> blog
AI-generated imagery vs stock photos vs real photography
We've mostly stopped using stock photography.
The three options
Real photography
Authentic, specific, builds trust. Always best for: people, products you sell, your physical premises, your team, your actual work. Worth the budget for hero images, About pages, and trust-critical content.
Stock photography
Generic, identical to other sites using the same source, often dated. Useful for: filler imagery where the specific subject doesn’t matter (abstract textures, generic backgrounds, illustrative concepts that don’t need uniqueness).
AI-generated imagery
Custom-generated to your brief, unique to your site, fast and cheap. Strong for: section illustrations, abstract concepts, branded imagery for blog posts, hero backgrounds where stock would feel generic. Weak for: anything specific (real people, real products, real locations).
Where AI imagery beats stock
Brand consistency. Stock photos come from different photographers in different lighting with different colour palettes. AI generates to a consistent brand brief: same lighting, same palette, same mood across 50 images.
Specificity. “Stock image of plumber working” gives you 200 generic options. “AI image of an Australian plumber in a domestic kitchen with the customer in the background, golden hour light” gives you exactly that, specific to your context.
Uniqueness. Other businesses can’t use the same image. You’re not on the same homepage as a competitor by accident.
Cost. Stock photo subscriptions: $20-50/month for limited downloads. AI image generation: pennies per image, generated on demand.
Where AI imagery falls short
Real people. AI faces still have artefacts: weird ears, miscounted fingers, eye asymmetries. For team pages, About pages, and anywhere visitors will scrutinise the image, use real photos.
Specific products. AI can’t generate accurate images of your actual products. Use real product photography.
Specific places. AI imagery of “Newcastle waterfront” will be a generic Australian waterfront, not the actual one. For location-specific content, use real photography.
Documentary content. Anything where the implication is “this is real,” AI is dishonest. Don’t fake it.
Our default approach
1. Use real photography for: hero images, About pages, team pages, products you sell, physical premises, completed work portfolios.
2. Use AI imagery for: section illustrations, blog post imagery, abstract concepts, branded backgrounds, anywhere generic stock would feel cheap.
3. Use stock photography rarely: only when AI can’t handle the brief and we don’t have budget for custom photography.
Most Jezweb-built sites end up 60% real photography, 35% AI-generated, 5% stock. The mix depends on the project.
The ethics question
AI image models were trained on existing imagery, much of it without explicit consent from photographers. This is genuinely contested ground. We follow the rules: licensed AI tools (Imagen, DALL-E, Midjourney with commercial licence) for client work, never imitate specific artists’ styles, never generate content meant to deceive.
Where ethics matter most: anything implying authenticity (people, places, events). Where ethics matter less: abstract illustration, decoration, mood imagery.