Which tool to choose
It depends on what you're looking for and where you watch.
- Recommendations that improve over time and a watchlist that grows: ChatGPT with memory on. It remembers your favorites and your nos, so every request starts from a profile already built.
- Titles available on a specific platform: state it in your request. Some streaming services have a direct link with the AI and send you straight to the title ready to watch; for the others, ask and then check availability yourself.
- Books with reasoned justifications, not just titles: Claude or ChatGPT explain well why a book resembles one you've loved, beyond genre.
The interfaces and integrations change often. If you can't find a feature linked to streaming, get the titles and search for them by hand on your app: the recommendation still holds even without the link.
How to do it
The difference between a banal recommendation and a spot-on one lies entirely in how much you tell about yourself. Three titles with the why are worth more than "I like good films".
- Choose three or four titles you really love and pin down why: the atmosphere, the pace, a theme, the directing.
- Add the nos and the constraints, which narrow things down more than the yeses. The working syntax:
Recommend me 5 films I'll probably love. I liked The Great Beauty, Parasite and Amelie, mostly for the atmosphere, the visual care and the bittersweet stories. I can't stand splatter horror and slapstick comedies. Tonight I'm in the mood for something intimate and under two hours. For each title tell me in one line why you match it to my tastes, and avoid the most obvious blockbusters.
- Give feedback on the first round: "the second was too slow", "the third perfect, I want that vein". The second round is almost always much more on target.
- Before you settle in, check that the title really exists and where it is: the AI sometimes confuses or invents.
A concrete example
Elena had been scrolling the catalogue for twenty minutes without choosing. She opened ChatGPT and gave three series she'd loved, explaining that she liked slow, psychological mysteries, not the action-packed kind. She asked for five recommendations, excluding the most famous titles she'd already seen. The AI proposed a Scandinavian series she'd never heard of, explaining in one line that it shared the dark pace and attention to character of her favorites. Elena checked: it was on the platform she already paid for. She started it that evening.
When it does NOT work (and how to fix it)
If it only recommends famous titles you've already seen
Left to its own devices, the AI draws from the most-cited titles, that is the most obvious ones. Fix: explicitly ask it to exclude the twenty most popular in the genre and to aim for lesser-known gems, or give a precise constraint (a country, a decade, an atmosphere). You force it to dig.
If it recommends a film or a book that doesn't exist
This happens especially with books: the AI can invent a plausible title or attribute it to the wrong author. Fix: before trusting it, verify title, author and year against an external source, the platform's listing or a quick search. A recommendation that doesn't exist is uncovered in ten seconds.
If the recommendations don't match your tastes
On the first round the AI takes a guess on the little you gave it. Fix: don't start over, correct. Say what was wrong with its suggestions, and with each piece of feedback the profile sharpens. Good recommendations arrive on the second or third exchange, not the first.
A tip from someone who really uses it
Build a "taste profile" and have it remembered with memory: titles loved, titles hated, genres to avoid, preferred length. Update it with a rating after each viewing, and the recommendations become more and more yours. Once in a while ask for a title "outside the comfort zone", but calibrated to your tastes: that's how the AI gets you to discover something you'd never have opened on your own.
Frequently asked questions
Can it be done with the free version?
Yes. Asking for recommendations, giving your tastes and receiving a watchlist is text: the free versions handle it. The memory that remembers your tastes from one session to the next is also available, with the limits of the version you use.
Do I really have to tell it what I like?
Yes, and that's the point. The more titles and reasons you give it, the less generic the result. A recommendation based on "I like action films" will be obvious; one based on three specific titles with the why will be targeted.
Does the AI know films and series better than a true enthusiast?
No, and believing that leads you astray. The AI hasn't watched anything: it works on the texts that talk about those titles, and for this reason it sometimes gets plots wrong or invents works. It's not a critic with a taste, it's an engine that matches patterns. It's worth using as a finder of candidates you didn't know about, provided you verify that they really exist and keep the final judgment for yourself.