weeklyAI · Week of 2 October 2026weeklyAI · Semana del 2 de octubre de 2026

← Your vote← Su voto

survey · SEEU Review · la publicación, 1 jun 2026 · gratis

Un estudio de 2026 preguntó a 252 usuarios de internet en Macedonia del Norte cuánta desconfianza sienten ante lo que ven en internet.

Versión breve · la versión detallada sigue, unos 6 min

Pregunte al chat de weeklyAI

Pregúnteme por este estudio: a quiénes se estudió, qué encontró y qué no dice.

Las conversaciones se guardan mientras exista weeklyAI, para mejorar la publicación. Se responde en el idioma en que usted escribe.

El estudio, de un vistazo
Quiénes
usuarios activos de internet en Macedonia del Norte
Cuántos
252
Dónde
Macedonia del Norte
Cuándo
2026
Tipo de estudio
encuesta
Quién lo hizo
Universidad del Sudeste de Europa
El límite que importa
La muestra no fue al azar: 73.4% albaneses y 22.2% macedonios, sin nadie de 25 a 34 años.
Composición de la muestra por etnia
Albaneses73.4%
Macedonios22.2%

Porcentaje de los 252 encuestados que pertenecía a cada comunidad; la muestra no fue al azar, así que no representa a todo el país.

Temor a que la inteligencia artificial se use para incitar odio entre etnias

Comunidad albanesafrente aComunidad macedonia

La comunidad albanesa mostró mucha más preocupación (4.10) que la macedonia (2.80).

Ver · 0:46

QUÉ PASÓ Investigadores de la Universidad del Sudeste de Europa preguntaron en 2026 a 252 usuarios activos de internet en Macedonia del Norte sobre su confianza en la información en línea. Más del 70 por ciento dijeron dudar seguido de lo que ven. El grupo de 35 a 49 años fue el más desconfiado. Entre los albaneses de la muestra, el temor a que la inteligencia artificial se use para incitar odio entre etnias promedió 4.1 de 5, contra 2.9 entre los macedonios. La muestra no fue al azar: el 73.4 por ciento era albanés y faltó el grupo de 25 a 34 años.

QUÉ PUEDE VERIFICAR USTED MISMO El documento menciona una forma de verificar: comprobar en un medio tradicional, como la televisión o los periódicos, si la información es cierta.

QUÉ DICE LA AUTORIDAD El documento menciona la Ley de Inteligencia Artificial de la Unión Europea, aprobada en 2024, que prohíbe los sistemas que manipulan el comportamiento humano. El documento dice que en Macedonia del Norte hay esfuerzos para armonizar sus leyes con esa norma, pero que esos esfuerzos son difíciles por la rapidez con que cambia la tecnología.

Qué significa para usted

Ese estudio no le dice qué hacer con su voto, porque se hizo en Macedonia del Norte y solo midió lo que la gente decía sentir. Cuando llegue a la casilla, fíjese en lo que sí puede revisar: si la información que vio durante la campaña tiene respaldo en un medio tradicional, y qué auditaron y publicaron las autoridades electorales sobre el registro y el conteo.

Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013

Quién pagó: El artículo no dice quién financió el estudio ni si los financiadores tuvieron alguna influencia; solo señala que el cuestionario fue validado por un panel de expertos de la Universidad de Europa del Sudeste.

Los hallazgos de otros estudios que aquí se mencionan los conocemos por este documento, que fue el que leímos; no abrimos cada uno de esos estudios.

Versión detalladaLos pasajes copiados del artículo, las ilustraciones y cada fuente con cuánto leímos de ella · 1261 palabras · unos 6 minLeerla →Cerrar
Así podría verse · ilustración generada por weeklyAI.watch, no es una fotografía

La regla y la evidencia: un estudio de 2026 preguntó a 252 usuarios de internet en Macedonia del Norte cuánta desconfianza sienten ante lo que ven en internet

### EL ACTO, CITADO Y ENLAZADO

En el expediente que tenemos esta semana no aparece ninguna resolución, ningún fallo, ninguna auditoría ni ningún proyecto de ley que esté en vigor. Los autores del estudio que revisamos piden que el país armonice su legislación con la ley europea de inteligencia artificial, y dicen que prohibir los sistemas que manipulan la conducta humana o usan perfiles biométricos debería ser prioridad nacional1. Pero eso es una recomendación de los investigadores, no una norma vigente. No hay texto que diga quién la aplica, desde cuándo ni qué puede hacer un ciudadano al respecto.

