A before-and-after image is evidence. It tells the reader what a specific edit changed. If lighting, angle, subject, text, and crop all change at once, the comparison cannot show the effect of the taught step. This is especially important when an AI image tool can make an entirely new scene look like a simple edit.
Begin with a source you can document
Use a screenshot, photograph, or design file you have the right to edit and publish. Keep the original unaltered. Make a copy for the tutorial step, and record the tool and operation used. If the post teaches a background-removal workflow, the “before” is the original image and the “after” is that same image with the background removed. Do not swap the subject for a better generated one and call it a removal result.
Choose the comparison format before editing. A side-by-side panel works when the changed area is broad; a slider can help with local changes but must remain usable on a phone and by keyboard. For a small detail, add a matched zoomed inset. The insets should use the same scale and position.
Hold the conditions steady
| Keep consistent | Why | Exception to label |
|---|---|---|
| Canvas size and crop | The reader should compare the edit, not a different composition. | If cropping is the lesson, show the original frame and the crop boundary. |
| Zoom and viewing scale | A larger “after” can look sharper without any real improvement. | Show a separate matched detail view when necessary. |
| Source content | Changing the subject, copy, or lighting adds a second variable. | List every change if the workflow intentionally transforms several elements. |
| Export quality | Different compression or resolution can create a false visual difference. | Explain any export setting that is itself being taught. |
Use AI for the step the article actually teaches
An AI-assisted cleanup may be the subject of the tutorial. In that case, show the real input and the real output. If the model changes more than requested—such as altering a product edge while removing the background—point it out or reject that result. A clean-looking “after” is not useful if it misrepresents what the tool did to the original.
Tools such as Photoshop and Canva can be used to assemble the comparison, but the editor is less important than preserving the source and labeling the operation. If an entire image was generated from a prompt, present it as a concept illustration, not as the “after” of an existing photograph.
Label the result without overclaiming
Put a short label on each panel: “Original” and “After background removal,” for example. A caption can add the tool, the date or version if it matters, and the exact edit. If you manually repaired edges after the AI pass, say so. If the article claims a measurable improvement, show the measurement and method separately; the image alone cannot prove speed, conversion, or accessibility.
Write alt text for the comparison’s meaning, not just “before and after.” If the visible caption already explains the full change, avoid repeating every word. The W3C images tutorial explains how text alternatives depend on what the image contributes. A slider should also expose both states to people who cannot drag it.
A final check before publishing
- Can you identify the original file and reproduce the edit?
- Are crop, scale, and export settings matched, or is the difference disclosed?
- Did AI alter an element outside the taught step?
- Can a reader understand the change at mobile size and without relying on the image alone?
If the image fails one of these checks, fix the comparison or describe the step in text until an honest visual is available. A smaller, accurate example teaches more than a dramatic result whose source cannot be accounted for.
