One meerkat at a time takes an elevated position and scans for predators while the rest forage with their heads down. It gives graded alarm calls that distinguish, at minimum, aerial from terrestrial threats and convey urgency, and the group responds differently to each. That much is well established and was demonstrated by playback experiments.
The interpretation of guarding as costly altruism is the part that did not survive testing. Kalahari field data showed that meerkats are more likely to guard after eating well, that sentinels position themselves near burrow entrances, and that a sentinel is usually first into cover when a predator appears. Guarding therefore looks closer to a low-risk activity a satiated animal can afford than to a sacrifice, though it still benefits the group.
Meerkat cooperation elsewhere is genuine and substantial. Non-breeding helpers feed pups, escort them while foraging and teach handling of dangerous prey by delivering disabled scorpions to young animals in graded difficulty. What remains contested is how much of this is explained by kin selection rather than direct benefit to the helper, since group members are usually but not always close relatives.