### EL CASO QUE RESPONDE

Esta semana no se encontró ningún caso real de deepfake electoral, de red de bots ni de acta manipulada con nombre, lugar y fecha. Los estudios que sí revisamos no documentan instancias concretas, sino experimentos y encuestas: en un experimento en Kenia, siete mil personas vieron un video generado donde dos candidatos presidenciales hablaban de un supuesto plan de corrupción, acompañado de comentarios con distintos grados de escepticismo23. En otro caso, los comentarios que no señalaban el origen sintético del video redujeron el apoyo al político, y los comentarios escépticos lo restauraron en parte4. En Ecuador, una investigación sobre las elecciones presidenciales de 2025 encontró que los grupos rurales y los adultos mayores mostraban más dificultad para detectar desinformación, y que las mujeres recibían más violencia política digital a través de contenido manipulado56. Estos son los tipos de caso que la recomendación de los autores parece tener en mente, aunque el texto no nombra ninguno.

### LO QUE SE MIDIÓ

El estudio se hizo en Macedonia del Norte, con 252 usuarios de internet, y comparó lo que la gente respondía en 2026 con indicadores de 2015. No midió un efecto en el mundo real: midió lo que la gente dice que piensa y siente. Más del 72% de los encuestados dijo dudar con frecuencia de la veracidad de lo que ve en línea7. La preocupación por la manipulación algorítmica subió un 45% frente a hace una década8. La confianza en que las elecciones puedan ser justas en un espacio digital sin regulación cayó un 18%9. La sensación de seguridad sobre la información que reciben quedó en 2,10 sobre 510. La verificación de lo que la inteligencia artificial les dice sobre política quedó en 2,2 sobre 5: aunque desconfían, casi no verifican11. La comunidad albanesa mostró mucha más preocupación (4,10) que la macedonia (2,80) por el uso de estas herramientas para incitar odio entre grupos12.

El estudio no es una muestra aleatoria de todo el país: los autores lo dicen ellos mismos. La distribución fue desigual, con 73.4% albaneses y 22.2% macedonios, y sin nadie del grupo de 25 a 34 años13. Por eso sus resultados no se pueden generalizar a toda la población14. Además, el estudio es de un solo momento: no puede mostrar si usar inteligencia artificial causa la desconfianza o simplemente la acompaña.

Vale la pena explicar cómo funciona lo que se mide, paso a paso. Primero, una herramienta de lenguaje como las que todos conocemos produce texto a costo casi cero, y eso abre la puerta a desinformación en masa15. Segundo, la misma tecnología permite manipular audio y video con precisión quirúrgica, de modo que un video falso puede verse tan real como uno verdadero16. Tercero, cuando el público sabe que eso es posible, ya no solo duda de lo falso: también puede descartar lo verdadero, porque cualquier evidencia incriminatoria se puede negar diciendo que fue fabricada17. La amenaza principal dejó de ser el sesgo humano y pasó a ser la realidad sintética producida por máquinas18.

### LA RESPUESTA DE LA OTRA REGIÓN A LO MISMO

En Estados Unidos, un experimento con 4,293 votantes registrados probó artículos generados por inteligencia artificial para anticipar rumores electorales antes de que circularan; encontró que redujeron la creencia en esos rumores, con efectos que persistían, aunque más débiles, una semana después, y sin señales de rechazo partidista1920. El mismo trabajo encontró que los artículos escritos con ayuda humana no fueron más efectivos que los escritos solo por la máquina una vez que la instrucción estuvo lista21. En Kenia, el experimento mostró que los comentarios que rodean un video sintético cambian cómo la gente lo interpreta, y que las personas descartan más fácilmente los deepfakes que atacan a candidatos que apoyan422. En Ecuador, el estudio sobre las elecciones de 2025 encontró que la capacidad de identificar contenido generado por inteligencia artificial se relaciona con valorar la educación mediática como forma de protegerse23. Esto es lo que otras autoridades y equipos de investigación han probado sobre el mismo método.

