Journalists’ Perceptions of Artificial Intelligence and Disinformation Risks

Academic Edi o s: Ma hias Felipe De- Lima-San os and Adeola Abdula ee Elega Recei ed: 30 June 2025 Re ised: 13 Augus 2025 Accep ed: 14 Augus 2025 Published: 30 Augus 2025 Ci a ion: Peña-Alonso, U., Peña-Fe nández, S., & Meso-Aye di, K. (2025). Jou nalis s’ Pe cep ions o A i icial In elligence and Disin o ma ion Risks. Jou nalism and Media,6(3), 133. h ps://doi.o g/ 10.3390/jou nalmedia6030133 Copy igh : © 2025 by he au ho s. Licensee MDPI, Basel, Swi ze land. This a icle is an open access a icle dis ibu ed unde he e ms and condi ions o he C ea i e Commons A ibu ion (CC BY) license (h ps://c ea i ecommons.o g/ licenses/by/4.0/). A icle Jou nalis s’ Pe cep ions o A i icial In elligence and Disin o ma ion Risks U ko Peña-Alonso *, Simón Peña-Fe nández and Koldobika Meso-Aye di Depa men o Jou nalism, Uni e si y o he Basque Coun y (UPV/EHU), 48940 Leioa, Spain; [email p o ec ed] (S.P.-F.); [email p o ec ed] (K.M.-A.) *Co espondence: [email p o ec ed] Abs ac This s udy examines jou nalis s’ pe cep ions o he impac o a i icial in elligence (AI) on disin o ma ion, a g owing conce n in jou nalism due o he apid expansion o gene a i e AI and i s in luence on news p oduc ion and media o ganiza ions. Using a quan i a i e app oach, a s uc u ed su ey was adminis e ed o 504 jou nalis s in he Basque Coun y, iden i ied h ough o icial media di ec o ies and wi h he suppo o he Basque Associa ion o Jou nalis s. This su ey, conduc ed online and ia elephone be ween May and June 2024, included ques ions on sociodemog aphic and p o essional a iables, as well as a i udes owa d AI’s impac on jou nalism. The esul s indica e ha a la ge majo i y o jou nalis s (89.88%) belie e AI will conside ably o signi ican ly inc ease he isks o disin o ma ion, and his pe cep ion is consis en ac oss gende s and media ypes, bu mo e p onounced among hose wi h g ea e p o essional expe ience. S a is ical analyses e eal a signi ican associa ion be ween yea s o expe ience and pe cei ed isk, and be ween AI use and isk pe cep ion. The main isks iden i ied a e he di icul y in de ec ing alse con en and deep akes, and he isk o ob aining inaccu a e o e oneous da a. Co-occu ence analysis shows ha hese isks a e o en pe cei ed as in e connec ed. These indings highligh he complex and mul i ace ed conce ns o jou nalis s ega ding AI’s ole in he in o ma ion ecosys em. Keywo ds: a i icial in elligence; isk pe cep ion; jou nalism; disin o ma ion; news media 1. In oduc ion In he pas decade, disin o ma ion has eme ged as one o he main conce ns in he ield o jou nalism (Cea & Palomo,2021). This phenomenon has gained a ele ance compa able o he exponen ial inc ease in he s a egic dimension o jou nalis ic p ac ice, d i en by he expansion o a i icial in elligence (Palomo e al.,2023;Ve ma,2024;Ke in-Ale echi e al.,2025). Thus, he ad ances in a i icial in elligence (AI) ha e su p ised us in ecen yea s due o hei speed and hei impac on almos e e y ield. Gene a i e AI, which ma ked a u ning poin in he de elopmen o his echnology, is pa icula ly ele an in con en c ea ion spaces (Gonzalo,2024;Newman e al.,2024). In he ield o jou nalism, i is inc eas- ingly e iden ha AI is beginning o make in oads ans e sally in he news p oduc ion p ocess, as well as in he s uc u e and unc ioning o media o ganiza ions (Túñez López e al.,2021;U a e Ruiz e al.,2023;Londoño-P oaño & Buele,2025). The p edic ions o hose who o esaw ha AI would ul ima ely de e mine he u u e de elopmen o media ans o ma ion seem o be coming ue (Zheng e