Organic toxins are persistent chemicals of global concern capable of accumulating in environment and food. Surface enhanced Raman spectroscopy (SERS) is a promising technique that facilitates onsite detection of organic toxins. However, the fabrication of a SERS substrate is complicated and difficult to provide flexibility, fastness and cost-effectiveness. This study aims to develop a paper-based SERS method using grape skin-gold nanoparticles/graphene oxide (GE-AuNPs/GO) as SERS substrate and evaluate its efficiency with rhodamine 6G (Rh6G) as a model organic toxin and a real water and food contaminant. GE-AuNPs synthesized by green method using grape skin waste extract and GE-AuNPs/GO showed a surface plasmon resonance at 536 and 539 nm, particle size 18.6 and 19.5 nm, and zeta potential −44.6 and −59.7 mV, respectively. Paper-based SERS substrates were prepared by coating a hydrophobic thin-film of 30% polydimethylsiloxane solution in hexane on Whatman no. 1 filter paper, followed by drop-casting GE-AuNPs or GE-AuNPs/GO and drying. The SERS signals of Rh6G showed an enhancement factor of 5.8 × 10⁴ for GE-AuNPs and 1.92 × 10⁹ for GE-AuNPs/GO, implying that a combination of electromagnetic surface plasmon, charge transfer and molecular resonances may be responsible for a higher enhancement of signal by the latter. A low detection limit of 7.33 × 10⁻¹¹ M in the linear range of 10⁻¹¹–10⁻⁵ M was obtained for GE-AuNPs/GO, while the relative standard deviation of repeatability and reproducibility was 9.6 and 12.6%, respectively. Paper-based GE-AuNPs/GO SERS substrate was highly stable as <20% loss in efficiency was shown over a 60-day storage period. Application to real samples showed a high recovery of Rh6G from tap water (93.9–100.8%) as well as food samples such as red chilli powder (91.0–95.4%), red glutinous rice ball (96.6–98.3%) and tomato ketchup (98.9–102.3%) after QuEChERS extraction. Collectively, the developed paper-based GE-AuNPs/GO can be a potential substrate for sensitive onsite detection of rhodamine 6G by SERS method.