### LA LECTURA DE LA BIBLIOTECA

Así lo leemos nosotros. La propaganda siempre ha buscado lo mismo: meter una versión de los hechos en la cabeza de alguien antes de que tenga tiempo de pensar. Lo nuevo no es el engaño, sino el canal. Antes el rumor llegaba por la plaza, la radio o el pasquín, y todos escuchaban más o menos lo mismo. Ahora llega por el mismo hilo donde uno recibe mensajes de la familia, y cada quien recibe una versión distinta. Eso cambia lo que un votante puede verificar solo. Puede preguntarle a un vecino si vio lo mismo. Puede buscar si un medio serio lo confirmó. Lo que ya no puede hacer solo es saber si lo que le llegó a él le llegó también a los demás, porque quizá no.

Lo que esto nos lleva a esperar en casas como la suya es una cosa concreta: que la duda se vuelva costumbre, y que la costumbre de dudar se confunda con la costumbre de no creer en nada. Cuando eso pasa, el votante deja de pedir pruebas y se conforma con la sensación de que algo pasó. Nos equivocaríamos si los votantes que más dudan de lo que ven en línea siguieran confiando igual en su autoridad electoral y en los medios de siempre, y si esa duda no se estuviera usando para justificar que se salten controles. Usted puede hacer algo con esto: cuando alguien le diga que ya no se puede creer en nada, pregúntele qué institución o qué medio le merece todavía confianza y por qué. Y antes de reenviar un video o un audio de campaña, pregúntese quién lo produjo y si alguna autoridad electoral o un medio serio ya lo verificó. Si nadie lo ha hecho, espere.

### LO QUE NO SE SABE

No se sabe si el miedo que la gente reporta se traduce en votos distintos, porque el estudio midió percepciones, no resultados electorales. No se sabe si la desconfianza que muestra la encuesta cambia el comportamiento real en la casilla, porque nadie observó eso. No se sabe si el patrón que aparece en Macedonia del Norte se repite en otros países, porque el estudio se hizo solo allí. No se sabe cuánto de la duda viene del uso de estas herramientas y cuánto de otras causas, porque el diseño no permite separarlas. No se sabe qué pasaría si se aprobara la ley que los autores piden, porque nadie la ha probado. Y no se sabe si la gente que dice verificar poco lo hace de verdad o solo responde así en una encuesta.

Lo único que usted puede hacer con lo que hoy sabemos es preguntar, antes de creer o reenviar: ¿quién produjo esto, y quién más lo ha confirmado?

