After the AI Hangover: What Happens to Hotel Photography?

AI, hotel photography, and the difference between generating images and building a vision

There is a sentence I hear more and more often: “With AI, you can do anything now.” Technically, that is almost true. Visually, it is only half true. And in hotel photography, it can even become dangerous.

Because a hotel photograph is not just a beautiful image. It is the first encounter between a guest and a place. It arrives before the booking, before check-in, before the real experience. It is a promise.

And a promise, if it works, must be desirable. But it must also be credible.

A room, a terrace, a lobby, or a restaurant must build expectation without betraying it. It must make people want to be there without inventing a place that does not exist. This is the fine line hotel photography has always walked: making a space stronger, clearer, more memorable, not more false.

AI is changing exactly that line.

Not because it replaces photography, but because it changes the way photography is conceived, produced, edited, and finalised.

Reality Has a Texture

For years, generative AI occupied a separate territory: Midjourney, dedicated platforms, spectacular experiments, images that were often beautiful but somehow suspended. It worked very well when the goal was to create an imaginary world.

Hotel photography lives somewhere else.

Here, the marble has to look like that marble. The view has to be that view. The room has to remain recognisable. Light can be interpreted, of course, but it cannot lie too loudly.

The real shift came when tools such as Adobe Firefly and Generative Fill were integrated directly into Photoshop, becoming part of the photographer’s daily workflow. AI stopped being a special effect and became a concrete part of post-production.

But it is not magic. It is a tool for visual direction.

The Set That Wasn’t There

Anyone who photographs hotels knows this: often, the property is not truly ready to be photographed at its full potential.

Flowers are missing. Plants are missing. Styling is missing. Lifestyle details, small objects, and support from the F&B department. Sometimes, very little would be enough to change the perception of a space: a plant in the right corner, a tray, a book, a cup, a detail that lets the scene breathe.

Before, all of this was solved through realistic retouching: finding the right images, inserting them, adapting perspective, light, shadows, colour, and grain. It could be done. But it was slow, expensive, and not always convincing.

Today, I can select a precise area of a scene, type “indoor plant,” and, in a few seconds, obtain an element that is coherent with the space's light and mood.

Interestingly, generic prompts often work better than overly specific ones. If I ask for a Kentia palm, the result may look forced. If I simply ask for an indoor plant, AI sometimes finds a more natural solution, less literal, more right for the image.

It is a small lesson in method: the prompt should not prove how well we can describe. It should help the image find its balance.

The Sky Is Not Enough

The most interesting leap, however, is no longer just about adding or removing objects.

The new frontier is AI’s ability to read the image as a whole: light, atmosphere, depth, volumes, and the relationship between space and perception.

Anyone who photographs hotels knows that the weather is not a detail; it is part of the production. Sometimes a shoot is brought forward, postponed, or completely rethought due to the weather, with all that entails for property availability, arrivals and departures, accessible rooms, and usable spaces. And quite often the opposite happens, too: a perfect sunny day is wasted because another logistical piece does not fall into place.

This happened to me recently in Rome, at a hotel with a beautiful swimming pool. The objective was very clear: aerial geometric shots, open umbrellas, empty pool, sharp light, clean shadows. After days of phone calls, planning, and weather checks, we finally found what looked like the right window. On paper, the sky was good. In reality, it was hazy. The blue was there, but the mood was not; the strong sun, the shadows, that visual tension the image needed were missing.

Replacing the sky in Photoshop has been possible for years. But a blue sky over a scene lit as if it were an overcast day is not credible. The sky changes, but the light remains wrong.

Today, AI can do something more sophisticated: it can rework the entire image as if it were truly bathed in bright, clear daylight. It not only changes the sky; it changes the quality of light, its direction, the contrast, the readability of volumes, and the presence of shadows.

It does not apply a filter. It rebuilds a mood.

Small Shooting Dramas

AI becomes truly interesting when it stops being a toy and becomes a problem solver.

A few years ago, at a hotel in Milan, I found myself facing one of those classic details that can ruin an image: carefully prepared rooms, good light, a ready composition, but beds with visible metal legs. I tried to recreate the missing bed bases using the generative tools available at the time, with the correct rooms as a reference. It worked, but with considerable effort: inconsistent results, perspectives to fix, details to correct.

Recently, something similar happened to me in Rome. Same problem, much more mature tools. With a clear reference and a well-calibrated prompt, the result arrived in a few minutes: coherent in style, credible in proportions, consistent from one image to the next.

The Risk of the Over-Correct Image

The real problem with AI is not that it produces bad images. Often, it produces images that are almost too pleasant.

The risk is the average aesthetic: perfect light, perfect styling, perfect tables, perfect plants, perfect people. A visual world that is clean, polished, premium, but also anonymous.

It is the “AI look”: correct images, but without territory. Without friction. Without memory.

And in hotel photography, this is a serious issue. Because a boutique hotel in Rome, a seaside resort, a business hotel in Milan, and a historic residence cannot all speak the same visual language.

It is not enough to make them desirable. They must be recognisable. Distinct.

It is similar to what happened with WordPress templates: suddenly, everyone could have an elegant website. But many websites started to look the same. The difference was no longer the template; it was how the template was interpreted, modified, and bent around an identity.

With AI, something similar is happening. The tool democratises access to visual production, but it does not guarantee a point of view.

Output Is Not Outcome

AI generates output. A lot of it. Fast. Spectacular.

But an output is not a finished image. And above all, it is not a vision.

The difference is still made by the eye: knowing what to remove, what to keep, where to push the light, and where to stop. Understanding when a scene is aspirational and when it becomes false. Understanding whether an image truly tells the story of a property or is simply following the average taste of the algorithm.

AI can generate infinite variations, but it does not know a hotel’s positioning. It does not know its history, its audience, or the kind of experience it promises.

That part remains human.

And in fact, today it is no longer just a matter of delivering “some photos.” We need to build a coherent visual world: hero images, room sets, details, lifestyle, F&B, exterior mood, visual continuity across the website, OTAs, social media, and campaigns.

AI does not remove this work. It makes it more efficient. But it does not decide which story the hotel should tell.

The Best Prompt Is Not Written in the Box

Hotel photography has always been a form of interpretation.

We choose the best moment, the best angle, the best light. We arrange pillows, hide cables, clean surfaces, and control perspectives. We do not show a space at its worst. We represent it in its most desirable version.

AI does not change this logic. It amplifies it.

That is why the point is not how much we can intervene in an image, but how far that intervention remains coherent with the place, the light, and the promise of the hotel.

We need to reposition AI in our minds. The generated output is not the final product. It is the beginning: raw material, a possibility, a very fast draft, sometimes surprising, sometimes completely off track.

The value is not in the fact that the machine produces something. The value is in what we decide to do with it.

Because, in the traveller’s journey, the first experience often does not happen in the hotel. It happens in front of a photograph.

And that photograph has to do something very difficult: make people want to be there, without promising a place that does not exist.

The prompt is not the photograph.
The prompt is only a tool.
The best prompt is the years of experience you bring to the image.