Someone had taken a real photograph and used AI to insert a fictional message
Days before Peru's 2026 presidential runoff, two fabricated images purporting to show opposing political sentiments in the city of Jaén circulated widely on Facebook, each appearing to document real grassroots activity around candidate Keiko Fujimori. Verified by La República's fact-checking unit using AI detection tools, both images exceeded 90 percent probability of synthetic origin — their contradictory nature suggesting not a partisan agenda, but a deliberate seeding of confusion. In the final hours before a democratic decision, the boundary between evidence and invention had been quietly erased.
- Two mutually contradictory images — one for Fujimori, one against — spread simultaneously on Facebook, accumulating thousands of reactions before anyone questioned their authenticity.
- Physical impossibilities embedded in the images, including a floating campaign sign with no support structure and a face that didn't match Fujimori's, betrayed the hand of an AI that does not understand gravity or likeness.
- Hive Moderation's detection tool confirmed both images as synthetic with over 90 percent certainty, anchoring suspicion in measurable, technical fact.
- Investigators traced the real plaza in Jaén through Google Maps imagery, confirming that an authentic photograph had been used as a canvas onto which fictional political messages were digitally painted.
- The dual fabrication — supporting and opposing the same candidate — points less to propaganda than to deliberate epistemic sabotage, designed to make voters distrust what they see.
- With the runoff imminent and authorship still unknown, the incident leaves open the unsettling question of whether this was an isolated act or the visible edge of a coordinated disinformation campaign.
Days before Peru's 2026 presidential runoff, two images began spreading on Facebook showing campaign signs installed in Jaén's central plaza — one declaring the city would vote for Keiko Fujimori, the other declaring the opposite. The first gathered over 2,100 reactions and 142 shares. Both looked like documentation of real political life. Neither was.
Verificador de La República subjected the images to technical analysis and found them entirely synthetic. The face presented as Fujimori didn't match her actual appearance, and the campaign sign appeared to float without any visible support structure — errors that emerge when AI generates images without understanding how physical objects occupy space.
Running both through Hive Moderation's AI detection tool confirmed what the visual inconsistencies suggested: the pro-Fujimori image scored 90.3 percent probability of AI generation; the anti-Fujimori image, 91.2 percent. The verdict was unambiguous.
What gave the fabrications their initial credibility was their grounding in reality. The plaza backgrounds matched an actual location in Jaén, verified against Google Maps imagery from June 2024 — the same pharmacy storefront, the same floral mural, the same cathedral visible in the distance. Someone had taken a real photograph and used AI to insert a fictional political message into it.
The contradictory nature of the two images is telling. Rather than advancing a single narrative, the fabrications appeared designed to sow confusion — flooding the information space with opposing "evidence" of public opinion so that no version of reality could be trusted. In an election's final days, that manufactured uncertainty is itself a form of influence. Who created the images, and whether they were part of a broader effort to manipulate Peruvian voters, remained unanswered.
Two images began circulating on Facebook in early June 2026, just days before Peru's presidential runoff election. Both appeared to show campaign signs installed in Jaén's central plaza—one declaring "In Jaén we will vote for Keiko," the other its exact opposite: "In Jaén we will not vote for Keiko." The first image accumulated 2,100 reactions, nearly 1,000 comments, and was shared 142 times. They looked like evidence of real political activity in a real place. They were not.
Verificador de La República, the newspaper's fact-checking unit, subjected both images to technical scrutiny and found them to be entirely synthetic. The face presented as Keiko Fujimori, the Fuerza Popular candidate, did not match her actual appearance. The campaign sign itself appeared to float in midair with no visible support structure—no rope, no frame, nothing holding it in place. These were not subtle errors. They were the kind of mistakes that emerge when artificial intelligence generates visual content without understanding the physical laws that govern how objects actually exist in space.
To confirm the suspicion, the verification team ran both images through Hive Moderation's AI-Generated Content Detection tool, a system designed to identify patterns in faces, textures, and other visual elements that signal synthetic creation. The results were unambiguous. The pro-Fujimori image showed a 90.3 percent probability of having been generated by AI. The anti-Fujimori image scored 91.2 percent. Both crossed well above the threshold of certainty.
What made the fabrication particularly sophisticated was its foundation in reality. The background of both images—the plaza itself, the surrounding buildings, the cathedral visible in the distance—all corresponded to an actual location in Jaén, a city in the Cajamarca region. Verificador confirmed this by comparing the viral images to Google Maps satellite imagery captured in June 2024 from the intersection of Bolívar Street and Mariscal Castilla Avenue. The same pharmacy storefront with its red sign appeared in both. The same mural of a woman's face surrounded by flowers was visible. The same cathedral, officially named the Parroquia Señor de Huamantanga, stood in the background. Someone had taken a real photograph of a real place and used artificial intelligence to insert a fictional political message into it.
The timing was not accidental. Peru's 2026 presidential election was entering its final stretch. The second round of voting was imminent. In this context, images purporting to show grassroots political activity—whether support or opposition—carry weight. They suggest momentum, organization, public sentiment. They are the kind of content that spreads because people want to believe them, because they confirm existing preferences or anxieties about which way the country is leaning.
The dual nature of the fabrication—one image supporting Fujimori, one opposing her—suggests an attempt to seed confusion rather than push a single narrative. By flooding social media with contradictory "evidence" of public opinion, the creator could muddy the waters around what people actually think, what is real versus what is manufactured. In an election season, that uncertainty itself becomes a form of influence.
Verificador's conclusion was direct: both images were false. They were created with artificial intelligence. The technical analysis left no room for ambiguity. What remained unclear was who created them, why, and whether this was an isolated incident or part of a broader campaign to manipulate Peruvian voters during a critical moment in the country's democratic process.
Notable Quotes
Both images were generated with artificial intelligence, with probability exceeding 90 percent— Verificador de La República analysis
The Hearth Conversation Another angle on the story
Why would someone go to the trouble of creating two contradictory images instead of just pushing one message?
Because contradiction itself is useful. If you flood the zone with conflicting "evidence," people stop trusting any of it. You don't need to convince someone to vote a certain way—you just need to make them doubt what's real.
But wouldn't people notice the floating sign, the wrong face?
Most people don't look closely. They see an image that confirms what they already suspect about their country, they react, they share. The technical flaws only matter if someone actually examines it carefully—which most people won't.
How did they get the background so right?
That's the clever part. They didn't create the background at all. They took a real photograph of the plaza and used AI to insert the sign and the face into it. It's easier to fool people when you're building on something true.
What does this tell us about elections in 2026?
That the information environment is now a battleground where the line between real and synthetic is deliberately blurred. A fact-checker can debunk these images, but by then they've already shaped how thousands of people think about what's happening on the ground.
Is there any way to know who made them?
Not from the images themselves. The AI tools can tell you something is synthetic, but they can't tell you who pressed the button. That requires investigation—tracking where the images first appeared, who shared them, what accounts amplified them. That's different work entirely.