De dónde sale cada dato de contexto, y cuánto leímos de cada documento

  1. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Harmonization with the EU AI Act is not just an integration criterion, but a critical policy priority to protect public opinion. Banning AI systems that manipulate human behavior or use biometric profiling should be a national priority."
  2. Wack M, Prochaska S. (2026). Making Sense of AI-Generated Disinformation: How Audience Interpretations Influence the Impact of Deepfakes in Kenya. Social Media + Society. 10.1177/20563051261462092 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "To answer this question, we rely on a survey experiment based in Kenya ( N = 7,000)."
  3. Wack M, Prochaska S. (2026). Making Sense of AI-Generated Disinformation: How Audience Interpretations Influence the Impact of Deepfakes in Kenya. Social Media + Society. 10.1177/20563051261462092 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "The experiment saw respondents view a generated clip of presidential candidates discussing their role in a fabricated corruption scheme coupled with embedded comments expressing varied levels of skepticism."
  4. Wack M, Prochaska S. (2026). Making Sense of AI-Generated Disinformation: How Audience Interpretations Influence the Impact of Deepfakes in Kenya. Social Media + Society. 10.1177/20563051261462092 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Using this instrument, we find that comments which failed to remark on the synthetic origins of the video reduced support for the politician, while skeptical comments partially restored it."
  5. Suing A, Ganazhapa H, Medina J. (2026). Media and information literacy in the face of election disinformation generated by artificial intelligence: experiences from Ecuador. Frontiers in Political Science. 10.3389/fpos.2026.1834093 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "The aim of this research is to analyse how MIL levels among vulnerable population groups in Ecuador condition their exposure to, and responses to, AI-generated electoral disinformation, with reference to the 2025 presidential elections."
  6. Suing A, Ganazhapa H, Medina J. (2026). Media and information literacy in the face of election disinformation generated by artificial intelligence: experiences from Ecuador. Frontiers in Political Science. 10.3389/fpos.2026.1834093 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Rural groups and older adults exhibited greater limitations in detecting disinformation, whilst women were found to be subject to increasing digital political violence of a gendered nature through manipulated content."
  7. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - el artículo del que trata esta nota — el artículo completo — el pasaje: "The results show an exponential increase in automated disinformation, with over 72% of the surveyed population reporting frequent doubts about the veracity of online information."
  8. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - el artículo del que trata esta nota — el artículo completo — el pasaje: "As illustrated in Table 4, concerns regarding manipulation have increased by 45% over the past decade."
  9. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - el artículo del que trata esta nota — el artículo completo — el pasaje: "This heightened state of vulnerability is paired with an 18% decrease in the belief that elections can remain fair in a digitally unregulated space."
  10. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - el artículo del que trata esta nota — el artículo completo — el pasaje: "When perceived informational security falls to critical levels (M=2.10 in Table 4), the public is left unprotected. This vacuum is filled not by the truth, but by synthetic narratives that fit defined beliefs, making the democratic process not a contest of ideas, but a contest of algorithms."
  11. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Although suspicion is high, verification remains low (M=2.2). This suggests that many citizens in North Macedonia may experience a form of “information isolation.”"
  12. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - el artículo del que trata esta nota — el artículo completo — el pasaje: "The finding that the Albanian community has a significantly greater concern (M=4.10) about the use of AI for interethnic hate speech, compared to the Macedonian community (M=2.80), reflects a historical and informational vulnerability."
  13. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - el artículo del que trata esta nota — el artículo completo — el pasaje: "The use of purposive sampling resulted in an uneven demographic distribution, specifically regarding ethnicity (73.4% Albanian vs. 22.2% Macedonian) and age (an absence of the 25-34 demographic cohort in the final dataset)."
  14. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Consequently, these results cannot be entirely generalized to the macro-population of North Macedonia. Instead, this study should be viewed as an exploratory analysis reflecting the specific digital, academic, and regional clusters sampled, offering valuable directional insights into how distinct sub-communities process algorithmic vulnerabilities."
  15. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Generative Artificial Intelligence (GenAI): Large language models (LLMs) like ChatGPT and Gemini have democratized content production while simultaneously paving the way for mass disinformation at a near-zero cost."
  16. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Deepfakes and Multimodal Manipulation: Beyond text, modern technologies allow audio and video manipulation with surgical precision."
  17. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - el artículo del que trata esta nota — el artículo completo — el pasaje: "In an environment where the public knows that AI can produce incriminating videos or audios that look real, corrupt politicians can dismiss real evidence as “AI-produced”."
  18. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - el artículo del que trata esta nota — el artículo completo — el pasaje: "The primary threat has transitioned from human-generated political bias to machine-generated synthetic reality."
  19. Linegar M, Sinclair B, van der Linden S, Alvarez RM. (2026). Towards scalable AI-assisted pre-bunking of election misinformation: evidence from a pre-registered US panel experiment. Royal Society Open Science. 10.1098/rsos.252226 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "In a pre-registered two-wave experiment with 4293 United States (US) registered voters, we test this framework against politically charged election misinformation—one of the most challenging domains for misinformation intervention."
  20. Linegar M, Sinclair B, van der Linden S, Alvarez RM. (2026). Towards scalable AI-assisted pre-bunking of election misinformation: evidence from a pre-registered US panel experiment. Royal Society Open Science. 10.1098/rsos.252226 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "We find that large language model (LLM)-generated pre-bunking significantly reduced belief in election rumours (effects persisting, though attenuated, one week later) and modestly offset declines in confidence in national election administration, with no evidence of partisan backlash."
  21. Linegar M, Sinclair B, van der Linden S, Alvarez RM. (2026). Towards scalable AI-assisted pre-bunking of election misinformation: evidence from a pre-registered US panel experiment. Royal Society Open Science. 10.1098/rsos.252226 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "After finalizing the prompt, articles written with human feedback were no more effective than articles using only AI, indicating that per-rumour human effort can be substantially reduced once the prompt is in place."
  22. Wack M, Prochaska S. (2026). Making Sense of AI-Generated Disinformation: How Audience Interpretations Influence the Impact of Deepfakes in Kenya. Social Media + Society. 10.1177/20563051261462092 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Partisanship also mattered, as respondents more readily dismissed deepfakes targeting candidates they supported compared to opposing candidates."
  23. Suing A, Ganazhapa H, Medina J. (2026). Media and information literacy in the face of election disinformation generated by artificial intelligence: experiences from Ecuador. Frontiers in Political Science. 10.3389/fpos.2026.1834093 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "A significant correlation was confirmed between the capacity to identify AI-generated content and the valuation of MIL as a mitigation mechanism."

Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013

Quién pagó: El artículo no dice quién financió el estudio ni si los financiadores tuvieron alguna influencia; solo señala que el cuestionario fue validado por un panel de expertos de la Universidad de Europa del Sudeste.

survey · SEEU Review · the paper, 1 Jun 2026 · free

Can I trust what I see before I vote? Here is what one 2026 survey in North Macedonia found, and what it cannot tell you.

Short version · the longer version follows, about 6 min

Ask the weeklyAI chat

Ask me about this study: who was studied, what it found, and what it does not say.

Conversations are saved for as long as weeklyAI exists, to improve the publication. Answers come in the language you write in.

The study at a glance
Who
active internet users in North Macedonia
How many
252
Where
North Macedonia
When
2026
Kind of study
survey
Who did it
South East European University
The limit that matters
Sample uneven by community and age; results cannot be stretched to the whole country.
Communities in the survey sample
Albanian73.4%
Macedonian22.2%

Shares of the 252 people surveyed, as the study reports them; the sample was uneven and the authors say it cannot stand for the whole country.

Fear that AI could stir hatred between communities, by community

Albanian communityagainstMacedonian community

Albanian community reported far more fear, 4.10 out of 5, than the Macedonian community, 2.80.

Here is the number that should stop a reader: asked whether they would check an AI tool's claim about a political event against television or newspapers, the people surveyed answered 2.2 out of 5 — high doubt, low checking.
weeklyAI's reading
Watch · 0:45

In 2026, two researchers at South East European University surveyed 252 internet users in North Macedonia about AI and public opinion. The survey was digital, with 13 questions, and it compared answers against a 2015 baseline. More than 72% said they often doubt what they read online. People aged 35 to 49 were more skeptical than younger ones. Albanian respondents feared AI-driven ethnic hate speech more than Macedonian ones. Fear of AI-manipulated elections rose 45% since 2015, and confidence in fair elections under AI fell. The authors say the sample was uneven by ethnicity and age, so the results cannot be entirely generalized to the whole population of North Macedonia.

The article does not say whether the survey names any way for a voter to check a video or a count.

The authors say efforts to harmonize North Macedonia's laws with the EU AI Act remain challenging, and they call for that harmonization. The authors urge that step; no local ruling is reported.

What this means for you

For your own ballot, the survey offers no checklist: it did not test any way to verify a video or a count. Watch instead for what your electoral authority publishes, what it audited, and what observers found. A survey of 252 people in North Macedonia cannot tell you what machines are doing to your vote.

Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013

Who paid: The article does not say who funded the study or whether funders had any say; it only notes that the questionnaire was validated by a panel of experts at South East European University.

The findings of other studies mentioned here are known to us through this document, which is the one we read; we did not open each of those studies.

The longer versionThe passages copied from the paper, the pictures, and every source with how much of it we read · 1209 words · about 6 minRead it →Close
How it could look · illustration generated by weeklyAI.watch, not a photograph

The rule and the evidence: Can I trust what I see before I vote?

### THE ACT

No resolution, ruling, audit or bill on artificial intelligence and elections in North Macedonia is in force in the passages we read this week; what the record carries instead is a recommendation, and it is a recommendation the study's authors address to their own country, not a rule anyone must obey12.

