I basically agree with John Blackburne.
There are many solutions to compile projects to html5. In addition getting AIR to compile html5 from an AS3 source is a huge job that nobody is going to address at this point.
The closest thing to export to html5 from the same source code is OpenFL (similar to AS3). But this solution unfortunately falls short of other vital aspects of development (video, Native Extensions, etc) ...
Finally, html5 canvas "engine" (which is the output that is usually better for multimedia projects and / or games), is from my point of view much worse in the global than what Flash Player achieved in terms of the balance between agility of development and final framerate of the product.
The only way to take advantage of html5 canvas to the fullest, is to program everything "manually" (without editor), perhaps with a very optimized framework, and optimizing each output element.
However, even in this case, html5 canvas changes a lot in execution speed depending on the platform. For example we developed a few months ago an interactive with animations. On a low / medium range Android was "slow" (stuttering). On a high-end Android, it was smoother, but stuttering at times. On an iPhone 5 (which is already years old), it was very smooth... On a computer with low end graphics card (in which Flash/AIR gpu interactives are very smooth), is good but not perfect. Etc...
For me this is unacceptable today. That an engine is so irregular in framerate according to platform, with the same (optimized) development.
Finally, the problem of AIR right now is not because of unpopularity or negativity towards the platform, but rather because few people know it, by Adobe abandoning it for years while doing the opposite advertising.
In our case when we develop a product in AIR (mainly interactive for events in controlled hardware), the client has no idea what technology we have used. They are simply 100% happy and ask no more.