Culturally shared images portraying the poor are important resources for legalizing income inequality. According to Stereotype Content Model (SCM), competence and warmth are the two fundamental dimensions of stereotype. Low-status people are often seen as incompetent, and threatening individuals are often perceived as cold. The poor have lower income, less education, and worse jobs or no jobs, and they are often perceived as status threats by others. Some studies have shown that the poor are often considered low-competence but high warmth. At first glance, the high warmth evaluation, for the poor, has compensated the low competence evaluation. However, poor people’s warmth score is only at the middle or low level, and far lower than that of the middle-class. Based on the SCM and existing research results, hypothesis is made: (H1) the stereotype towards the poor is likely to be low-competence and low-warmth. The source-monitoring framework suggests that general knowledge or beliefs about sources can lead to biased guessing. Empirical studies also show that stereotypes are an important factor influencing guessing bias in source monitoring tasks: When the source memory is blurred, participants tend to guess that kitchenware was in the kitchen, and toiletries were in the bathroom; medical advices came from doctors, and legal advices came from lawyers; “worry about their attractiveness” came from young people, and “worry about their health” came from older people. Given people’s negative stereotype towards the poor and the effect of stereotype on guessing bias, it can be inferred: (H2) when the source is forgotten, participants are likely show negative guessing bias towards the poor. Two experiments were carried out to test the two hypotheses. College students participated these experiments. Experiment 1 employed two single category-IATs (SC-IAT) to explore participants’ implicit stereotypes towards the poor. SC-IAT is very useful, as it allows for testing the automatic association between “the poor” and “competence” (Exp1a) and the automatic association between “the poor” and “warmth” (Exp1b) without having a complementary target category (such as“the rich”). Experiment 2 required participants to complete the source monitoring task and estimated the guessing bias towards the poor via the two-high threshold source memory model (2HTSM). For the source monitoring task, participants, during the learning phase, were asked to perform a two-choice judgment for individually presented trait words (90 words) paired with their sources (the poor or well-off people). During the testing phase, 180 words were randomly presented on the computer screen. Participants were instructed to indicate whether the word had been presented. When a word was judged as an “old” word, participants then need to indicate its source. In experiment 1, when the words representing the poor and the words representing low-competence (Exp1a) and low-warmth (Exp1b) were categorized on one response key, participants responded faster; the D score in Exp1a was larger than that in Exp1b, suggesting that the poor are stereotyped as low-competence and low-warmth, and their competence is perceived as more negative. In experiment 2, when trait-words were detected as “old”, but the source was forgotten, participants tended to guess that negative trait words came from the poor, but positive words came from well-off people, suggesting that participants have negative guessing bias towards the poor. Taken together, results from the two experiments provide evidences supporting our hypotheses: participants not only hold negative stereotype, but also show a negative guessing bias towards the poor. These may further exacerbate income inequality. Being treated unfairly, the poor faces severe difficulties in the process of poverty alleviation. Therefore, future research needs to explore effective methods to decrease or eliminate the negative stereotypes and guessing bias towards the poor.