Fungal toxins are secondary metabolites of fungi. Food is highly susceptible to contamination by various fungal species that produce fungal toxins during production and storage. Fungal toxins can cause either acute or chronic poisoning from long-term, low-dose ingestion. Therefore, fungal toxins have become a topic of international interest as a food safety issue. Deoxynivalenol (DON) is a single-terminal sporam toxin produced predominantly by Fusarium graminae and Fusarium pinkosa. DON is globally one of the most common fungal toxins contaminating grain, food, and feed. Various methods have been applied for screening and detecting DON; however, these methods utilize expensive instruments and entail complex operations, poor repeatability, and low sensitivity. Therefore, the development of a simpler, more rapid, and sensitive sensing technology for DON detection is important for applications within the agriculture and food industry. Recently, surface-enhanced Raman scattering (SERS) has become a rapidly developing spectral analysis technology with unique advantages, including high sensitivity, high throughput, and rapid response rates. Therefore, attempts have been made to apply the SERS technique to detecting DON. However, due to the limitations concerning SERS substrates, the currently established SERS method exhibits serious problems, including low sensitivity and weak anti-interference ability, and cannot meet the requirements of sample detection. Recently, our group has prepared aggregated silver nanoparticles (a-AgNPs/CDs) with high SERS activity by using single-layer carbon-based dots (CDs) as a capping agent. Moreover, the obtained materials (a-AgNPs/CDs) were combined with hydrogel technology to prepare novel hydrogel SERS chips. The obtained SERS chips exhibited several advantages over traditional SERS substrates, such as high sensitivity, long-term stability, improved uniformity, and strong anti-interference capabilities. Herein, a novel SERS method for rapid screening and detection of DON in grains was established using a portable Raman spectrometer based on the developed hydrogel SERS chips. The main experimental conditions were optimized before the SERS detection of DON; this included the optimization of the hydrogel SERS chip soaking temperature and time in the DON solution. It was found that the optimal soaking temperature and time were 40 ℃ and 5 min, respectively. Under the optimal SERS detection conditions, the linear response range of DON was 1-10000 μg/kg (correlation coefficient (R2)=0.9967), and the limit of detection (LOD) was 0.14 μg/kg. Due to the unique pore size structure of the hydrogel, common sugar, protein, oil, pigment, and other interfering substances in the sample matrix were blocked outside the hydrogel. Therefore, only simple extraction was required while detecting complex samples. This method was applied to detecting DON in wheat flour, yielding recoveries of 97.3%-103% with relative standard deviations of 4.2%-5.0%. The established SERS method for DON detection exhibits a broader response range, high sensitivity, good repeatability, rapid response, simple operation, and strong anti-interference capability. This shows that the laboratory-constructed hydrogel SERS chip has excellent potential for rapid screening and detection of biotoxins in food.