Ask HN: Would You Use a Dating App That Matches Based on Your YouTube History?

[cyphix]

've been thinking about a different approach to online dating—what if instead of swiping based on looks or filling out long personality tests, you were matched based on your YouTube watch history and recommendations?

The idea is to use real behavioral data (what you actually watch) to match people with shared interests. Instead of awkward small talk, the app would also suggest conversation starters based on mutual content (e.g., "You both watch a lot of Veritasium—what's your favorite experiment?").

A few questions for the HN community:

Would you trust an app like this with your YouTube history (or would manual topic selection work better)? Do you think real behavioral data is better than self-reported interests for dating? What are the biggest flaws in this approach that I might be missing? Right now, I’m considering an MVP where users upload their YouTube history manually (Google Takeout) or connect their account via OAuth. Curious to hear your thoughts—does this sound like something that solves a real problem in dating apps?

[ok_dad]

Don't make another trendy AI-based dating service, people are sick of those. They only use them because they are forced to do so. First, figure out how to create a safe dating web site, ensure personal data security and don't collect anything you don't absolutely need, don't use too much telemetry and don't keep it for too long and especially don't use location-based telemetry, eliminate the bot problem which is inevitable, eliminate the problem of child predators which is inevitable, and things like that. The goal is to ensure that you only hold the data you require, and avoid any further data, since in a security breach user data is radioactive. Once you know how to do those things, just make a site where people can post a deep profile and search based on those profiles and perhaps have some sort of questions that help match people on a basic level with hobbies, values, etc. People will figure out how to find folks they might like to date and will take it from there. If you used Match back in ~2008 or so when it was profiles and dating questions, that's the goal. I met my wife there, and it was an alright experience. From stories I hear today, I would probably never date again!

[JohnFen]

I wouldn't be into something like this. I don't think my YouTube history gives a reasonable insight into who I am or who would be a good match for me. Even if it did, I don't think that automated analysis of that history can be done with reasonable accuracy (I'm basing that on how awful YouTube recommendations are).

[cyphix]

"i don't think my Youtube history gives a reasonalble insight into who i am or who would be a good match for me" i think it does tell about your interests if you spend a lot of time on youtube..... about the automated analysis of the history, i don't know that i would need to try it But thanks for your constructive criticism .. let me build something and will show it to you

[toomuchtodo]

How does this determine signal wrt shared core values (honesty, trust, respect, open communication, compassion, authenticity, equality)? Interests ebb and flow over a lifetime, who you are as a person is more material to a successful, lasting partnership.

(married ~17 years, n=1, ymmv)

[cyphix]

i would ask you how the other dating websites signal these core values ?

[toomuchtodo]

They do not (maybe eharmony?). What leads you to believe they are doing a good job? The challenge is that traditional meeting opportunities have declined (church, friends, third spaces), dating apps took off, but everyone is unhappy that they cannot Uber the person they want.

https://www.npr.org/2024/11/07/nx-s1-5050713/people-are-unha...

https://www.theguardian.com/lifeandstyle/2023/oct/28/its-qui...

https://www.pewresearch.org/short-reads/2023/02/02/key-findi...

https://www.psypost.org/the-reality-of-romantic-preferences-...

[cyphix]

I dont believe they are doing a good job , my thought process is that my platform would focus more on shared interests rather than these core values

[toomuchtodo]

Please report back with findings! Wishing you an enjoyable experiment. People are tricky.

[coldtea]

>The idea is to use real behavioral data (what you actually watch) to match people with shared interests.

Many shared interests are not necessarily a good proxy for spousal or romantic compatibility.

[cyphix]

Yes correct, "many shared interests are not necessarily a good proxy for spousal or romantic compatibiliy" after reading your comment for like 10 times , i think then it would not be a typical dating site where your sole aim is to find people with spousal or romantic compatibility , it is more of a "connecting" app where you can have deep conversations with people and some of them might result in romantic relationships. Thinking about it reddit and facebook also do something similar but not exactly similar . they dont match people based on shared interests, not by analysing their youtube history , recommandations and watch later playlist.

Looking forward to your reply

[coldtea]

yeah, for a connecting app it could work well -- especially with the loneliness epidemic and everything going on

[implmntatio]

try the (or some) watch later playlist, with all the stuff a person would do or wants to do 'if' they had [time, xyz] ...

no better way to start doing and getting to know people.

a picture, quick info from what height one jumps from, and an attempt to pack the brutality of one's entire personality into one paragraph.

[cyphix]

hey great idea , watch later playlist also tells a lot about people