Most people spend all their prompting energy on what they want in an image.
The professionals also spend time on what they do not want.
A negative prompt is an instruction that tells an AI image generator what to exclude from the output. Instead of describing what should be in the image you are describing what should not be. Used correctly negative prompts solve specific recurring problems that positive prompt language cannot fix on its own.
Used correctly negative prompts solve specific recurring problems that positive prompt language cannot fix on its own.
This guide explains exactly how negative prompts work, when they are worth using, and gives you 30 ready-to-use phrases you can copy into your prompts today.
How negative prompts work
Different AI tools handle negative prompts differently. Understanding the mechanics of your specific tool determines how to write them effectively.
In Midjourney, negative prompts are added using the --no parameter. You list the things you want excluded after --no at the end of your prompt.
Basic structure: [your prompt] --no [things to exclude]
Example: 'Product still life of a skincare bottle on marble, soft window light --no text, logos, watermarks, people, hands'
Multiple exclusions are separated by commas. There is no limit to how many things you can exclude but practical experience shows that 3 to 8 specific exclusions work better than long lists of dozens of items.
In ChatGPT image generation, negative prompts are written conversationally into the main prompt using phrases like 'without any text', 'no people in the scene', 'avoid showing hands', 'exclude any logos or branding'. ChatGPT does not use a separate parameter - the exclusions are part of the natural language description.
In Stable Diffusion, negative prompts have a dedicated text field separate from the positive prompt. This is where the negative prompt concept originated and where it is most powerful and precise.
When negative prompts actually help
Negative prompts are not magic. They do not fix bad positive prompts. They do not substitute for specific descriptive language. They work best for a specific set of problems.
Use negative prompts when:
You keep getting an unwanted element that your positive prompt is not preventing. You have run the prompt multiple times and the same problem keeps appearing. A specific exclusion instruction is more direct than trying to phrase your positive prompt around the problem.
You want to prevent common AI defaults. AI image generators have default tendencies - they add watermarks to some content, place people in scenes when not asked, generate text that nobody requested, add logos or branding elements automatically. A standing negative prompt prevents these defaults.
You are working in a specific content category with known problems. Product photography almost always benefits from excluding hands and people unless specifically requested. Portrait work benefits from excluding multiple subjects. Brand content benefits from excluding any text or logos that might appear randomly.
Do not use negative prompts when:
Your positive prompt is vague and you are hoping negative prompts will compensate. If your base prompt is weak add more specificity to the positive description first. Negative prompts cannot save a vague positive prompt.
You are excluding things that are not actually appearing in your outputs. Only exclude what is actually causing problems. Long lists of preemptive exclusions can confuse the AI and produce unexpected results.
The most useful negative prompt categories
Unwanted people and body parts
AI image generators frequently add people, hands, and body parts to scenes even when they are not described in the prompt. This is one of the most common negative prompt use cases.
--no people, hands, fingers, body parts, human figures, faces, crowds
Use this for: product photography where a hand keeps appearing, landscape and interior scenes where people appear uninvited, any content where a human-free scene is required.
Text and typography
Random text appearing in AI images is extremely common and almost always unwanted in brand photography. Signs, labels, words, and symbols appear spontaneously in backgrounds, on surfaces, and on products.
Random text appearing in AI images is extremely common and almost always unwanted in brand photography.
--no text, words, letters, numbers, typography, signs, labels, watermarks, captions, subtitles
Use this for: lifestyle scenes with background environments, product photography where labels should be blank or minimal, any image where accidental text would look unprofessional.
Logos and branding
AI generators sometimes add brand logos, trademark symbols, and corporate branding elements to scenes unprompted - particularly in tech, retail, and lifestyle contexts.
--no logos, brand names, trademarks, copyright symbols, registered marks, corporate branding
Use this for: competitive brand photography where competitor logos might appear, any professional brand content where accidental logos would be problematic.
AI artifacts and quality issues
Certain visual problems appear consistently in AI-generated imagery - extra fingers, distorted faces, impossible anatomy, blurry areas, and other telltale signs of AI generation.
--no extra fingers, distorted hands, blurry faces, double faces, morphed features, anatomical errors, skin artifacts, noise, grain, pixelation, low quality, blurry, out of focus
Use this for: portrait and beauty work where anatomical accuracy matters, any content where AI-looking artifacts would undermine professional quality.
Unwanted styles and aesthetics
If your outputs keep defaulting to an aesthetic you do not want, naming it in the negative prompt can pull the generation away from that direction.
--no cartoon, illustration, anime, painting, sketch, drawing, digital art, 3D render, CGI
Use this for: when you want photorealistic results but the AI keeps producing illustrated or artistic outputs. When your stylize value is high but you still need photographic quality.
Composition problems
Specific composition elements that keep appearing unwanted can be excluded directly.
--no borders, frames, vignette, letterbox, multiple panels, collage, split screen, watermark, photo frame
Use this for: when outputs keep appearing with decorative borders or frames, when multiple image panels appear instead of a single image.
Overused AI aesthetics
Certain aesthetic qualities have become so associated with AI generation that they immediately read as artificial to trained eyes.
--no oversaturated colors, HDR, lens flare, oversharpened, plastic skin, perfect symmetry, too smooth, artificial bokeh, stock photo feel
Use this for: when your outputs look obviously AI-generated and you want more natural photographic quality. Often combined with --raw for maximum photographic realism.
'--no text, watermarks, logos, extra hands, people (when not requested), blurry areas, anatomical errors, oversaturated colors, stock photo feel, artificial bokeh, borders, frames'
Add this to the end of any brand photography prompt as a starting point. Remove exclusions that are not relevant to your specific content type and add specific exclusions for problems you encounter in your outputs.
Negative prompts vs better positive prompts
There is an ongoing debate in the AI prompt community about whether negative prompts are better than improving positive prompts.
The honest answer is that they solve different problems.
A better positive prompt prevents problems by being specific enough that the AI has no room to add unwanted elements. When you describe exactly what surfaces, lighting, and composition you want there is less space for random additions.
A negative prompt is a direct exclusion instruction for things that keep appearing despite specific positive prompting. It is not a substitute for good positive prompting - it is an additional layer of control.
A negative prompt is a direct exclusion instruction for things that keep appearing despite specific positive prompting. It is not a substitute for good positive prompting - it is an additional layer of control.
Use both. Start with the most specific positive prompt you can write. Then add targeted negative prompts for specific recurring problems. This combination produces the most consistent and controllable results.
Fix prompts that need negative language
If your current prompts keep producing the same unwanted elements our free Prompt Fixer can help. Paste your prompt and describe what is going wrong - the tool diagnoses the problem layer by layer and rewrites the prompt with specific positive improvements plus targeted negative prompt additions where appropriate.
CropUp Ai