Hands‑On Review of Meituan’s AI Assistant “XiaoMe”: Real‑World Strengths, Flaws, and How It Works

As one of the earliest internal‑test users, I received an invitation code for Meituan’s AI assistant XiaoMe and spent two days exploring its capabilities. The main interface immediately presents common use cases, aiming to guide users who have just installed the app.

When I want to eat something, instead of opening Dianping or searching manually on Meituan, I can simply ask XiaoMe to recommend nearby good restaurants. The assistant quickly lists several options and even offers to place the order for me, which dramatically speeds up the ordering process.

However, two notable problems emerged. First, XiaoMe’s automatic matching sometimes selects a store that is not the closest, leading to longer delivery times. In my test, ordering from a nearby Xibei restaurant and applying a 17‑yuan coupon cut the cost by almost half, highlighting a missed opportunity when the assistant chose a farther venue.

Second, the assistant’s understanding of certain requests is limited. When I asked XiaoMe to order a “shark” (a menu item named “shark”), it failed to place the order but still suggested shark‑flavored snacks, which was unexpected. Asking for a “carrier‑aircraft” (航母) resulted in XiaoMe admitting it could not handle the request.

Additional interactions showed mixed results: searching for my CSDN profile returned decent information, but asking XiaoMe to hail a taxi or grab a KFC coupon was unsupported. The assistant also mis‑interpreted the phrase “hand‑tear product manager” (手撕产品经理) as “first‑time” (首次), indicating shortcomings in language nuance.

From a technical standpoint, XiaoMe runs on Meituan’s self‑developed LongCat‑Flash‑Chat model, which offers fast inference and strong personalized recommendation abilities. Natural‑language ordering lowers the usage barrier, but the current feature set remains tightly bound to Meituan’s ecosystem.

In summary, XiaoMe feels similar to Alipay’s earlier “支小宝” – both use AI to serve daily life needs. While the product is still in public‑beta and exhibits many “pain points” (limited task coverage, occasional misrecognition, ecosystem lock‑in), its rapid response and personalized suggestions suggest promising future improvements.

For readers seeking an invitation code, the author notes that many reviewers share codes on XiaoMe’s official Little Red Book (Xiaohongshu) account.