Naver Knowledge iN: search users wrote the answers, and Naver indexed every one

When Naver launched in 1999 with Korea's first homegrown search engine, the Korean-language web was small — everyday questions in Hangul had nowhere to live. In October 2002 NHN answered that with Knowledge iN (지식iN): a place to ask anything and get answers from other users, with every exchange stored and 'continuously databased' so it could surface in search results, as Digital Times later described in its history of the company.

The loop compounded. Answers earned 'naegong' points and answerer ranks; answered questions became pages that searchers actually clicked. In its first decade Knowledge iN collected over 100 million questions from 13 million askers, answered by 7 million users. NHN took the top spot in Korean search from Daum in 2003, and in 2008 became the first Korean internet company to reach 1 trillion won in revenue — with Knowledge iN cited as its representative service.

The corpus was also a moat: Knowledge iN content was searchable only inside Naver, so the pages Google could crawl were exactly the ones Naver's machine kept out of its index. The mechanism degraded over time — a cited study found roughly 35% of adopted answers were never corrected, answer-farming and ads cluttered results, and by 2011 columnists were publishing 'search Google instead' pieces. Naver responded by layering verified experts over the crowd.

Don't crawl the answers — have users write and index them

The real constraint was language-scale content, and only native users could produce it at the needed rate — no crawler or editorial team could.

Indexed Q&A pages made each contribution permanent and findable, so a single answer kept paying the search engine back for years.

Points and ranks turned a help forum into a status game, which is what sustained 7 million answerers without cash payment.

Keeping the corpus searchable only inside Naver meant the flywheel's output was proprietary — a rival's better crawler had nothing to crawl.

When the content you need doesn't exist, build the loop that makes users produce it. Points and ranks can out-produce any crawling budget, and the corpus you own compounds.

Knowledge iN outlived every early-2000s rival portal and by 2019 had accumulated over 325 million answers. But the crowd's quality problems became Naver's reputation problem: uncorrected answers, joke posts and advertising pushed Korean users toward Google for reference-style queries through the 2010s. Naver's fix was hybridization — adding lawyer and doctor consultations and expert answers on top of the crowd corpus rather than abandoning it.