The $dx$ term intuitively stands for a really small (infinitesimally small) change in the $x$ variable.
Remember how we define integration in the first place. $\int \limits_{a}^{b} f(x) \,dx$ is defined by first taking Riemann sums that look like $\sum f(x^{*})(x_{i+1} - x_{i})$, where $x^{*}$ is a point in $[x_{i}, x_{i+1}]$. If you remember from Calculus, this sum represents the sum of the areas of rectangles that all together give us an approximation to the area under the curve $f(x)$. In what sense are these Riemann sums an approximation of the area? In the sense that, as the base of the rectangles gets smaller and smaller (so that the rectangles get thinner and thinner), you get a better and better approximation.
When you take the limit as the base of the rectangles goes to $0$, then intuitively $(x_{i+1} - x_{i})$ goes to an infinitesimally small value, so that $\sum f(x^{*})(x_{i+1}-x_{i})$ becomes, intuitively, $\int f(x)\,dx$, where the $\int$ symbol means we are "adding" up the now infinitely many rectangles, and $f(x)\,dx$ is intuitively the area of each rectangle. So, to me, from this understanding of the limit, I always interpret the integral my mind as meaning $\underbrace{\int}_{\text{adding}} \underbrace{f(x)\,dx}_{\substack{\text{area of rectangle} \\ \text{w/ height }f(x)\text{ and base } dx}}$. Of course, this is just intuition, because $\int f(x)\,dx$ is just notation for the limit. $dx$ doesn't mean anything on its own, at least not at the level of Calculus, though as someone else pointed out, there is a way to make this intuitive interpretation rigorous. It involves defining infinitesimals (search the topic of Nonstandard Analysis if you are interested in this).