Photos have become one of the most important criteria by which people choose a vacation destination. However, in a time when images generated with artificial intelligence are increasingly difficult to distinguish from the real thing, many tourists end up discovering only at their destination that the place looks completely different from what they saw online.
A survey of 2,000 adults found that only 5% of participants were able to correctly identify all real photos in a test where they were compared to AI-generated images. At the same time, 74% of respondents say they would not book a vacation without first seeing photos of the destination.
Price remains the main criterion in choosing a vacation (62%), but photos and videos take second place, indicated by 47% of participants, The New York Post reports. For people who travel at least six times a year, the influence of images is even greater: almost half (48%) say that they weigh heavily in the final decision.
Why AI-generated images seem more real to us than genuine photos
For Alexandru Dan, professor and researcher in the ethics of artificial intelligence at ASE Bucharest, this survey reflects a change in the way we relate to the online environment. Today, a photograph is no longer a guarantee of reality.
“By the nature of my work, I analyze the evolution of AI-generated content and the deepfake phenomenon on a daily basis. Therefore, I see this study as a confirmation of a reality that I have been documenting for a long time: a simple image no longer proves that that place, landscape or object really exists.” explains the expert for “The Truth”.
What is really striking about the results of the US test is not just the confusion of the respondents, but their clear inclination to fake. The numbers are telling: only 5% of the participants were able to correctly identify all three real photos. Moreover, in the case of a landscape with a lake, the balance tipped in favor of artificial intelligence – 23% believed that the AI-generated image was the authentic one, while only 22% chose the real photo.
“If people had guessed completely at random, statistically they would have performed better. This shows us that we are not just dealing with an inability to distinguish truth from falsehood, but with a real preference: the image created by AI has come to seem more ‘authentic’ to us than reality itself.” emphasizes Alexandru Dan.
The researcher recalls that a similar paradox was documented as early as 2023. A specialized study published at the time showed that computer-generated portraits were perceived as “more humane” than the faces of real people. Moreover, the people who were wrong the most were precisely the most convinced of the correctness of their choices.
According to the specialist, the explanation of this phenomenon strictly depends on the way human perception works.
“Our weakness . . . which these images exploit is instead a general one, because such a program produces an ideal version of the world, with perfect light, intense colors, symmetry, no clutter and no clutter, just as we, when we look at a picture, judge ‘how real it is’ by these signs on the surface, so that reality, with its small imperfections, comes to seem less believable than the polished image, and the stake I always repeat is that, until the current generation of programs, a photograph was proof that something had really existed in front of the camera, while the generated image breaks this link and proves nothing, even though it looks in every way like proof.” is the opinion of the researcher.
Alexandru Dan draws attention to a dangerous paradigm shift: “The point I keep stressing is this: until these algorithms came along, a photo was absolute proof that an object actually existed in front of the camera. The AI-generated image definitively breaks that link. Today, a photo no longer proves anything, even though, visually, it has all the force and weight of proof.”
The overconfidence trap: the illusion that we can detect fakes
The survey also reveals another problematic aspect: the paradox of certainty. More than half of the participants (52%) declared themselves very confident in their ability to detect artificial images. In reality, their performance contradicted this belief. Almost 40% of those who were fooled by at least one image admitted that they were downright surprised by the result, being absolutely convinced that they had made the right choices.
In this context, the expert sounds the alarm: in the current state of technology, there is no longer any simple and infallible method by which the general public can distinguish, just with the naked eye, a real photo from one generated by artificial intelligence.
The mistake many parents make when children argue: “Empathy, responsibility and forgiveness are not built by repeating formulas”
He cautions against the idea that there are certain “signs” by which the average person can recognize a fake image. “Still, the classic signs are there, from meaningless text on companies or menus, wrongly drawn hands, buildings that don’t stand up, shadows and reflections that come from somewhere other than light, to patterns that repeat themselves perfectly in water, tiles or foliage, textures that are too smooth and edges that seem to melt, but the problem is that these traces are reduced from one version of the programs to another. And the survey confirms an uncomfortable thing, namely that the most convinced that they recognize a false image are often the ones who are wrong, so looking with the naked eye is not a method that the general public can rely on.” he confesses.
In fact, adds the specialist, the second, important confusion is that between an image made completely by AI and one that is only heavily edited, because a large part of hotel photography is not created from scratch, but processed, through very wide lenses that make the camera seem much larger, by erasing the clutter from the frame, by replacing the sky and by pushing the colors. “And these cases are not caught by looking for AI traces, because the basis is a real photo, only embellished, which is why the most reliable clue is not in the picture, but around it, in the match between what the picture shows and the rest of the information, whether it’s a pool that is present in the brochure but does not appear in any customer pictures, a “sea view” that does not match the map, or an image with no clue where originates, so multi-source verification remains better than visual inspection,” emphasizes Alexandru Dan.
Seven out of ten tourists say they were disappointed by the reality
According to the survey, seven out of ten respondents say they have at least once arrived at a destination that looked noticeably different from the photos they saw before booking. Of these, more than 80% believed or suspected that the differences were caused by the use of artificial intelligence.
The experience has made many people change the way they check out a destination. Thus, 38% say they now prefer to ask people they know, 37% look for recent photos instead of promotional images, 35% look especially at photos published by other tourists, and 34% give more trust to official sources of the destination.
However, the professor points out that not every difference between photos and reality is produced by AI: “The effect is already visible in the numbers, since in the same survey seven out of ten people say they arrived at a destination that looked noticeably different from the pictures they chose it from, and over 80% of them believed or suspected that AI was involved, which forces me to clarify correctly, that we’re talking about a hunch on their part, not proof that AI was used, because people now blame AI for everything disappointment, whether it was involved or not, and the loss of trust becomes a problem in itself.”
What we can check before booking
In the opinion of the specialist, the best protection is not to try to “guess” if a photo is generated by AI, but comparing information from multiple sources.
“I recommend not to rely on a single check, but to use several sources that confirm each other, and the most useful are recent photos of other tourists, pictures from reviews and posts with the location passed on them, because they always lag behind the campaign images, while a video material is more difficult to fake from head to tail than a still picture, although even there things are no longer certain. For promises like «at the shore of the sea’ or ‘five minutes from the centre’, the satellite map and Street View clarify in seconds, as a reverse image search shows you if the same picture appears in other places or if it is taken from image banks, without forgetting the date, because when the pictures and reviews were published matters as much as what they show” points out Alexandru Dan.
The vicious circle that affects more and more young people: loneliness pushes them to spend more and more time online, and the Internet isolates them even more
Moreover, starting from August 2, 2026, some of the images generated with artificial intelligence will have to be marked in accordance with the provisions of the European regulation on artificial intelligence. However, the specialist warns that this measure will not completely eliminate the problem.
“My view, as an AI ethics researcher, is that a visible label on absolutely anything… would do more harm than good, because we would get used to it and not see it again, and at the limit it would lead us to believe that an image is either completely fake or completely real, although there are many nuances between the two. A balanced solution would combine a provenance mark that accompanies the file wherever it circulate with a visible label kept only for the really deceptive cases, i.e. deepfake materials and fabricated places, displayed where the human makes the decision, i.e. on the booking page. However, these markings can be deleted, and the platforms recompress the images anyway, which is why the law only requires marking “to the extent technically possible”, so the tag reduces the risk, but does not eliminate it, and part of the verification work remains, after all, ours. he concludes.
