The higher-resolution image of Donald Trump kissing Natalie Harp on a golf course appears to give her six fingers.
That is the detail that stops the story cold.
The photo spread online as if it captured a real moment. Trump. Harp. A golf course. A kiss. Then the clearer version showed the hand holding the golf club. The fingers do not line up with a real hand.
That is not a small flaw.
The image also carries a watermark linked to OpenAI tools. Its golf equipment and background show other odd shapes. The higher-resolution file looks like an AI image, or an image changed by AI.
This is the part that still amazes me. The fake did not need to look perfect. It only needed to look good for a few seconds on a phone. Long enough for someone to share it. Long enough for a caption to turn a picture into a claim.
Then the image gets blurry.
The lower-resolution version is harder to judge. The six-finger problem fades. The strange details disappear. So does the easy proof. A bad image can hide the signs that would expose it.
That creates a nasty gap between what people see and what the file can support.
The sharp version has several red flags. The blurry version does not prove the scene happened. It also does not offer enough detail to settle the question on its own. The image remains disputed, but the available evidence points strongly toward AI generation or digital editing in the clearer version.
The identity of the woman is also not certain from the image. Her face is turned away. People online said she was Natalie Harp, but the photo itself cannot establish that.
This matters because the claim is political even before anyone says a word about politics.
An image of Trump and Harp kissing can affect how people judge Trump, Harp, the White House, and the claims already moving around them. A real photo would carry one meaning. A fabricated photo would carry another. The image can do damage before anyone checks which world they are in.
That is the power of synthetic media. It does not need to win a fact check. It needs to arrive first.
The two versions also show how quickly AI can exploit image quality. The clearer file reveals the extra finger and the watermark. The blurrier file looks more convincing because the errors are hidden. This is a strange reversal. Better quality can expose a fake, but lower quality can help it travel.
I find that both thrilling and unsettling.
The technology can create a believable political scene from nothing. It can also create the evidence that makes people doubt real images later. The same tools push in both directions. They help make false pictures. They also force us to inspect every picture more closely.
But inspection has limits.
A person can spot six fingers. A person may notice broken golf clubs or uneven detail. A person cannot always recover the source, date, camera, or real setting from a compressed file. When an image has no trusted origin, visual judgment is only one part of the check.
The White House called the image fake. That is a claim from the White House, not proof by itself. The higher-resolution version gives that claim support, but the lower-resolution version remains less clear. Those facts need to stay separate.
That separation is important. It is easy to say, “The picture is fake,” and move on. It is harder to say exactly what the evidence shows.
The evidence shows strong AI signs in the higher-resolution image. It does not show a verified source for the photo. It does not prove that a kiss happened. It does not prove who the woman is. It does not turn the blurry copy into a real photograph.
And it shows how little a viral image may actually contain.
A picture can look like a record of an event. It may be only a record of a prompt, an edit, or a repost. That is a huge change for public life. The file can feel concrete even when its history is missing.
I do not want to become calm about that. Calm can sound like surrender.
The exciting part is that we now have tools that can reveal hidden AI marks and visual errors. The catch is that these tools do not give every image a clean answer. A watermark can support a finding. Its absence cannot prove reality. Blur can remove clues without adding truth.
So the Trump-Harp image should not be treated as proof of a kiss. The sharper version has too many signs of AI generation or editing to trust. The blurry version has too little evidence to rescue the claim.
That is not a satisfying ending.
It is the honest one.