The fastest way to make a good website look fake
AI image tools have crossed a line in the last two years. They no longer produce obviously wrong images with extra fingers and melting furniture. They produce images that look right for about two seconds — long enough to pass a quick glance, short enough to fail the moment a visitor actually looks.
That two-second window is the problem. A visitor skims your homepage, your team page, your portfolio. Their brain is quietly asking one question the whole time: “is this a real company?” Every image on the page is evidence, and an AI image in the wrong place is evidence that points against you.
The mistake most businesses make is not that they use AI images. It is that they use them where the image has to prove something true.
Where AI images quietly cost you trust
The rule is simple: the closer an image gets to a claim about reality, the more it costs you when it is fake. Here is where that shows up.
Faces. Nothing triggers the “this is not a real company” reflex faster than an AI-generated face. The skin is too smooth, the light does not sit right, the eyes are slightly off. Your customers may not be able to say what is wrong. They just know something is. A team page full of AI faces tells a visitor that you could not even show them who works there.
Your actual work. If you build houses, an AI render of a “beautiful modern house” tells the customer nothing about the house you would build for them. Worse, if they later discover the image was generated, every real claim on your site becomes suspect. Portfolio and case-study images are the last place to use AI, not the first.
Your products. A generated product photo that looks slightly better than the real thing is a promise you cannot keep. When the delivered product does not match the picture, you have not sold a product — you have manufactured a return, a dispute, and a one-star review.
Anything a customer can compare to reality. Your office, your workshop, your location, your equipment. If a visitor can drive past it or walk into it, the image had better be real.
The test for every image fits in one sentence: if a customer believed this was real and it is not, would that hurt us? If the answer is yes, use a real photo.
Where AI images genuinely help
None of this means AI images are off-limits. They are simply the wrong tool for images that carry a claim. For images that set a mood or explain an idea, they are often better than anything you could afford to shoot.
Abstract backgrounds and textures. Gradients, grain, soft geometry, subtle patterns. These do not claim anything about reality, so there is nothing to be caught out on. AI is faster and cheaper than buying stock, and the result is unique instead of the same stock texture every competitor uses.
Concept illustrations. When you need to explain a process, a service, or a workflow that has no physical form, a consistent set of AI illustrations can be clearer than any photo. A consultancy explaining “how we run a website project” does not need a photo of a whiteboard — it needs a clean visual metaphor.
Visual exploration before a real shoot. AI is excellent at answering “what should we actually photograph?” You can generate twenty concept directions for a hero image, agree on a direction with the client, then shoot or source the real thing. The AI image never ships; it just removes the expensive guessing.
Editing real photos. This is the most underrated use. Removing a cluttered background, cleaning up a product shot, upscaling an old photo to modern resolution, recoloring a logo to sit on a dark background. The photo stays real; AI just removes the reasons it looked amateur.
The pattern: AI is great for the visuals that support your message, and wrong for the visuals that are supposed to prove it.
The consistency problem nobody plans for
A single AI image can look excellent. Ten AI images on one page usually do not, and the reason has nothing to do with image quality. Each one was generated in a slightly different style — different lighting, different color temperature, different level of detail — and the eye reads the mismatch as sloppiness before it reads any single image as beautiful.
Real photography gets consistency for free because it happens in one place, in one light, with one camera. AI gets consistency only if you force it.
The fix is to treat your AI images like a brand guideline. Before generating, write down the style once: the model, the style prompt, the aspect ratio, the color palette, the lighting. Reuse it for every image. If you need multiple images in a series, keep the same seed and change only the subject. The result is a set of images that look like they belong to the same company, not ten images that happened to live on the same page.
A five-minute workflow that avoids the worst mistakes
You do not need to become an image expert to use AI images well. You need a short checklist:
- Decide what the image must do. Is it proving a claim, or setting a mood? If it proves something, it has to be real.
- Write the style once. One style prompt, one palette, one aspect ratio, reused everywhere.
- Review at full size. AI artifacts hide in hands, text, and logos. Look at every image at 100% before it ships.
- Run the trust test. Would a customer believe this is real? If the answer is “yes, and that is a problem,” replace it.
- Check the license. Most AI tools allow commercial use, but read the terms — some limit it, and none will protect you if the image includes a recognizable person, brand, or logo.
The checklist takes minutes. The damage it prevents is measured in lost enquiries.
Real photos and AI images are not competitors
The strongest websites we see do not choose between real photography and AI imagery. They use both, in the right places. Real photos carry the weight — your team, your work, your product, your space. AI images carry the mood — the backgrounds, the textures, the illustrations that make the real photos feel part of a considered design.
Get that split right and AI saves you a real amount of money and time without costing you a single customer. Get it wrong — put a generated face next to a real price — and the best copy and the best layout in the world will not recover the trust you just spent.
If you are not sure which of your images is doing real work and which is quietly working against you, that is exactly the kind of review we start every website project with.
Related service: Web design and content strategy