I have a CR subscription.
They liked the handling, agility, and quietness. They also liked the ride quality, so Mazda got the balance between handling and comfort just right, according to CR.
They also complemented the "presence" of the car, and they liked the interior materials.
What they didn't like: lack of seat adjustments, small-ish interior, "tedious and cumbersome" infotainment (that lacks android auto and apple carplay).
They gave middling reviews to the engine - it gets 1 mpg better than larger, V6 powered competitors, but isn't as smooth, which isn't a good tradeoff in their opinion.
Overall, it is ranked pretty low in midsize SUVs (ouch). Highlander, Sorento, Santa Fe, Pilot, Santa Fe Sport, Murano, Edge, Pathfinder, Atlas, 4Runner all score higher.
The "predicted reliability" is 1/5. Unfortunately, the web site does not appear to break out the specific areas and just reports overall reliability. The 2016 and 2018 both get the same score, but there is no score for the 2017 because they didn't have enough data.
Looks like they didn't even drive a 2018. Lack of seat adjustments? Adjusts in all the usual ways. Even the passenger seat has height and lumbar adjustment, which some competitors don't have.
It beats all those mentioned competitors in almost every comparison test by actual car publications. What are they smoking?
1/5 in "predicted reliability" based on what, a crystal ball? It's really quite laughable, but also sad. The 2016 came out a year and a half ago and has performed largely flawlessly. There's examples with over 25,000 miles with no problems. All the while they're "guessing" that this is the next Fiat 500L.
If anyone has any data to show that this is built like a Chrysler, I'd pay to see it.