Making the Call

AI-generated, stock, rendered or real: which photos belong on a manufacturer's website?

Summarize with AI

Is that picture of your plant floor a photo of your plant floor? Most buyers looking at a manufacturer's website assume it is. That's the whole answer to the question of which photos to use. If an image says "this is us" (your floor, your machines, parts you made, your people), it has to be a photo of exactly that. AI-generated and stock images can't make that claim, so they belong only where nobody reads them as proof: backgrounds, diagrams, decoration. A render is honest when it's accurate and labelled as a render. Touch-ups are fine until they change what a buyer would rely on.

What is a photo on an industrial website actually for?

Proof, mostly. A buyer comparing suppliers uses the pictures to check what the copy only claims: the size of the operation, the age of the equipment, how parts are handled. Why stock fails that check is the subject of the credibility signals buyers look for, and what's worth shooting is in the manufacturing website examples piece.

Sort images by what each one claims. A wide shot of a machining floor claims you have that floor. A close-up of a welded frame claims you made it. A blurred abstract background behind a headline claims nothing. Only the first two are evidence, and they're the ones buyers look at hardest.

That matches what we've heard. On one call this year, a fabricator said buyers had told them that the site didn't show the work they do. Another prospect wanted their own floor photos and short process videos in place of stock.

Can buyers tell an AI-generated factory from a real one?

Sometimes, and an engineer or plant manager is better placed than most, because the details image generators tend to get wrong are ones they'd notice: machines with no cabling or guarding, tooling that couldn't work, floors cleaner than any running plant, safety signs that don't say anything.

But spotting it today isn't the point. A generated floor is a claim you can't back up, and it gets checked when a buyer visits or an auditor walks through. Once a buyer finds one invented picture, they have a reason to doubt the certifications next to it.

The file can also give it away. Google's documentation says Google Images reads the IPTC "digital source type" field, including a value for images "created algorithmically using a model derived from sampled content", and that when an image carries C2PA metadata, Google may show in "About this image" how it was made or edited (Google Search Central, image metadata, updated 2025-12-10). In practice, CMS uploads and image-optimization pipelines often strip IPTC and C2PA data when they resize or compress a file. That matters if you list products in Google Merchant Center, whose image rules say images created with generative AI must carry metadata saying so and that it mustn't be removed (Merchant Center, image_link). Check what your site and feed actually keep.

Where are AI and stock images fine?

Anywhere the image isn't standing in for something real: a texture behind a heading, an illustration, an icon, a process diagram.

Editing real photos is fine too, AI tools included, when the edit keeps what's shown: color and exposure, dust and glare, a stray cable, a cleaner background on a product shot. It stops being a touch-up when it changes what a buyer would rely on. Adding machines you don't own, widening a bay, smoothing a weld or changing a finish all cross that line. A useful test: if the buyer stood in front of the real thing, would they feel misled?

When is a render the honest choice?

More often than people think. A CAD render shows the product as designed. It's legitimate for a product that isn't built yet, a configurable line with too many variants to photograph, or a part whose real photos you can't publish. Caption it as a render.

The line is the same one that applies to retouching. A render or a retouch is fine when what the buyer would rely on is accurate: the part's geometry, finish, material and tolerances. Studio lighting and a clean background don't change any of that. A render showing a surface finish your process can't produce, or a material you don't use, has crossed the line, however clearly it's captioned.

That also answers the company with hundreds of SKUs and no photographer. Consistent catalog shots or CAD renders for the long tail are fine as long as they're accurate. Spend the real photography on the pages that carry proof: the floor, the equipment, finished work, your people.

Where a render goes wrong is where AI does: standing in for capability, like a rendered machining cell you don't own.

What can you show when customers' parts are under NDA?

