White House responds to photo appearing to show Trump kissing his assistant on golf course
But as the photograph spread across social media, the story quickly became less about what people thought they were seeing—and more about whether the image could be trusted at all.
Within hours, screenshots, cropped versions, enlarged copies and reposts were appearing across X and other platforms.
Some users treated the picture as genuine. Others immediately suspected artificial intelligence. And soon, fact-checkers began examining something that most people scrolling past the image would never notice: the tiny visual details hidden inside the photograph.
According to Snopes, two noticeably different versions of the picture were circulating. One was relatively clear and high-resolution. The other was much blurrier.
That difference turned out to matter enormously.
When specialists examined the clearer version, they found several warning signs associated with AI-generated or AI-edited imagery. The uploaded report notes apparent problems with the hands and describes a watermark indicating that artificial-intelligence software may have been involved in creating or modifying the picture.
Deepfake-detection specialist Hany Farid told Snopes that the higher-resolution version was clearly AI-generated. But the blurry version created a more complicated problem: because so much visual information had been lost, investigators could not reach the same definitive conclusion about that version from the available evidence. Snopes therefore described its authenticity as inconclusive rather than confirming it as genuine.
That distinction is important.
An inconclusive image is not the same thing as an authentic image.
And an AI-generated version circulating online does not automatically explain where every lower-quality version originated.
Meanwhile, the White House rejected the image outright. A spokesperson described it as “clearly a fake photo,” according to the report and subsequent coverage.
Yet even after that denial, the photograph continued traveling across social media, where the mystery surrounding the unidentified blonde woman added another layer to the story.
Only the back of the woman’s head can be seen clearly enough for viewers to make general observations. That did not stop many people online from identifying her as Natalie Harp, one of Trump’s aides. But the viral picture itself does not establish the woman’s identity, and the assumption that she is Harp remains just that—an assumption.
Harp has drawn attention before because of reports describing her unusually close working relationship with Trump. The uploaded article references a letter reported by The New York Times in which Harp allegedly apologized for potentially embarrassing Trump while accompanying him during a golf trip to Scotland. The letter reportedly ended with the words, “With all my heart, Natalie.”
Another report cited in the original story described an alleged episode in which Harp traveled in the trunk area of an SUV after being told there was no available motorcade seat while trying to accompany Trump to court.
Those stories have helped fuel intense online speculation about the nature of their relationship.
But speculation is not proof.
Snopes previously examined online claims alleging a romantic relationship between Trump and Harp and reported that the rumors remained unproven. Public reporting about Harp’s devotion to Trump and her proximity to him has generated attention, but it has not established an affair.
That is why this viral photograph has become such a striking example of how difficult the AI era has made breaking news.
Years ago, a dramatic photograph might have been treated as powerful evidence on its own.
Today, every pixel can become part of the investigation.
Hands must be examined.
Lighting must be examined.
Shadows must be examined.
Compression must be examined.
Metadata, watermarks and image history can all matter.
And sometimes even experts cannot immediately determine what happened to an image before it reached millions of screens.
The situation becomes even more complicated when multiple versions appear online at different resolutions. A high-quality AI-generated picture can be compressed, screenshotted and reposted until many of the flaws that originally exposed it disappear.
Conversely, a blurry image cannot automatically be declared fake simply because investigators cannot verify it.
That leaves responsible fact-checking trapped between two dangerous conclusions: believing everything and believing nothing.
As of September 18, 2026, the strongest publicly reported evidence points to the higher-resolution viral image being AI-generated, while analysis of the lower-resolution version has remained inconclusive. The White House has denied that the photograph is real.
And there is another detail worth remembering.
The photograph originally began appearing publicly on X around September 14, according to the uploaded report. By September 16, it was already circulating widely enough to generate international coverage and a White House response.
In barely two days, an image whose authenticity was disputed had traveled far enough to become a political news story.
That speed is part of the bigger story.
Because once millions of people have seen an image, correcting the record becomes far harder than spreading the original claim.
Some people will remember the denial.
Others will remember the fact-check.
But many will remember only the picture.
And that may be the most consequential question surrounding this controversy—not simply whether one photograph was created by artificial intelligence, but how easily an uncertain image can become attached to a real person, a real relationship and a real political narrative before anyone has established what actually happened.
For Natalie Harp, the renewed attention means old reports about her relationship with Trump are once again being circulated alongside a photograph that has not been authenticated as showing what social-media users claim.
For Trump, the White House position is straightforward: the picture is fake.
For fact-checkers, however, the technical investigation produced a more precise conclusion: the clearer version showed strong evidence of AI generation, while the blurrier version could not be conclusively authenticated or debunked from the image alone.
And that leaves one mystery still hanging over everything:
Where did the earliest version of the photograph actually come from—and can investigators trace the image far enough back to discover what existed before the internet began copying, compressing and transforming it?
Because if that answer emerges, this story could take another unexpected turn…
