AI Illustration Instead of Stock Photo: When to Use
Reach for an AI illustration instead of stock photo when the idea on your slide has no honest photograph. If you are trying to show a concept, a process, a future state, or something abstract like "trust" or "data flowing between teams," a staged photo of people pointing at a laptop will always look like filler. A generated illustration can carry the idea directly. Keep the stock photo for the one thing it still does better: a real, specific, recognizable person or place.
That is the short version. The longer version is that most slides fail on the wrong visual, not the missing one, and knowing which situation you are in saves you an hour of scrolling a stock library for a picture that does not exist.
The real question is not AI versus stock, it is concept versus subject
I spend most of my week inside decks, and the visual arguments almost always come down to one distinction. Are you illustrating a concept or showing a subject?
A subject is a thing that exists and that the audience needs to recognize as itself. Your actual product screen. The founder's face on the About slide. The warehouse you are proposing to lease. For a subject, you want the real thing, and if you cannot photograph it yourself, a stock photo of the closest real equivalent is the honest choice.
A concept is an idea with no fixed appearance. "Our roadmap." "How the referral loop compounds." "Three types of customer." There is no photograph of a referral loop. When teams try to force one, they end up with the same tired stock genre: smiling colleagues, a glowing lightbulb, hands stacked in a circle. That is exactly the imagery that makes a slide read as generic, and I wrote about that failure mode in more detail in a breakdown of what looks cheap inside an AI deck. A concept is where a generated illustration earns its place.
When an AI illustration instead of stock photo is the right call
Here is the decision I actually run, in order:
- Does the slide need a specific real person, product, or place? If yes, use a photo you own or a licensed stock photo. Do not generate a fake human and present them as real.
- Is the idea abstract or a process? If yes, an illustration almost always beats a photo, and a generated one beats hunting for a metaphor photo that does not exist.
- Do you need several images that match each other across the deck? Illustration wins here by a wide margin, because you can hold one style, palette, and line weight across ten slides. Stock photos from different shoots never match.
- Is speed the constraint? Generating three directions and picking one is usually faster than filtering a stock library and still settling.
The pattern most teams land on, and the one I use, is simple: illustrations carry the conceptual slides, and a small stock budget covers the handful of slides that need a genuine photograph. One team that ran a month-long swap of stock for generated images reported using AI for roughly the large majority of their visual needs while keeping a small pay-as-you-go stock credit for authentic photos of people. That ratio matches what I see in practice.
Where a photo still wins, and I will not pretend otherwise
Generated illustration is not a universal upgrade. It loses in three places, and pretending otherwise is how decks end up looking worse, not better.
Real people. Faces, hands, and genuine expressions are still where generation is weakest and where a fabricated result is also dishonest. If a slide implies "here is our customer" or "here is our team," a generated person is both a quality problem and a trust problem. Use a real photo.
Photoreal product shots. If the audience will buy or evaluate the actual object, they need to see the actual object. A generated approximation of your own product is a strange thing to put in front of a buyer.
Anything the audience will read as a claim. A photo of a specific building, a specific city, a specific document reads as evidence. If it is generated, it is evidence of nothing, and a sharp audience will notice.
The failure I see most often is teams reaching for photoreal generation, trying to make an AI image look like a photograph, and landing in the uncanny middle where it is neither a clean illustration nor a convincing photo. If you have hit that wall, the fixes I go to are in where AI slide makers still fail and how to work around them. The short answer is to commit to illustration as illustration. Stylized and clearly drawn beats almost-real every time on a slide.
How to keep a generated illustration from looking cheap
Getting the concept-versus-subject call right is half the work. The other half is execution, and this is where slides go wrong even when the choice was correct.
- Pick one style and hold it. Flat, line, isometric, or painterly, choose one and prompt every image toward it. A deck where slide 3 is a flat vector and slide 7 is a moody 3D render looks broken.
- Match your brand palette, not the generator's defaults. Most tools drift toward saturated purples and teals. Name your actual colors in the prompt, or recolor after.
- Leave room for text. Generate with negative space where your headline and body will sit. A gorgeous full-bleed illustration is useless if it fights your copy.
- Downscale the ambition. A simple, clear illustration of one idea beats a busy scene with six half-formed metaphors. Slides are read in seconds.
- Keep the resolution honest. Export large enough that it stays crisp on a projector and on a shared screen. A soft, upscaled image is the fastest tell that something was generated in a hurry.
If you are choosing a tool to do this inside your deck rather than in a separate image app, the trade-offs matter more than the feature lists, and I walk through them in how to choose an AI presentation tool for your situation.
The one thing nobody selling you images will mention: rights
This is the part that gets skipped, and it is the part that can actually cost you.
An illustration you generate is not automatically yours to own the way a commissioned drawing is. In the United States, the Copyright Office addressed this directly in Part 2 of its Copyright and Artificial Intelligence report, published on January 29, 2025 (read on the Copyright Office site on September 12, 2026): human authorship is the bedrock of copyright, works generated entirely by AI are not copyrightable, and prompts alone do not currently give the user enough control to count as the author. In plain terms, a purely generated image is difficult to protect as your own exclusive asset. That is fine for a one-off slide. It is a problem if you are about to make the image your logo or a core brand mark.
The other half of the rights question is whether you are allowed to use the image commercially at all, and that depends entirely on the tool. Some generators are built specifically around this. Adobe, on its Firefly business page (read September 12, 2026), states that Firefly was trained on licensed Adobe Stock content, openly licensed content, and public domain content where copyright has expired, and it offers IP indemnification on paid plans. Other tools make no such promise. Before you put a generated image in a client deck, an investor deck, or anything public, read the specific tool's terms for that plan. Do not assume the rules are the same across tools, because they are not.
For a pitch or a sales deck, my rule is boring and safe: use generated illustration freely for internal and conceptual slides, prefer a commercially-safe generator or a licensed stock photo for anything client-facing, and never generate a logo or trademark. The structure I use for a pitch deck outline keeps the visual load light enough that this is easy to honor.
A quick decision table
| Your slide needs… | Best choice | Why |
|---|---|---|
| A real person or team | Photo (owned or licensed) | Generated people fail on trust and on hands and faces |
| Your actual product | Real screenshot or photo | Buyers need to see the real thing |
| An abstract concept or process | AI illustration | No photo exists; illustration carries the idea |
| A matched set across many slides | AI illustration | One style holds; stock shoots never match |
| A logo or brand mark | Commissioned or licensed art | Purely generated work is hard to own |
FAQ
When should I use an AI illustration instead of stock photo?
Use it when the slide shows a concept, a process, or something abstract that no real photograph captures, when you need several images that match in style, or when speed matters. Stay with a photo for real people, your actual product, and anything the audience will read as literal evidence.
Do AI illustrations look unprofessional on slides?
They look unprofessional when they try to imitate photographs and land in the uncanny middle, or when the style changes from slide to slide. Committed, clearly-drawn illustration in one consistent style reads as intentional and professional.
Can I use AI-generated images in a client or investor deck?
Often yes, but it depends on the specific tool's license for your plan. Some generators are built to be commercially safe and even offer indemnification; others are not. Read the terms for the exact tool and plan before anything goes public.
Do I own the copyright to an AI-generated illustration?
Under current US Copyright Office guidance, a work generated entirely by AI is not copyrightable, because human authorship is required and prompts alone are not enough. That is acceptable for a disposable slide image, but it is a reason not to use generated art for a logo or core brand asset.
Is it faster to generate an image or search stock?
For conceptual slides, generating two or three directions and picking one is usually faster than filtering a stock library for a metaphor that may not exist. For a specific real subject, a good stock search is faster and more honest.