The label vanished within hours, but a forensic photo comparison points to what really tripped X's detector #
Karoline Leavitt announced her exit as White House Press Secretary on Wednesday, and within minutes X slapped a 'Made with AI' label on her farewell post. The tag vanished by nightfall. What tripped it tells you how to read the same flag on your own feed.
Leavitt, 28, said she will leave at the end of August. President Donald Trump confirmed the departure in a Truth Social post, saying she would step back to spend more time with her two young children and stay on as one of his top outside advisers heading into the November midterm elections. Neither Trump nor Leavitt named a successor.
The resignation matters because Leavitt is the youngest White House Press Secretary in US history, and one of the administration's most visible faces. Yet the detail spreading fastest online wasn't her exit. It was the small grey note at the bottom of her post.
What the 'Made With AI' Label Actually Means #
Washington Post technology reporter Drew Harwell was among the first to flag the tag, noting on X that Leavitt's goodbye message carried a 'Made with AI' disclosure. Reporters at CBS News and Status quickly confirmed it.
Plenty of readers assumed the label meant a machine wrote her statement. It doesn't. Under X's own guidelines, the platform applies the tag to any media it determines is AI-generated or manipulated. That covers photos and graphics, not just text. The label says nothing about who wrote the words.
X did not respond to requests for comment on what triggered the note. The White House also stayed silent.
The Photo That Likely Set It Off #
Leavitt's post included two images of her with Trump. A forensic comparison by The Daily Beast found the first one, showing the pair speaking to reporters outside the White House, appeared altered against a near-identical frame shot by Getty Images photographer Anna Moneymaker.
In the version Leavitt posted, her hair looked fuller, her cheekbones higher, her chin narrower, her waist slimmer, and her chest enlarged. Those edits are the kind of change X's system is built to catch.
How the Detection Works #
X does not read a hidden watermark inside the pixels. It scans the file's C2PA metadata, an industry provenance standard that logs which tool created or edited an image. Content Provenance and Authenticity credentials get written automatically by editors like Photoshop's Generative Fill, Google Photos' Magic Eraser, and several AI generators.
When X spots that signature at upload, it applies the label on its own. On automatic tags, the setting is greyed out, and there's no appeal button. That's why an edited photo of a real person can still get flagged as AI, even when nothing was invented from scratch.
Spotting the Flag Yourself #
The takeaway for everyday users is simple. A 'Made with AI' note on any post usually points to the image, not the caption. If you edit your own photos with common AI tools before posting, expect the same tag. Stripping the metadata before upload removes the trigger, though X can still flag visuals its classifier recognises.
Leavitt's label had disappeared by around 8 p.m. Eastern, roughly three hours after her post went live. The reason it appeared in the first place stayed on the record.
© Copyright IBTimes 2025. All rights reserved.