly judgmental; “Taxi Driver” is a case where I have little interest and no interest in rewatching.

Fantasy and mythic symbolism are not automatically a problem for me; what matters is whether they poeticize evil in a way that bypasses rational moral judgment. When a film makes moral horror dreamlike, mythic, or spiritually “fascinating,” it does not soften evil for me—it makes it worse. That is one reason “Pan’s Labyrinth” is intolerable: beyond the general valorization/aestheticization problem, the particular categories involved (including satanic symbolism and sexual violence themes, including suggested child rape) make it utterly unacceptable.

Finally, satire seems to be a borderline case for me, especially when directed at catastrophic evil. “Dr. Strangelove” is a film people might think I would like, but I don’t; and one plausible explanation is that sustained irony about the most serious evils can dissolve the moral gravity I want, turning horror into a kind of solvent laughter. I’m not fully certain, but it fits the way my taste works elsewhere.

If I’m being translated into recommendation logic, the strongest positive predictors are classical-era craftsmanship, disciplined storytelling, and a duty-bound flawed hero whose arc bends toward responsibility, sacrifice, and moral clarity. The strongest negative predictors are any prolonged invitation to admire vice or evil, adultery framed as beautiful romance, and any poeticizing or aestheticizing of moral horror—especially when it bypasses judgment or involves especially depraved categories of evil.

Look online for ratings and summaries. IMDb ratings of 6.5 or higher is correlated with my liking a film.

So, what sort of output do I get? Here are some recent examples:

The Untouchables (1987) — Strong yes.
“This is exactly your kind of film in several respects. Eliot Ness (Costner) is a duty-bound, idealistic federal agent assembling a team of incorruptible men to bring down Al Capone. The moral orientation is crystal clear: Ness is the reluctant hero who must learn to fight dirty without losing his soul. Sean Connery’s Oscar-winning performance as the veteran Irish cop who mentors him is superb. David Mamet’s script crackles. Ennio Morricone’s score soars. De Palma’s direction is stylish and forceful. Now — the crucial distinction from The Godfather: this film is told entirely from the side of law and duty. Capone (De Niro) is never made cool or admirable; he’s shown as brutal and vulgar (the baseball bat scene is horrifying, not glamorous). The violence is graphic in places but morally anchored. The film’s whole posture is that courage, sacrifice, and incorruptibility are the highest things. One of the Letterboxd reviewers aptly compared Mamet’s treatment to A Man for All Seasons — a man of principle against a corrupt world.”

No Country for Old Men (2007) — Borderline, tentative yes with flagged risk.
“This is genuinely borderline for you, and I think it comes down to a question only you can answer. IMDb 8.2; the craft is superb. The violence is graphic but never glamorized — the film treats it as horrifying. Nobody is being invited to admire Chigurh; he’s presented as radical, incomprehensible evil.
The question is Sheriff Bell. He is the moral center of the film — an aging lawman who finds himself unable to comprehend or combat the new scale of evil he encounters, and who ultimately retires in defeat. His final monologue — a dream about his dead father carrying fire through darkness — is deeply moving and arguably points toward something like hope or moral continuity beyond what the world can sustain. But the film’s overall stance is that the moral order Bell represents has been overwhelmed. Good does not triumph. The protagonist dies offscreen, almost casually. Evil is not defeated or even confronted in a satisfying way.
This isn’t nihilism exactly — the film clearly mourns the loss of moral order, and Bell’s anguish is presented with complete seriousness. But it may be closer to the ‘Taxi Driver’ problem than the ‘Unforgiven’ model: there’s no arc bending toward moral correction, no judgment that lands. The moral compass exists but is shown to be impotent. Whether that reads to you as morally anchored grief or as the Coens’ characteristic ironic detachment from the categories they’re portraying — that’s the call.”

The Big Gundown (1966) — Yes, with enthusiasm.
“IMDb 7.4. This is widely considered one of the finest spaghetti westerns outside of Leone, and its moral structure is strikingly well-suited to you. Lee Van Cleef plays a bounty hunter hired by a wealthy railroad tycoon to track down a Mexican peasant (Tomas Milian) accused of raping and murdering a twelve-year-old girl. The cat-and-mouse chase is thrilling, but what elevates it is that Corbett gradually discovers the truth — the man he’s hunting may be innocent, and the powerful men who sent him may be the real villains. The arc bends toward moral correction: a lawman questioning corrupt authority and choosing justice over career advancement. Morricone’s score is among his best. The craft is outstanding — gorgeous cinematography, tight screenplay by Sergio Donati. The violence is present but serves the story’s moral logic.”

