Load your audio file, pick a target peak level in dBFS, and press Normalize. The tool reports the original peak and the gain it applied, then you download the normalized WAV.
The tool finds the single loudest sample in your file and multiplies the whole recording by whatever factor puts that peak on your target level, usually -1 or -3 dBFS. The balance between loud and quiet parts is unchanged, only the overall level moves.
Leaving a small amount of headroom avoids inter-sample peaks that can distort when the audio is later encoded to MP3 or AAC. -1 dBFS is a safe default for most uses and -3 dBFS is safer still if the file will be re-encoded.
No. This is peak normalization, which targets the highest sample. Perceived loudness standards such as LUFS measure average loudness over time and require a different, much heavier analysis than a purely local browser tool performs.