Recent studies have shown that, when comparing multidigit numbers that differ in length (e.g., 2384-107), a decision is made considering length but also other attributes like the value of the initial digits (i.e., left-most digit/length congruity effect: faster responses to 2384-107 than to 2675-398). A non-solved issue is whether participants choose the number with more digits by exactly computing the number of digits in the string (e.g., 3 vs. 4) or whether they simply choose the perceptually larger item. In our first study participants were presented with pairs of different length numbers (3 vs. 4-digits) and were requested to decide which multidigit starts with a larger digit. Results showed more difficulties when the smaller digit was in the 4-digit length number, suggesting that length was automatically processed even although it was irrelevant for the task. In a second study, we presented participants with pairs of 3- and 4-digit-length numbers but obscured the processing of perceptual length by including a letter at the end of the 3-digit-length numbers (e.g., 8567-342M). Additionally, we manipulated the left-most-digit/length congruity effect and presented one string in each pair in a larger font than the other. Then we requested participants to do a physical size decision task. Together with an influence of the leftmost digit in the string, no effects of digit-length were observed, thus suggesting that the exact number of digits in the string is not processed automatically. It seems participants rely on perceptual information when comparing multidigit numbers that differ in length.