Pixelate, blur or solid box: which one to use
Pixelate turns the area into large squares, each one the average colour of the pixels under it; it shows clearly that something was hidden and suits faces and number plates in photos you post. Blur smooths the area into a soft haze and looks gentler, which suits a background, a crowd or a screen behind you. The solid box paints the area black and keeps nothing of what was there, so it is the right choice for text and numbers: ID cards, passports, bank cards, IBANs, boarding passes, addresses and phone numbers. The strength slider sets the size of the squares or the amount of blur, measured against the size of your photo, so a setting looks the same on a small screenshot and a 48-megapixel camera picture.
Why a light blur is not always enough
Researchers showed in 2016 that a computer trained on blurred or pixelated pictures can still recognise faces and read handwritten digits that people cannot make out, and a public proof-of-concept tool, Depix, recovers text from pixelated screenshots. Text is the easiest to recover, because the letters come from a known font. This tool makes recovery harder in three ways: squares are never smaller than 8 pixels of the full image, areas painted with the brush are widened to whole squares, and blur works on an averaged, smaller copy of the area with a little random noise added. Even so, only the solid box guarantees that nothing can be read back. Use it for any number or word that must stay private.
Finding faces automatically
Some browsers include a face detector that comes with the device, described in the Shape Detection API draft. It is not yet a web standard, and many browsers do not offer it. When your browser has it, a "Find faces" button appears: it marks every face it finds with a box using your current effect, and you can remove or adjust each one. When it is missing, the page says so and you mark faces by hand; nothing extra is downloaded and no photo is sent anywhere to look for faces. Always check the result yourself, since a detector can miss faces that are small, turned away or partly covered.
Full size kept, hidden data removed
The saved copy has exactly the width and height of your photo; only the areas you marked change. JPG photos are saved as JPG, PNG as PNG with any transparent background kept, and WebP as WebP where the browser can write it. Phone photos are turned the right way up first. Because the new file is drawn from the pixels, the camera data stored in the original, such as the GPS location, the date and the phone model, is not copied into it. To see what a photo reveals before you share it, open it in the EXIF viewer. HEIC photos from an iPhone open only in browsers that can read them; otherwise convert them to JPG first. Photos up to 50 megapixels work in most desktop browsers; phones may refuse very large images.