In complicated electromagnetic environments, synthetic aperture radar (SAR) can be threatened by various kinds of malicious interference, of which deceptive jamming is the intentional and efficient one. For jamming effect evaluation on SAR images, traditional methods are mostly based on the change of image quality, while not suitable for evaluating the confusion caused by high-fidelity false targets. In this paper, a novel framework to evaluate the effect of deceptive jamming on SAR is proposed based on visual consistency (VC). Three levels of vision, namely detection, recognition and semantics, are fused for efficient deception evaluation along with the corresponding metrics system. Fully considering the imaging characters of deceptive jamming, specifically designed detection and recognition flows are introduced to quantitatively evaluate the deception. Furthermore, to evaluate whether the generated false targets are with reasonable context, a unprecedent concept named semantic accuracy is proposed via considerations of statistical differences compared with that of the background template. Besides, the cases that deceptive jamming with several common non-ideal issues are considered when evaluating the effects. Sufficient experiments have proved the practicality and superiority of the proposed evaluation framework under different deceptive jamming with various non-ideal factors.