The recommendation asks that North Macedonia bring its law into line with the European Union's Artificial Intelligence Act, and it names two things as national priorities: banning AI systems that manipulate human behavior, and banning those that use biometric profiling1.

The study also proposes what it calls a new digital social contract, resting on three things: that the workings of algorithms be made transparent, that a person's own mind be protected, and that traditional media be strengthened as the place where information is verified2.

Nothing in these passages states who would enforce such rules, from what date they would apply, or what a single voter could do and by when; the passages describe a policy priority, not an enacted law12.

### THE CASE IT ANSWERS

No real instance of a deepfake, an avatar, a bot network or a manipulated tally form was found this week with a name, a place, a date and a second source; that part of the record is short, and we say so plainly.

What the passages do carry is a description of the kind of thing the recommendation is meant for: tools that generate text at almost no cost3, and tools that alter audio and video with what the study calls surgical precision4.

It would seem this is the kind of thing they mean: the study says such technology can undermine the idea that seeing is believing, leaving public opinion open to campaigns meant to destabilize elections5.

It would seem this is the kind of thing they mean: the study describes a shift in the main threat from bias made by people to a synthetic reality made by machines6.

### WHAT WAS MEASURED

The measurement comes from a survey of 252 active internet users in North Macedonia, reached in 2026, with 52 percent women and 48 percent men, and by community, 73.4 percent Albanian, 22.2 percent Macedonian and 4.4 percent from other communities7.

More than 72 in every 100 of those surveyed said they often doubt whether what they read online is true8.

Fear of manipulation by algorithm stood at 3.92 out of 5, which the study reports as a rise of 45 percent against its 2015 starting point9; belief that information is truthful stood at 2.10, a fall the study puts at 12 percent; confidence that elections can stay fair under AI stood at 2.45, a fall the study puts at 18 percent10.

The people aged 35 to 49 showed the most doubt of any age group, at 4.08 out of 5, with the 18-to-24 group at 3.86 and the over-50 group at 3.20.

The Albanian community reported far more fear that AI could be used to stir hatred between communities, at 4.10 out of 5, than the Macedonian community, at 2.8011.

Here is the number that should stop a reader: asked whether they would check an AI tool's claim about a political event against television or newspapers, the people surveyed answered 2.2 out of 5 — high doubt, low checking12.

The study's own authors say the sample was uneven by community and age, that the 25-to-34 group was missing entirely, and that the results cannot be stretched to the whole country713.

We found no passage measuring this in an actual election rather than a survey; every number above comes from what people said, not from what happened at a polling place.

### THE OTHER REGION'S ANSWER TO THE SAME THING

A survey experiment in Kenya put a made-up clip of presidential candidates discussing a invented corruption scheme in front of 7,000 respondents, with comments underneath expressing different levels of doubt1415.

That study found that comments which never mentioned the video was synthetic lowered support for the politician shown, while skeptical comments partly brought that support back16, and that people more readily dismissed a fake aimed at a candidate they supported than one aimed at a candidate they opposed17.

A separate experiment with 4,293 registered voters in the United States tested articles written to warn people in advance about election rumors, and found belief in those rumors fell, with the effect weaker but still present a week later, and no sign of backlash along party lines1819.

A study in Ecuador, looking at the 2025 presidential elections, surveyed 405 people, interviewed 12 experts and held six group discussions, and found that rural residents and older adults had the most trouble detecting false content, while women faced growing political violence through manipulated material202122.

### THE LIBRARY'S READING

Here is how we read it. A false story about a candidate has always done one thing well: it leaves a mark that no correction fully removes. What is new is not the lie but the speed and the cost. A rumor once needed a printer, a crowd or a party. Now it needs a prompt and a few seconds. What a public can learn to check is the source, the date, the second account. What no one can check alone is whether a voice is real. So the checking moves from the eye to the institution — and that is where it gets hard, because the institution is exactly what the doubt has already reached.

