Annotate Screenshots With Arrows and Text
How to point at things in screenshots effectively — arrows, boxes, circles, numbered steps and labels — using a free browser-based image annotator.
A screenshot with a clear annotation answers a question in one glance. The same screenshot with five overlapping arrows and three colors of highlight creates a new question: "what am I supposed to look at?" Annotation is a small skill, and the rules below cover almost every case — bug reports, how-to guides, reviews, and feedback on designs.
One annotation, one message
The single biggest mistake is annotating everything at once. Each screenshot should answer one question: where is the setting, what is broken, what changes in this step. If a screenshot needs three things explained, either the explanation belongs in text below the image, or the screenshot needs cropping into three images.
The editor's annotation tools are direct on the canvas — text, arrow, rectangle, circle — and each element stays individually editable: drag to move, pull a handle to resize, rotate, set the size and color per element. Nothing is flattened until you export.
Arrows: point from empty space
An arrow works when its tail starts in a quiet area and its head lands next to the target, not on top of it. Covering the target with the arrowhead defeats the purpose — the reader should see the arrow and the thing it points to simultaneously.
- Start arrows from a margin or empty background, never from another annotation.
- Keep arrows at consistent angles within one image — a fan of three arrows from one corner reads better than three crossing diagonals.
- One color for all arrows. Orange on a light screenshot, or white on a dark one, stays readable on both.
Boxes and circles: for regions, not points
A rectangle says "this area", a circle says "this thing". Use a rectangle when the boundary matters — a UI element, a row of text, a setting. Use a circle sparingly: on busy screenshots a circle reads as noise, and an ellipse drawn hastily around text looks unpolished.
Keep the outline 2-3px, one color, and prefer the rectangle unless the target is genuinely round. Fill nothing: a translucent fill hides the pixels underneath, which is the opposite of what annotation is for.
Text: number the steps
For guides, numbered labels are the most durable annotation of all. A "1", "2", "3" next to each step survives cropping, re-sharing, and embedding — whereas a paragraph below the image does not travel with it.
- Put step numbers inside the image, next to the region they refer to.
- Keep label font size proportional to the screenshot: roughly 1/20th of the image height is readable at full size and still legible when the image is embedded smaller.
- If a label must sit over the screenshot, give it a small solid backing (the editor's text tool supports per-element color) rather than raw text over busy pixels.
The classic combinations
Three recipes cover most requests:
- Bug report: one rectangle around the broken element, one arrow from a margin, one line of text naming the expected behavior.
- How-to step: cropped screenshot of the relevant screen, one numbered label, one arrow if the target is small.
- Review or comparison: the annotated screenshot plus the side-by-side comparison layout — annotations explain the difference, the layout frames it.
Building it in the editor
- Open the collage editor and drop the screenshot onto a single-cell layout (or the template matching the final size — a blog header, a thumbnail).
- Add the arrow first, then the box or circle, then the text — in that order, because text sits at the visual hierarchy's top and is easiest to place last.
- Adjust each element directly on the canvas; the undo history lets you walk back through annotation decisions.
- Export as lossless PNG — annotation over screenshots is text and line work, and JPG compression smudges exactly those pixels.
If the explanation spans several screenshots in sequence, that is the stitching job: join them into one long image and annotate each step with its number as you go.
The squint test
Zoom the finished image to 25% and squint: the annotation should be the first thing you see, and its meaning should still be guessable. If the annotation disappears at 25%, it is too thin or too small; if the meaning is unclear, it is trying to say two things at once. Split it and try again.