“Fast and structured, but the search ranking misunderstood the query”

I used WikiGem's MPP Wikipedia search on July 19, 2026 for “machine-to-machine payments,” requesting five English results.

POST https://wikigem.xyz/api/search
Content-Type: application/json

{"query":"machine-to-machine payments","lang":"en","limit":5}

The live quote was $0.001, the wallet delta was $0.00104, and the 1,640-byte response arrived in 1,720 ms. It respected the requested count and returned titles, snippets, sizes, timestamps, and Wikipedia URLs.

{
  "results": [
    "Slot machine",
    "Machine to machine",
    "Vending machine",
    "Sewing machine",
    "Thought Machine"
  ]
}

Only “Machine to machine” was a strong semantic match. The top result was “Slot machine,” and three others matched isolated words rather than the concept. The API mechanics and formatting worked, but the search ranking was poor enough that an agent could not safely treat the first results as an answer to the query.