I'm reading this as you are searching a Chinese-language project (like
zh.wikipedia.org), and getting results that are mixed Traditional and Simplified Chinese. If that's not the case, please elaborate!
My understanding, which is admittedly incomplete, is that the text for Chinese-language projects is stored however it was entered (Traditional or Simplified), and is converted at display time. If you look at the main page of
zh.wikipedia.org[2] today without being logged in (or in a private browsing window), the featured article link has this text: "2007年欧洲冠军联赛決賽", which uses both 赛 and 賽, with 赛 being the Simplified version of Traditional 賽.[3] If you request the zh-cn version of the page,[4] the text is "2007年欧洲冠军联赛决赛", and both are Simplified "赛". If you request the zh-tw version of the page[5], the text is "2007年歐洲冠軍聯賽決賽", and both are Traditional "賽". So, I believe that explains why you are seeing mixed Traditional and Simplified results.
What to do about it? I can't get the Opensearch API to do the conversion in place, but there is a separate API that does the conversion: Parsing wikitext.[6] Unfortunately, I can only get the API to do the conversion (which is based on the uselang parameter) when I submit the text as wikitext,[7][8] which adds some additional tags and a long comment to the results. \u-formatted input doesn't work, and I can't get the conversion to work for json input (i.e., the result of the Opensearch call). That doesn't mean it isn't possible, just that I haven't figured it out.
I hope that points you in the right direction, and maybe inspires someone who knows this stuff better than me to help out.
—Trey