A business licenses a stock photo of a smiling contractor pointing at a clipboard. So do four hundred other businesses in the same category. The image looks fine on any one site. Across the category, it is the same signal repeated four hundred times, which is another way of saying it is not a signal at all.
That used to be purely a branding problem. It still is one, but it is now also a search problem, because both classic search and answer engines read images as evidence, and evidence that is identical everywhere proves nothing anywhere.
The Same Photo On Ten Thousand Sites
A stock file has one name, one set of embedded metadata, and often the exact same alt text pasted in by whichever template or theme shipped it. Search engines can and do detect duplicate images across the web the same way they detect duplicate text. A photo that appears identically on ten thousand domains cannot be used to distinguish any one of those domains from the others.
Original photography is the opposite case. A photo taken on your job site in Phoenix, of your team, at your location, exists nowhere else. It is a unique data point search engines can associate with one entity, and that association compounds every time the same faces, the same location, or the same equipment reappear across other original photos on the same site.
What AI Overviews and Multimodal Search Actually Do With an Image
Answer engines like ChatGPT, Perplexity, and Gemini are increasingly multimodal, which means they read an image alongside the text around it rather than skipping past it. A photo becomes part of the evidence a model weighs when deciding what to state as fact and what to show in a visual answer.
A stock photo gives that process nothing to extract. It carries no location, no real detail, and no way to confirm it depicts your actual work. An original photo, described honestly in the surrounding copy and the alt text, gives the model something specific: a real place, a real service, a real detail worth repeating. This is the same corroboration problem we described in our post on why AI recommends your competitor. Engines favor whoever has evidence beyond a claim, and a stock image is a claim with a picture attached, not evidence.
The Structural Fix: Alt Text, Filenames, and ImageObject Schema
Three things turn a photo into something machines can actually use. Alt text that states the specific subject, service, and location rather than a generic description or nothing at all. A filename that describes the image instead of a camera-assigned string of numbers. And, on pages carrying real proof, ImageObject schema naming the subject, the creator, and where relevant the location the photo was taken.
None of this works if the underlying photo is generic. Schema and alt text can only describe what is actually in the frame. Applied to a stock photo, accurate alt text just states more clearly that the image is generic. Applied to original photography, the same fields become a genuine fact the page can be cited for.
Original Photography Is Also an Experience Signal
We wrote before about E-E-A-T for AI search and how answer engines weigh whether real experience stands behind a page. Photography carries that same weight. A page showing the actual team, the actual space, or the actual work reads as written by someone who did the thing, while a page filled with stock imagery reads as written by someone describing a category from the outside.
That distinction is invisible to a casual glance and obvious to an engine cross-referencing a claim against the images around it. A remodeling company that shows the same kitchen in three original photos and the finished result is corroborating its own text. A remodeling company that shows a stock kitchen no different from a competitor's is asking to be trusted on words alone.
Scope the Shoot for the Structure, Not After It
The alt text, the filenames, and the schema all get harder to write well after the fact. A photographer who knows the final page needs an image naming a specific city, a specific service, and a specific detail can compose and log the shot with that in mind. A photographer handed a generic brief and told to fill a website with images cannot, and the result is technically original but functionally as generic as the stock file it replaced.
This is the same argument we made about production and search structure needing to be scoped together rather than handed off between two vendors after the fact. A studio that only shoots, or an agency that only writes alt text after images are delivered, is solving half the problem each. Neither half is enough on its own.
Replacing stock photography with original imagery, and building the alt text, filenames, and schema that make it citable, is part of the work inside our SEO / AEO / GEO Foundation, from $7,500. A photo that exists on one site and nowhere else is worth more to a search engine than a better-lit photo that exists on a thousand.