al.,2018;Cui,2025). Jou nal. Media 2025,6, 133 h ps://doi.o g/10.3390/jou nalmedia6030133 Jou nal. Media 2025,6, 133 2 o 15 Peña-Fe nández e al. (2023) s a e ha , al hough AI b ings bene i s o jou nalism, i s adop ion en ails e hical challenges ha mus be p oac i ely add essed. I s eme gence has ekindled he deba e ega ding he ambi alen na u e o echnological de elopmen s (B ennen e al.,2018), ha is, he need o assess he balance be ween hei po en ial bene i s and he new insecu i ies hey in oduce (V abiˇc Dežman,2024;Wei e al.,2024;Schae z e al.,2025). I is e iden ha he use o a i icial in elligence and i s applica ion o jou nalism e eal signi ican oppo uni ies o imp o emen in he e iciency o news p oduc ion and dis ibu ion, bu also he h ea s posed by i s use o unde mine quali y jou nalism (Ma eos Aba ca & Gamonal A oyo,2024) o o dissemina e disin o ma ion on a la ge scale (As obiza,2024). These challenges a e compounded by conce ns ega ding he e aci y o con en , e hics and anspa ency in da a usage, in usion in o p i a e li e, and i s con ibu ion o ein o cing exis ing social di ides (Hansen e al.,2017;B undage e al., 2018;Pihlaja inne & Alén-Sa ikko,2022). Media p o essionals hemsel es a e no immune o his deba e. On he one hand, AI can assis jou nalis s by eeing hem om he mos ou ine and low alue-added asks (Wu e al.,2019;Cools & Diakopoulos,2024;Cools & de V eese,2025;Ga cía de To es e al.,2025), which could help s eng hen he mo e cogni i e aspec s o he p o ession, such as c ea i i y, c i ical hinking, cu iosi y, o skep icism (Thu man e al.,2017;Guzman & Lewis,2020;Mo a ec e al.,2024). Fo media o ganiza ions, AI may se e as he key o accessing business models ha imp o e hei economic iabili y (Lindén & Tuulonen,2019; Ko enidis & Veglis,2021;Oh & Jung,2025) o help hem educe cos s (Wilczek e al.,2024). Howe e , he main ese a ions ega ding he implemen a ion o AI come om he wo ke s hemsel es, who see i mo e as a h ea han as a complemen (Mond ía Te ol,2023; an Dalen,2024). In ac , some s udies conclude ha media manage s adop au oma ion as a s a egy aimed p ima ily a educing cos s and imp o ing p oduc i i y, esul ing in a educ ion in he numbe o employees (Kim & Kim,2017;Rick & Hani zsch,2023; Thäsle -Ko donou i & Ba ling,2023) o an inc ease in job insecu i y (López Jiménez & Oua iachi,2020;Ananny & Ka ,2025). In any case, he aim is o make he jou nalis ’s wo k mo e e icien (Papadimi iou, 2016;Dodds e al.,2025), which makes i s implemen a ion in he media essen ial (Au o , 2015;Ros amian & Mo adi,2024). A he same ime, he adop ion o AI demands new p o essional p o iles (Six o Ga cía e al.,2021;Mølle e al.,2024), which equi e no only lea ning o use ce ain ools (Mu cia Ve dú & La a Ramos,2024), bu also ocusing hei aining on h ee cen al pilla s: ounda ional knowledge, echnical skills, and e hical compe encies (Lopezosa e al.,2023). In his ega d, a ious au ho s a gue ha AI li e acy will be a undamen al equi emen o u u e jou nalis s (G ae e,2016;Sonni e al.,2024; Sonni,2025). The in eg a ion o AI in jou nalism is no wi hou e hical challenges. Beyond he po en ial bene i s a ibu ed o i , one o he mos deba ed dilemmas is he inhe en bias in AI algo i hms (Binns,2018;Deuze & Becke ,2022), compounded by he isk o la ge-scale dissemina ion o disin o ma ion (Ke yso a,2018). Addi ionally, he e is he h ea aced by news consume s in disce ning he sou ce and e aci y o in o ma ion, which can lead o us issues in he media (Mon al & Reich,2017;Wu,2024;Mo osoli e al.,2025). AI can acili a e ce ain speci ic ac i i ies wi hin he news p oduc ion p