NDAs come up again and again on calls about proof, and one business had removed every client logo because client contracts forbade them. Describing the work in words is covered in what to leave off your website. For the pictures:

  • Photograph the capability, not the part. The machine that made it, the inspection area, the weld cell, material on the rack.
  • Make a sample. A test part in the same material and process, made to be photographed, shows what you can hold without showing anyone's design.
  • Go close. A tight crop of a surface finish or a weld bead often reveals nothing about the design. Check before publishing.
  • Ask. Some customers will approve a specific photo, especially a cropped or partial one.

A rendered or AI-generated copy of a customer's part doesn't solve the NDA. It shows the design just as much.

What should a distributor do with the manufacturer's photos?

Keep them. The manufacturer's shot is how a buyer confirms it's the right part, as long as your dealer agreement covers using it.

The trouble is that they show the product, not you. One distributor told us a single brand's photos had taken over their site, so it read like a dealer for one line. What only you can show is the branch shelf, the kitting bench, how an order is packed, your trucks. A handful of those on the pages buyers land on says something the catalog shot can't. The rest is in distributor marketing.

Do you have to tell buyers an image was made or changed with AI?

Here's what the primary sources say. It isn't legal advice.

The FTC's advertising guide for small businesses says an ad is deceptive if it's likely to mislead consumers acting reasonably and the point is material, and that the FTC looks at "words, phrases, and pictures" to judge what an ad conveys (FTC, Advertising FAQs). It sets out no AI labelling rule.

At least one state has gone further. Pennsylvania's Artificial Intelligence Deceptive Advertising Act (SB 806, Act 99 of 2026, signed October 22, 2026, effective 60 days later) bars an advertiser from knowingly, recklessly or willfully using synthetic content, meaning content "significantly modified or generated" by AI, in a way that creates a false consumer perception about the nature, origin or authenticity of the goods or services, without a clear and conspicuous disclosure (bill text, PN 1808). Content that's evidently fictional, parody, satire, commentary or criticism, or limited to text is exempt. It's a Pennsylvania statute, enforced by the state Attorney General and district attorneys, and a court can award a civil fine of up to $3,000. The text doesn't say how it applies to ads aimed at businesses.

The practical reading is simpler than the law. On a page whose job is proof, an image that needs a disclaimer to avoid misleading anyone is usually the wrong image. A generated plant floor with a disclaimer under it is still a plant you don't have.

In a rebuild, that's where manufacturing website design starts: what a buyer needs to see before trusting the rest.

Common questions

Can buyers tell if the images on a manufacturer's website are AI-generated?

Sometimes, because industrial buyers know what real equipment looks like: missing cabling or guarding, impossible tooling, floors too clean to be running. Google can also read AI metadata in the file, though many sites strip it on upload. The bigger risk is that a generated image of your plant gets checked on the first visit or audit.

Is it OK to use AI-generated images on a manufacturer's website?

Yes, for backgrounds, illustrations and diagrams. No for your facility, equipment, people or parts you've made, because buyers read those images as proof.

Can you use stock photos on a website?

Legally, yes, within the license. Whether you should depends on what the image is for. Stock is fine as decoration, never as a picture of your company: a stock photo of a generic factory is evidence of nothing.

Are product renders OK instead of real photos?

Yes, when they're accurate and captioned as renders: for products not yet built, configurable lines and parts you can't photograph. Don't use them to suggest equipment or finished work you don't have.

Our agency used AI to touch up photos of our products and shop floor. Do we need to tell customers?

Routine fixes like color, exposure and dust usually don't mislead anyone, which is the FTC's test. Edits that change what a buyer would rely on, such as adding equipment or changing a finish, are different, and Pennsylvania's law, in effect from December 2026, requires a disclosure when content significantly modified or generated by AI creates a false impression about what's being sold. Ask your counsel about specific images.

Should a distributor use the manufacturer's product photos?

Yes, to identify the part, if your dealer agreement allows it. Then add photos of what only you can show, such as branch stock, kitting and packing, so the site shows your business and not just the brands you carry.

CMS (content management system)Trust signals (credibility signals)
Share

Let's build something together.

A 30-minute call. We'll listen, dig into the details, and tell you honestly whether we're the right partner, or point you to someone who is.

Book Intro Call ↗