In order to achieve the pre-set funding goal, some entrepreneurs may engage in malicious fraud, that is, using fraudulent textual descriptions to attract monetary contribution from crowd. Thus, fraud is inevitable in the online financial market. Fraudulent texts are not strictly equivalent to low-quality campaigns, but fraudulent content can jeopardize users’ perceptions of project quality. Thus, the fraudulent text has great drawbacks for the development of crowdfunding model, leading investors lose confidence in this newborn financing model. Through text mining, 4 indicators are adopted to measure the linguistic feature related to fraud. And 126,593 campaigns from Kickstarter is employed to estimate the impact of linguistic feature related to fraud on the fundraising outcomes. Multi text levels are selected as the study objects include abstract, detailed description and the reward narratives. The results show that in general, lower linguistic feature related to fraud attracts the investors to contribute more money, the predictive model also validates this conclusion. However, some fraud indicators have no significant negative impacts on financing, or even show positive influences. Moreover, the detailed delivery terms in the reward text, the higher ratio of successful funding. This study provides a guideline for the founders to generate attractive description for the crowdfunding campaigns.