My answer is yes, they do work, but probably not for the reason you expect. There are many, I have used MyFitnessPal the most, but it doesn’t really matter. The reason why they work is because when you know you have to account for all the food you ingest, you end up being a lot more careful about what you eat. I don’t typically use a food tracker without a plan - I always have a calorie goal. I don’t let them factor in my exercise, I do that separately, I only want to account for the calories coming IN from food. So the times when I have a set goal (typically my BMR) and track what I eat, I end up losing weight every single time. That’s because when I eat at my BMR, any exercise I add contributes towards weight loss.
I really wanted to prove to myself that if I were 100% faithful in tracking calories IN and calories OUT that I could lose weight, so I did an experiment and documented it on my blog. But in my article I also talk about the inherent inaccuracy of not only food tracking, but fitness trackers, scales, pretty much any tool you use. But the great thing is that if you’re tracking and recording, you can tweak what you’re doing until it produces the results that you want.
Last point - is this sustainable? In other words, is it practical to track your food on an ongoing basis, forever? I would say some people might have that discipline, but most people do not, including me. So I actually just use food tracking as one tool among many. I keep records of my weight, take photos, body part circumferences, and fat measurement. If I can tell that I need to tighten up my food consumption (i.e. stick better to my caloric goals), then I use food tracking for a few weeks to get myself back on track.
So consider it an effective tool. Try it for 4 weeks, inaccuracies and all, and I guarantee that you’ll see results. And to your point about how to log food you get from restaurants…well, if it’s a large chain restaurant, MFP has many of those foods already. If it isn’t, then I type in what I ate (like Chicken Parmesan) and review a few different versions and take one that appears to be average. That’s good enough.
Here are the relevant articles: