[frabcus]

This isn't always the case - APIs often change. Facebook, for example, is (at least was, a few years ago) notoriously bad at changing in an unpredictable and buggy way, and I stopped using it for that reason. Some HTML scrapers are more reliable than that.

As for efficiency, again not such an issue. HTML is very good these days, compared to 10 years ago, a simple CSS selector often does the job.

[thomasrambaud]

This is true, but APIs are often versionned.

Concerning efficiency this is true CSS / XPath processors, at least, both offer very nice performances.

But download 70KB of HTML each time you only need a single data, where the API request cost only a few (avg < 2KB), can be such a pain if you need to do this frequently. This can be handled by a scalable configuration but I find it a bit the overkill.