Look:

Here you see a Circle of 1.5 radius. So it barely touches the outside of the 3 by 3 grid.
Now, you are checking a point that is (1, 1)
from the center of the circle. Where is that? That is the center of one of the corner cells of the 3 by 3 grid. And yes, that point is inside the circle.
Notice that only the center cell is fully inside the circle.
You want some criteria that will include the side cells but not the corner cells (making a cross). Well, observe that:
- Most of the corner cell is outside.
- Only one of the corners of the corner cell is inside. But the side cells have two corners inside.
Checking corners of the cell is easier. You would check four points for the cell:
(0.5, 0.5)
from the center, which is inside (distance is 0.707107... < 0.5).
(1.5, 0.5)
from the center, which is outside (distance is 1.58114... > 0.5).
(0.5, 1.5)
from the center, which is outside (distance is 1.58114... > 0.5).
(1.5, 1.5)
from the center, which is outside (distance is 2.12132... > 0.5).
Checking how much of the cell is inside is harder, but also more interesting, because you could use it to shade the cells proportionally.
You can reach an approximation by checking an array of points inside of the cell. In fact, checking the four corners is a first approximation.
But we can do better by cutting some corners. Once you know which corners are inside, you also know which sides does the circle cut. Then by computing the intersection of the circle with the sides, you can get this approximation:

You would have to identify which case it is based on which corners are inside. The area inside the segments drawn should be simple to compute. And the missing area is a circular segment, so you can apply the formula for that.
I doubt you would need to go into figuring out how much of each pixel is inside the circle for your use case. Checking the corners should be sufficient.
You might also be interested in Midpoint circle algorithm, Marching squares.