We quickly form first impressions about newly encountered people guiding our subsequent behaviour (approach, avoidance). Such instant judgments might be innate and automatic, being performed unconsciously and independently to other cognitive processes. Lying detection might be subject to such a modular process. Unfortunately, numerous studies highlighted problems with lying detection paradigms such as high error rates and learning effects. Additionally, humans should be motivated doing both detecting others' lies and disguising own lies. Disguising own lies might even be more challenging than detecting other people's lies. Thus, when trying to disguise cheating behaviour, liars might display a mixture of disguising (fake) trust cues and uncontrolled lying cues making the interpretation of the expression difficult (perceivers are guessing). In two consecutive online studies, we tested whether seeing an increasing amount (range 0-4) of lying cues (LC) and non-lying cues (NLC) on a standard face results in enhanced guessing behaviour (studies 1 and 2) and that enhanced guessing is accompanied by slower responding (study 2). Results showed that pronounced guessing and slowest responding occurred for faces with an intermediate number and not with the highest number of LC and NLC. In particular, LC were more important than NLC to uncertain lying decisions. Thus, only a few LC may interfere with automatic processing of lying detection (irrespective of NLC), probably because too little lying cue information is yet available.