<p>Whether highly novel inventions experience delayed recognition in examination, often labeled as “bias against novelty”, remains underexplored. Guided by confirmation bias theory, we test whether highly novel inventions face longer time-lags using 530,525 invention patents granted to China’s listed firms (2000–2022). We measure novelty with the disruption (<i>D</i>) index and define grant lag as months from the substantive examination effective date to the grant date. We document a U-shaped relationship between novelty and grant lag with a turning point around 0.325—beyond which grant lag increases with novelty, consistent with intensified delayed recognition. Two countervailing mechanisms account for the pattern: a retrieval-convenience effect shortens lag at low-to-moderate novelty, whereas an interpretive-burden effect lengthens lag beyond the threshold; the latter predominates in driving delayed recognition at high novelty. We further show that larger inventor teams, greater interdisciplinarity, broader examiner experience, and patent-agent representation mitigate the right-tail increase in grant lag. Although prioritized examination and automated prior art recommendation reduce overall delays, they do not fully attenuate delayed recognition for highly novel inventions. Results are robust across alternative applicant samples, alternative novelty measures, an EPO-based grant-lag definition, survival models incorporating four examination outcomes, and HDFE/2SLS models. This study reveals a challenge for balancing rigor and timeliness in examining highly novel inventions and informs organizational and policy levers to better support radical innovation.</p>