ocess (Cana - ilhas,2022), among which he igh agains disin o ma ion s ands ou (Man edi Sánchez & U a e Ruiz,2020;Saeidnia e al.,2025), a ma e o pa icula conce n in democ a ic coun ies (Rod íguez Ma elo e al.,2023). In ac , he ad en o AI has inc eased he possibili ies o comba ing disin o ma ion (Mo eno Espinosa e al.,2024) and can help Jou nal. Media 2025,6, 133 3 o 15 dis inguish be ween u h ul in o ma ion and he dis o ion o eali y (Flo es Vi a ,2019; Lange & Lech e man,2021;San os,2023). Balles e os and del Olmo (2024) s a e ha AI has he capaci y o analyze la ge amoun s o in o ma ion quickly and e icien ly, iden i ying pa e ns and anomalies ha can e eal he e aci y o a news i em. In his ega d, some jou nalis ic o ganiza ions ha e exp essed hei dis us ega ding he e olu ion o he sec o , due o he lack o capaci y o comba disin o ma ion and un e i ied in o ma ion (Subiela He nández e al.,2023). To add ess his conce n, ac - checke s ha e eme ged as con ex ual agen s o shape a new ecosys em ha igh s agains disin o ma ion (Pilo Ga cía e al.,2024). Indeed, some ad oca e o he inco po a ion o AI in o con en - e i ica ion p ocesses (Tejedo & Sancho Ligo ed,2023). Thus, he po en ial o AI o in o ma ion e i ica ion—a key p ocess in communica ion—is ecognized (Cua ielles e al.,2023), and i has been demons a ed ha he use o AI can be i al o coun e ac ing he dis up i e e ec s o disin o ma ion (Rubin,2022), as i educes de ec ion ime and inc eases he capaci y o espond o i (Cua ielles e al.,2024). Un o una ely, hese same AI ools can also p o e e ec i e in ampli ying he each and impac o disin o ma ion by enabling he gene a ion o alse con en . The misuse o AI ools ep esen s a signi ican challenge in he in o ma ion age (La Rosa & Luján,2024), as i is possible, a a ela i ely low cos , o use AI o dis ibu e alse in o ma ion wo ldwide. This possibili y is seen as a h ea by he public and by a la ge pa o in o ma ion p o es- sionals, since messages p oduced using a i icial in elligence (AI) echniques can be alsely a ibu ed o non-exis en in o ma ion sou ces, inco po a e alse o inaccu a e con en , and be used o pa isan o malicious pu poses (Peña-Fe nández e al.,2023). The p oblem o disin o ma ion, unde s ood as ha which se es a speci ic pu pose and includes an in en ional componen , poses a g ea e isk when si ua ed wi hin he con ex o a hype connec ed socie y, whe e any use o digi al en i onmen s can simul aneously be bo h a ic im and a conscious o unconscious sou ce o disin o ma ion (Benaissa Ped iza, 2024, pp. 106–107). The malicious use o hese models o gene a e alse and c edible con en au oma ically, massi ely, and a no cos cons i u es one o hei main isks and is he mos immedia e a gumen o hose calling o u gen egula ion (Sanguine i,2023, p. 13). Mo eo e , AI no only has he capaci y o c ea e con incing and ealis ic con en , bu i also aises se ious e hical challenges in mul iple a eas, especially ega ding p i acy, da a p o ec ion, and copy igh , among o he s (Ga cía,2024). In ac , he eme gence o AI in jou nalism aises ques ions abou he e hics, quali y, and anspa ency o au oma ed in o ma ion p ocesses. I is c ucial ha jou nalism p o essionals and media o ganiza ions unde s and hese challenges and wo k o add ess hem e ec i ely (Es eban Regules & Calle Mendoza,2024, p. 251). In his con ex , his a icle aims o analyze Basque jou nalis s’ pe cep ions o he impac o a i icial in elligence on he phenomenon o disin o ma ion. To add ess his main objec i e, his esea ch poses he ollowing ques ions: RQ1. How is AI going o impac in o ma ion diso de s (disin