Despite challenges, the COVID stay-at-home has provided opportunities to reveal unforeseen complexities within built environments, particularly from indoor health lenses, that had not been sought before the pandemic. This research aimed to evaluate occupants’ feedback on impacts of the stay-at-home on the indoor air quality (IAQ) perception in buildings nationwide in the U.S. during the first year of the pandemic (2019) and compare it with the baseline (2019). We used geo-tagged big textual data obtained from Twitter platform and developed Natural Language Processing (NLP) models based on Sentiment Analysis (SA) approach to compute the occupant feedback for the two consecutive years. We built the SA models through developing a six-step workflow, including data acquisition; data cleaning; text tokenization; analysis; accuracy evaluation; and data visualization and mapping. We used the QDAP dictionary and nrc lexicon to develop the SA models. We also developed automation scripts to improve the simulation performance. Results illustrate that occupants’ complaints on IAQ increased during 2020 compared with the baseline (2019). Findings further suggest that occupants with less access to operative Heating, Cooling, and Air Conditioning (HVAC) systems posted more dissatisfied feelings on Twitter. This research aids decision-makers to better understand the impacts of lockdown on occupants’ health experience to rethink the design and operation of buildings in a way to promote human-centered design and improving building resilience against future pandemics.