I Tried to Build a Travel Map From YouTube Food Videos and Stopped at a Legal Wall
With a trip coming up, I got greedy. If I could pull just the restaurants out of my favorite food YouTubers' videos and pin them all on a map, I'd never get lost abroad. I had an AI do it, and technically it really worked. But building it made me more and more uneasy, and I stopped at a wall. This is a non-developer learning that 'can do' and 'may do' are different.
1. The Idea: Restaurants From Videos Onto a Map
Food videos are full of information. The problem is it's trapped inside the video. I didn't want to re-scrub a 30-minute clip going "what was that ramen place called?" So I thought: what if the AI watched the video and pulled just the restaurant list?
2. The Surprise: The AI 'Watched' the Video Directly
I imagined scraping subtitles, but the AI proposed something else. Hand it the video URL and it watches the video and audio directly, returning a list of "restaurants you can actually visit in this video" — name, area, menu, as a table. When it worked I got a little chill. It even caught a shop name shown only on a sign, absent from any transcript.
3. First Stumble: A Restaurant Pinned in the Wrong Region
To put it on a map you need coordinates. But when it couldn't find a location exactly, I'd told it to "fill in something close enough" — and one restaurant landed hundreds of km from its real region. On the map it looked plausible; I nearly believed it.
Lesson: 'fill in something' creates quiet lies. 'Leave it blank if unknown' is better. A blank is visible; a plausible wrong answer isn't. So I removed the fill-in logic entirely.
4. Then the Real Wall — "Am I Even Allowed to Do This?"
As I built, I kept feeling uneasy. Is it right to pull the essence out of someone's hard-made video and use it in my service? So I actually looked into it. There were three lines.
4-1. Copyright — 'Facts' Are Fine, 'Expression' Isn't
Objective facts like name, location, menu aren't protected by copyright. But a YouTuber's commentary, impressions, evaluations, and phrasing are their creative work. So I told the AI in no uncertain terms: "never reproduce the presenter's comments — extract dry facts only."
4-2. Unfair Competition — a 'Copy-the-Whole-Effort Substitute' Is Risky
This was the most concrete wall. Taking what someone built with substantial effort and investment and using it as-is in your own business to replace the original is where trouble lies. If my one map removes any reason to watch that YouTuber's channel, that's exactly the situation.
4-3. Accuracy — Wrong Info Is Another Liability
Restaurants close and relocate. Asserting 'go here' from stale info is its own problem.
5. So I Stopped, and Changed Course
Not giving up — redirecting. I set lines to hold.
- Extract facts only. Don't reproduce the presenter's expression.
- Cross-check extracted facts against other public sources (public maps, ratings).
- Always credit the source channel so people go to the original. Be a guide, not a substitute.
- Refresh periodically, and flag or drop anything uncertain.
The result became less 'a map that copies others' and more 'fact notes for me + links back to the original.' A bit less fun, but a lot easier on the conscience.
6. Lessons a Non-Developer Took Away
- Just because it's technically possible doesn't mean it's allowed. That's the biggest thing I learned this trip.
- No 'fill in something.' A blank is more honest than a plausible wrong answer.
- Don't ignore the uneasy feeling. "Am I allowed to do this?" is usually a signal to go find out.
7. FAQ
- Q: So, are you using that map?
A: Only as personal 'fact notes' for reference. Whether to push it as a service to others, I'm considering only within the lines above. - Q: Is extracting data from videos itself illegal?
A: Referencing 'facts' is generally fine, but copying expression or substituting the original is risky. For specific cases, expert advice is the answer. - Q: Isn't this judgment hard for a non-developer?
A: It is. So before building, I ask the AI "is there any legal problem with this?" first. It's not perfect, but it at least prevents the 'never even thought about it' kind of accident.
8. Closing
This time it's not a 'built something cool' story but a 'stopped' story. Yet to me it was worth more. The more we live in an era where anyone can build without code, the more 'may I build it' matters over 'can I build it.' Being able to ask yourself that question — maybe that's the real skill a non-developer needs.


