Step 1: State the buffer as a distance-based region.
Mathematically, the buffer of an input feature $F$ at distance $d$ is the region $B = \{ p : \text{dist}(p, F) \le d \}$, the collection of every point $p$ in the plane whose shortest distance to $F$ is not more than $d$. This is fundamentally a two dimensional region, since it is defined by a distance inequality, not by a single distance value.
Step 2: Apply the formula to a point.
If $F$ is a single point, then $\text{dist}(p, F) \le d$ describes every point $p$ lying inside or on a circle of radius $d$ centered at $F$. This filled circular region is exactly a disk, which GIS software represents and stores as a polygon.
Step 3: Apply the formula to a line.
If $F$ is a line segment or polyline, then $\text{dist}(p, F) \le d$ describes a stadium-shaped region, a rectangular strip of width $2d$ running along the line, with semicircular end caps of radius $d$ at the two extreme endpoints. This filled strip is again a closed, bounded two dimensional area, stored as a polygon, not as a line or polyline.
Step 4: Note why the output type cannot change with the input type.
The reason both cases give a polygon is that a buffer is not a copy or extension of the input feature's own geometry type, it is an entirely new region built from a distance condition, and any region satisfying "distance less than or equal to $d$" is inherently area-like. This is why GIS software always classifies the buffer output as a polygon feature class, regardless of whether the input feature class was points, lines, or even polygons.
Step 5: Conclusion.
Buffering a point or a line always produces an area-type feature, that is, a polygon.
\[ \boxed{\text{Polygon}} \]