Every agency has had this conversation with a client: “We need images for the website.” What follows is usually a negotiation between budget, timeline, and quality — pick two.
Custom photography is expensive. Stock subscriptions add up. And the client’s existing image library is four blurry photos from a 2019 company picnic.
On a recent project, we tried a different approach. Instead of the usual stock-plus-photographer combo, we used AI image generation for the majority of the site’s visual content. The results — both in quality and cost savings — changed how we think about image sourcing for client work.
The Project
The client was a mid-sized financial advisory firm launching a new website. Content-heavy: 40+ pages including service descriptions, team bios, a resource center with 25 articles at launch, and a blog that would publish weekly.
Their visual needs broke down roughly like this:
- 15 hero/banner images for main pages
- 25 featured images for resource center articles
- 12 team member headshots
- 20+ supporting images throughout service pages
- Social sharing thumbnails for all content
In a traditional workflow, we’d budget for a half-day photo shoot ($2,000-3,000 for the photographer alone), supplement with stock photography ($50-100/month subscription), and spend significant design hours sourcing, editing, and formatting everything.
Total estimated image budget using our standard approach: $8,200 across photography, stock licensing, and design hours.
The Alternative Approach
We proposed using AI Photo Generator, a realistic AI image generator, as the primary image source for the project. The client was skeptical — understandably. “AI images” still conjured thoughts of weird artifacts and uncanny valley faces for them.
So we ran a blind test. We generated five images, mixed them with five stock photos, and asked the client’s team to identify which were which. They got six out of ten right — barely better than guessing.
That sold them on trying it.
What We Used AI Generation For
- Hero images: Scenes of professional settings, office environments, client meetings — all generated to match the firm’s brand colors and aesthetic
- Article featured images: Specific to each topic. An article about retirement planning got a tailored image, not a generic “money” stock photo
- Supporting imagery: Lifestyle and environmental shots woven throughout service pages
- Social thumbnails: Purpose-built for each platform’s dimensions
What We Still Photographed
- Team headshots: We used AI to generate professional variations from casual photos the team submitted, but two senior partners insisted on traditional photography. Fair enough — their photos appear on regulatory filings
- Office exterior: One shot of the actual building for the contact page. AI can’t photograph a specific real location
The Workflow We Built
For this to work at agency scale, we needed a repeatable process — not one person experimenting with prompts for an hour per image.
Step 1: Visual Brief Per Page
During the design phase, our designer wrote a one-line image brief for each placement. Not a full prompt yet — just the intent.
Examples from the actual project:
- Homepage hero: “Advisor and client shaking hands in bright modern office”
- Retirement planning page: “Couple in their 60s reviewing documents together at home, relaxed”
- Market commentary article: “City skyline at dawn, financial district, optimistic tone”
Step 2: Prompt Generation
We translated each brief into a detailed prompt. This is where specificity matters. We learned quickly that prompts need to include:
- Subject and action (who’s doing what)
- Setting (where, what kind of space)
- Lighting (natural, studio, warm, cool)
- Mood (professional, relaxed, energetic)
- Composition notes (headroom for text overlay, left-weighted for sidebar layouts)
That last point is important for web work. Stock photos rarely account for where text will sit. With AI generation, you can prompt for negative space exactly where your design needs it.
Step 3: Generate, Review, Optimize
We’d generate 3-4 variations per brief, pick the strongest one, and run it through our standard optimization pipeline — resize, compress, convert to WebP, generate srcset variants.
Average time per final image: 8 minutes from prompt to optimized file.
For comparison, our average time sourcing a stock photo (searching, previewing, licensing, downloading, editing to fit) was around 20 minutes.
Step 4: Client Review
We presented all images in context — placed in the actual design, not as standalone files. This is key. An AI-generated image in isolation invites pixel-peeping. The same image in a finished page layout, with typography and spacing, just looks like a website.
The client approved 90% of images in the first round. The few rejections were about content preference (“can we make the couple look older?”), not quality.
The Numbers
Here’s the cost comparison for the full project:
Traditional Approach (Estimated)
| Item | Cost |
| Half-day photo shoot (photographer + assistant) | $2,800 |
| Photo editing/retouching (12 hours @ $85) | $1,020 |
| Stock subscription (3 months of project) | $150 |
| Stock individual licenses (premium images) | $280 |
| Design hours for sourcing/formatting (30 hours @ $110) | $3,300 |
| Total | $7,550 |
AI Generation Approach (Actual)
| Item | Cost |
| AI generation credits | $120 |
| Partner headshot photography (2 people) | $650 |
| Design hours for prompting/review/optimization (12 hours @ $110) | $1,320 |
| Total | $2,090 |
Savings: $5,460 (72%)
Even if we’re conservative and call it 60% savings (accounting for some estimates in the traditional approach), that’s a significant number — especially on a project where the total web build budget was $35,000. We saved the client roughly 15% of their entire project cost on images alone.
What We Passed On to the Client
We didn’t pocket the savings. We reduced the image line item in the proposal, which helped us win the project in a competitive bid against two other agencies. The client got a better price, we got the work, and the final product was — honestly — better than what stock photos would have delivered.
The images were more cohesive because they were all generated with the same aesthetic guidelines. They were more relevant because each one was created for its specific context. And they were completely unique — no risk of a competitor’s site showing the same hero image.
Lessons for Other Agencies
After running this on three more client projects since, here’s what we’ve learned:
Build a Prompt Library
Don’t start from scratch every time. We maintain a shared doc of prompt patterns organized by industry and use case. “Professional headshot — male, 40s, business casual” has a base prompt we tweak per project. This cuts generation time significantly.
Set Client Expectations Early
Mention AI image generation in the proposal phase. Some clients love it (cost savings, uniqueness). A few have concerns (usually about authenticity). Better to have that conversation upfront than surprise them in the design review.
AI Images Are Best for Illustrative Content
Where AI generation shines: lifestyle imagery, environmental scenes, conceptual visuals, portraits from reference photos. Where it doesn’t: specific real products, specific real locations, legally sensitive contexts where “this is a real photo” matters (some healthcare, legal, real estate). Know the boundaries.
Quality Control Is Non-Negotiable
Every image gets reviewed before it goes into a design. Check for artifacts, check that text is readable if present, check that faces look natural. We caught maybe 5% of generations with subtle issues that needed a re-generate. Quick to fix, but you have to look.
Pair with Real Photography When It Matters
AI generation doesn’t have to be all-or-nothing. Our best results come from combining AI-generated environmental and lifestyle imagery with real photography for key brand moments — the CEO portrait, the office space, the actual product. The mix feels authentic because the important things are real.
The Bigger Takeaway
For agencies, the cost of visual content has always been a friction point in proposals. Clients want beautiful imagery but don’t want to pay $5,000+ for a photo shoot on top of the build cost. That tension usually gets resolved by defaulting to stock — which means the site launches looking like every other site in the industry.
A realistic ai image generator changes that equation. You can deliver custom, on-brand imagery at a fraction of the cost, in a fraction of the time. The client gets a better-looking site. You get more competitive proposals. The project stays on budget.
We’re not retiring photographers or canceling every stock subscription. But for the majority of web content imagery — the hero shots, the article graphics, the lifestyle photos that fill out a 40-page site — AI generation has become our default approach. The quality is there, the cost isn’t, and our clients can’t tell the difference.
That’s a hard combination to argue with.
Leave a Reply