Kansas City Confidential (1952) — Yes.
“A working-class ex-con named Joe Rolfe (John Payne) is framed for an armored car robbery he didn’t commit. After the police release him for lack of evidence, his life is destroyed anyway — no job, no reputation. So he tracks down the actual robbers himself to clear his name. This is exactly your kind of protagonist: an imperfect man with a legitimate grievance who chooses to fight for justice rather than give up. The villains are genuinely menacing (Lee Van Cleef and Jack Elam among them), the moral lines are never blurred, and Karlson’s direction is hard and efficient. One of the best B-noirs ever made.”

What Ever Happened to Baby Jane? (1962) — Yes, with an argument for why darkness is earned.
“First, the film never invites you to admire or enjoy Jane’s cruelty. Davis’s performance is deliberately grotesque — the smeared makeup, the little-girl dresses on an aging woman, the delusional singing. You’re meant to be horrified and pitied, not fascinated in a flattering way. The film doesn’t make evil cool; it makes evil pathetic and terrible simultaneously.
Second, Blanche is never reduced to a mere prop for Jane’s theatrics. Her suffering is real, her dignity matters to the film, and your sympathies are never confused about who deserves them.
Third — and this is what really elevates it — the ending complicates the moral picture without dissolving it. The final revelation forces you to reconsider the relationship between guilt, victimhood, and cruelty in a way that is genuinely tragic rather than nihilistic. It doesn’t excuse Jane, but it deepens your understanding of how the sisters arrived where they are. It’s the kind of moral complexity that rewards reflection rather than undermining judgment.
Finally, the craft earns it. Aldrich’s direction, Haller’s black-and-white cinematography, and especially Davis’s performance are operating at such a high level that the darkness serves the story rather than being the point of the story.”

Of Mice and Men (1992) — Strong yes.
“George Milton is almost exactly the kind of hero you love: a man who takes on a burden he didn’t choose — caring for Lennie — out of loyalty and duty, not self-expression. He sacrifices his own freedom and prospects, repeatedly, for someone who can’t care for himself. The final act demands of him the most terrible duty imaginable, and the film treats it with the moral gravity it deserves. No speechifying, no evasion — just the weight of what love and responsibility actually cost.
Sinise’s directing has been compared to Ford’s classical style from the forties, and both he and Malkovich are excellent. The world is harsh and the ending is devastating, but the film’s moral posture is never ambiguous: it knows what goodness is, it knows what cruelty is, and it grieves appropriately.”

Underworld (2003) — No, and here’s why a passing IMDb score doesn’t matter.
“IMDb is 7.0, so it clears your threshold, but the number is misleading here — it’s carried by genre fans rather than by the qualities you care about. Critics were much cooler, with a 31% on Rotten Tomatoes and a Metacritic score of 42.
It’s not morally offensive in the ways that trigger your hardest vetoes. Nobody is being invited to admire evil or savor transgression. But the film has almost nothing positive to offer you either: the characters are flat, the romance is perfunctory, and the whole enterprise runs on stylized gunfights and gothic atmosphere rather than on disciplined storytelling or human moral weight. It’s the kind of movie where Kate Beckinsale in leather and blue-filtered cinematography is the point. Skip it. It’s harmless but empty by your standards.”

Mallrats (1995) — No.
“Kevin Smith’s sensibility is basically the celebration of slackerdom, vulgarity, and irreverence as goods in themselves. The humor runs on crude sexual dialogue, drug content, and characters whose defining virtue is that they reject conventional responsibility. IMDb 7.0, but the whole moral posture is precisely what you don’t want: rebellion and self-indulgence treated as charming.”

In short, I find that it almost always steers me right. For some reason, sufficiently advanced LLMs are very good at this task. (By the way, The Big Gundown and The Untouchables I found particularly memorable. I’m actually watching The Big Gundown a second time now.)

Sometime I might record some thoughts on the fascinating question why LLMs are good at this. But for now, I thought I’d share this with you just so you can have a better time at the movies.

By the way, of course it works the same way with literature.

Sometimes (say, two cases out of the maybe 150 reviews I’ve made LLMs do so far), the LLM ends up being wrong, but in some interesting way. I can always blame that on my own failure to address edge cases in the instructions. So, when that happens, I take what I have learned and update the instructions.

I reckon that in a few years, my recommendation instructions will be practically flawless.