Here is what that leads us to expect in homes like yours. Where most people already doubt what they see and few bother to confirm it, the doubt does not turn into carefulness. It turns into withdrawal: a real document and a fabricated one start to look equally arguable, and the official record of a vote loses the power to settle anything. You would know we are wrong if the people around you who distrust their feeds still treat an audit or a ruling as final. Watch for that. And when someone waves away real evidence as probably AI, ask which check would settle it — and whether anyone actually ran it. Ask your electoral authority what it published and found, and read that document rather than a clip about it.

### WHAT IS NOT KNOWN

The passages do not tell us whether AI caused the doubt measured, since a survey taken at one moment cannot show what produced what13.

They do not let us speak for all voters in North Macedonia, because the authors themselves say the sample cannot be stretched that far13.

They do not tell us whether any law, program or rule worked, because no such thing was tested1.

They do not tell us what a deepfake did to a single voter in a single election, because no such instance was found this week with a name, a place and a date.

So the honest position for a voter is this: the fear is measured, the rule is not yet written, and the one thing you can do before the date fixed by law is decide now which document you will trust when a clip tells you the count is wrong.

Where each piece of context comes from, and how much of it we read

  1. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - the article this story is about — the whole article — the passage: "Harmonization with the EU AI Act is not just an integration criterion, but a critical policy priority to protect public opinion. Banning AI systems that manipulate human behavior or use biometric profiling should be a national priority."
  2. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - the article this story is about — the whole article — the passage: "The research proposes that North Macedonia initiate a “new digital social contract.” This contract should be based on algorithmic transparency, protection of the cognitive sovereignty of the individual, and strengthening traditional media as verification authorities."
  3. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - the article this story is about — the whole article — the passage: "Generative Artificial Intelligence (GenAI): Large language models (LLMs) like ChatGPT and Gemini have democratized content production while simultaneously paving the way for mass disinformation at a near-zero cost."
  4. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - the article this story is about — the whole article — the passage: "Deepfakes and Multimodal Manipulation: Beyond text, modern technologies allow audio and video manipulation with surgical precision."
  5. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - the article this story is about — the whole article — the passage: "This technology, according to DiResta (2024), has the power to undermine the concept of "visual evidence", making public opinion vulnerable to sophisticated campaigns for influence (influence operations) aimed at destabilizing electoral processes."
  6. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - the article this story is about — the whole article — the passage: "The primary threat has transitioned from human-generated political bias to machine-generated synthetic reality."
  7. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - the article this story is about — the whole article — the passage: "The use of purposive sampling resulted in an uneven demographic distribution, specifically regarding ethnicity (73.4% Albanian vs. 22.2% Macedonian) and age (an absence of the 25-34 demographic cohort in the final dataset)."
  8. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - the article this story is about — the whole article — the passage: "The results show an exponential increase in automated disinformation, with over 72% of the surveyed population reporting frequent doubts about the veracity of online information."
  9. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - the article this story is about — the whole article — the passage: "As illustrated in Table 4, concerns regarding manipulation have increased by 45% over the past decade."
  10. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - the article this story is about — the whole article — the passage: "This heightened state of vulnerability is paired with an 18% decrease in the belief that elections can remain fair in a digitally unregulated space."
  11. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - the article this story is about — the whole article — the passage: "The finding that the Albanian community has a significantly greater concern (M=4.10) about the use of AI for interethnic hate speech, compared to the Macedonian community (M=2.80), reflects a historical and informational vulnerability."
  12. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - the article this story is about — the whole article — the passage: "Although suspicion is high, verification remains low (M=2.2). This suggests that many citizens in North Macedonia may experience a form of “information isolation.”"
