{"id":31862,"date":"2026-10-19T12:41:38","date_gmt":"2026-10-19T10:41:38","guid":{"rendered":"https:\/\/wordlift.io\/blog\/en\/?p=31862"},"modified":"2026-10-19T12:41:41","modified_gmt":"2026-10-19T10:41:41","slug":"first-hit-wikidata-search","status":"publish","type":"post","link":"https:\/\/wordlift.io\/blog\/en\/first-hit-wikidata-search\/","title":{"rendered":"Why first-hit Wikidata search breaks AI agents"},"content":{"rendered":"\n

A lot of AI agents identify things the same way: search Wikidata<\/a> for the name, take the first result, move on. It is one line of code and it usually works. We measured how often it does not, and what happens when it fails.<\/p>\n\n\n\n

The one-line identity step<\/h2>\n\n\n\n

Here is the pattern, in some form, in many agent tools and RAG pipelines:<\/p>\n\n\n\n

hits = requests.get(\"https:\/\/www.wikidata.org\/w\/api.php\", params={\n    \"action\": \"wbsearchentities\", \"search\": name,\n    \"language\": \"en\", \"type\": \"item\", \"format\": \"json\"}).json()[\"search\"]\nentity_id = hits[0][\"id\"] if hits else None<\/code><\/pre>\n\n\n\n
wbsearchentities<\/code> is a good search box. It matches labels and aliases and ranks the results. It never sees the sentence the name came from, so the first hit is the most likely meaning of the name in general, not in your text.<\/p>\n\n\n\n
Try it on this sentence:<\/p>\n\n\n\n
\n
Orange reported higher mobile revenue this quarter.<\/p>\n<\/blockquote>\n\n\n\n
The first hit for “Orange” is Q187796<\/a>, a commune in Vaucluse, France. The company is Q1431486<\/a>. An agent that takes the first hit now believes a town in Provence publishes quarterly results.<\/p>\n\n\n\n
How often it goes wrong<\/h2>\n\n\n\n
We ran the six gold sets from the resolve() repository<\/a>: 126 short sentences in English, Italian, French, German, Spanish and Portuguese, with 351 marked names and the correct Wikidata<\/a> ID for each. We gave the same names to the first-hit lookup and to resolve()<\/a>, our entity resolution API, which reads the sentence before it decides. These are our own development sets, not held-out test data, so they favour resolve(); the first-hit errors below do not depend on that.<\/p>\n\n\n\n