If you use the chassis of the car, that is very problematic.

If you had two parallel electrodes of the identical metal type and a fix length exposed to the brake fluid, then readings would be meaningful, if not precise.

Let's say you made such a test probe and calibrated it with fresh brake fluid that is mixed with various percentages of distilled water. Now you could put this probe into a random brake fluid reservoir and get a repeatable result. The resistance reading would depend on the electrode separation as well as the uninsulated length.

It would probably provide an inaccurate reading due to soluble metals, salts and minerals contaminating the brake fluid. However that would only cause errors on the side of caution. If it gave a high enough resistance reading, that would indicate the reservoir was dry.

The next issue to worry about is: how fast does moisture migrate from wheel cylinders back to the master cylinder reservoir? I don't know. However, you could flush a small quantity from each bleeder valve and test that in some very small sample container (or directly in-line with the bleeder hose). Flush until the reading is good enough. This would be an excellent way to conserve on brake fluid, if that was a significant business expense.