  13. Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013 - the article this story is about — the whole article — the passage: "Consequently, these results cannot be entirely generalized to the macro-population of North Macedonia. Instead, this study should be viewed as an exploratory analysis reflecting the specific digital, academic, and regional clusters sampled, offering valuable directional insights into how distinct sub-communities process algorithmic vulnerabilities."
  14. Wack M, Prochaska S. (2026). Making Sense of AI-Generated Disinformation: How Audience Interpretations Influence the Impact of Deepfakes in Kenya. Social Media + Society. 10.1177/20563051261462092 — only the abstract - the full text could not be fetched — the passage: "To answer this question, we rely on a survey experiment based in Kenya ( N = 7,000)."
  15. Wack M, Prochaska S. (2026). Making Sense of AI-Generated Disinformation: How Audience Interpretations Influence the Impact of Deepfakes in Kenya. Social Media + Society. 10.1177/20563051261462092 — only the abstract - the full text could not be fetched — the passage: "The experiment saw respondents view a generated clip of presidential candidates discussing their role in a fabricated corruption scheme coupled with embedded comments expressing varied levels of skepticism."
  16. Wack M, Prochaska S. (2026). Making Sense of AI-Generated Disinformation: How Audience Interpretations Influence the Impact of Deepfakes in Kenya. Social Media + Society. 10.1177/20563051261462092 — only the abstract - the full text could not be fetched — the passage: "Using this instrument, we find that comments which failed to remark on the synthetic origins of the video reduced support for the politician, while skeptical comments partially restored it."
  17. Wack M, Prochaska S. (2026). Making Sense of AI-Generated Disinformation: How Audience Interpretations Influence the Impact of Deepfakes in Kenya. Social Media + Society. 10.1177/20563051261462092 — only the abstract - the full text could not be fetched — the passage: "Partisanship also mattered, as respondents more readily dismissed deepfakes targeting candidates they supported compared to opposing candidates."
  18. Linegar M, Sinclair B, van der Linden S, Alvarez RM. (2026). Towards scalable AI-assisted pre-bunking of election misinformation: evidence from a pre-registered US panel experiment. Royal Society Open Science. 10.1098/rsos.252226 — only the abstract - the full text could not be fetched — the passage: "In a pre-registered two-wave experiment with 4293 United States (US) registered voters, we test this framework against politically charged election misinformation—one of the most challenging domains for misinformation intervention."
  19. Linegar M, Sinclair B, van der Linden S, Alvarez RM. (2026). Towards scalable AI-assisted pre-bunking of election misinformation: evidence from a pre-registered US panel experiment. Royal Society Open Science. 10.1098/rsos.252226 — only the abstract - the full text could not be fetched — the passage: "We find that large language model (LLM)-generated pre-bunking significantly reduced belief in election rumours (effects persisting, though attenuated, one week later) and modestly offset declines in confidence in national election administration, with no evidence of partisan backlash."
  20. Suing A, Ganazhapa H, Medina J. (2026). Media and information literacy in the face of election disinformation generated by artificial intelligence: experiences from Ecuador. Frontiers in Political Science. 10.3389/fpos.2026.1834093 — only the abstract - the full text could not be fetched — the passage: "The aim of this research is to analyse how MIL levels among vulnerable population groups in Ecuador condition their exposure to, and responses to, AI-generated electoral disinformation, with reference to the 2025 presidential elections."
  21. Suing A, Ganazhapa H, Medina J. (2026). Media and information literacy in the face of election disinformation generated by artificial intelligence: experiences from Ecuador. Frontiers in Political Science. 10.3389/fpos.2026.1834093 — only the abstract - the full text could not be fetched — the passage: "A mixed-methods approach combining quantitative and qualitative methods was employed, with a descriptive and explanatory scope, executed through a survey of 405 participants, 12 semi-structured interviews with experts, and six focus group discussions with 35 participants representing heterogeneous profiles."
  22. Suing A, Ganazhapa H, Medina J. (2026). Media and information literacy in the face of election disinformation generated by artificial intelligence: experiences from Ecuador. Frontiers in Political Science. 10.3389/fpos.2026.1834093 — only the abstract - the full text could not be fetched — the passage: "Rural groups and older adults exhibited greater limitations in detecting disinformation, whilst women were found to be subject to increasing digital political violence of a gendered nature through manipulated content."

Dika, Z., Bajrami, D. (2026). Transforming Public Opinion in the Era of Artificial Intelligence: A Comparative Analysis in North Macedonia (2015–2026). SEEU Review. https://doi.org/10.2478/seeur-2026-0013

Who paid: The article does not say who funded the study or whether funders had any say; it only notes that the questionnaire was validated by a panel of experts at South East European University.

The findings of other studies mentioned here are known to us through this document, which is the one we read; we did not open each of those studies.