Hi!
I checked the files you sent me. This is a sample output from TrID:
59.0% (.AI) Adobe Illustrator graphics (97501/2/16)
39.6% (.EPS) Encapsulated PostScript (65501/2/14)
1.2% (.PS) Postscript document (2000/1)
The file formats structure is very similar, neary identical. .AI files are in fact a form of EPS (in my understanding).
Considering that theese are indeed .EPS files created by Illustrator, I don't think the results are too far off / bad.
There's the possibility that TrID's definition for .AI files (ai.trid.xml) may contains some tokens or patterns that's not really specific of that fileformat, but only of the files that I have the chance to analyze (over 60 anyway, so that's not a small number). This obviously would cause the score to be a little higher than the optimal.
If you have a large set of .AI files, you may try to refine the .AI def and see if you can get a more specific one.
But, I repeat, I don't see the present results really as wrong.
Thanks for your feedback,
Bye!