o ma ion, alse con en , and deep akes)? RQ2. A e he e signi ican di e ences among jou nalis s based on a iables such as gende , yea s o p o essional expe ience, he ype o media in which hey wo k, and hei posi ion wi hin he o ganiza ion? RQ3. How do Basque jou nalis s pe cei e isks in hei p o essional use o AI? A which le els—indi idual and associa ed—a e hese isks pe cei ed? Jou nal. Media 2025,6, 133 4 o 15 2. Ma e ials and Me hods This s udy adop s a quan i a i e app oach, based on a s uc u ed su ey adminis- e ed o jou nalis s wo king in he Basque Coun y. The sample consis s o 504 jou nalis s, iden i ied h ough he Open Communica ion Guide o he Basque Go e nmen (Basque Go e nmen ,n.d.), which lis s he ac i e media ou le s in he egion, and wi h he collabo- a ion o he Basque Associa ion o Jou nalis s (Basque Associa ion o Jou nalis s—Basque College o Jou nalis s,n.d.). The dis ibu ion o he su ey also bene i ed om he suppo o his associa ion. O he pa icipan s, 276 a e men (55.1%), 226 a e women (44.5%), and 2 indi iduals iden i ied wi h o he gende s (0.4%). Acco ding o a ailable da a (Basque Go e nmen ,2022;Pé ez e al.,2023), i is es ima ed ha app oxima ely 5000 indi iduals a e employed in he media sec o in Euskadi. The e o e, o a 95% con idence le el, he su ey’s ma gin o e o is ±4.15% (Sa ionandia e al.,2025). Da a acquisi ion was ca ied ou be ween May and June 2024 h ough anonymous su eys, wi hou eco ding any pe sonally iden i iable in o ma ion o sensi i e da a. A no poin we e esponden s asked o p o ide hei name, add ess, email, elephone numbe , o any o he in o ma ion ha could allow o hei di ec o indi ec iden i ica ion. Pa icipa ion in he su ey was en i ely olun a y and based on in o med consen , which was p o ided a he beginning o he ques ionnai e. This in oduc o y sec ion clea ly explained he pu pose o he esea ch, he con iden iali y o esponses, and he esponden ’s igh o decline o comple e he o m i hey wished. As he s udy in ol es no physical o psychological isks, does no p ocess specially p o ec ed da a, and complies wi h he p o isions o he Gene al Da a P o ec ion Regula ion (GDPR) and cu en na ional legisla ion, i is conside ed ha o mal e hics commi ee app o al is no equi ed, in line wi h s anda d guidelines o low- isk esea ch in aca- demic se ings. The ques ionnai e was adminis e ed p ima ily h ough an online pla o m, which maximized geog aphic co e age and pa icipan accessibili y, and was complemen ed by elephone suppo o acili a e pa icipa ion. The su ey was de eloped ad hoc o he p esen s udy. P io o i s adminis a ion, he ques ionnai e unde wen a alida ion p ocess, which included a pilo es wi h a small g oup o jou nalis s o ensu e he cla i y, ele ance, and comp ehensibili y o he i ems. The ques ionnai e included closed and mul iple-choice ques ions, o ganized in o wo main sec ions: (a) sociodemog aphic and p o essional a iables (gende , yea s o expe ience, ype o media ou le , p o ince o wo kplace, main ac i i y in he media, and le el o esponsibili y); and (b) pe cep ions and a i udes ega ding he impac o AI on jou nalism. Among he speci ic issues analyzed a e pe cep ions o he inc ease in he isks o in o ma ion diso de s (disin o ma ion, alse con en , and deep akes), he ise o biases and digi al di ides (gende , social class, e c.), and he iden i ica ion o he main isks associa ed wi h he use o AI in he jou nalis ic ield (allowing o a maximum o wo op ions o be selec ed). Only ully comple ed ques ionnai es we e included in he analysis. Pa ially comple ed su eys and hose missing in o med consen we e excluded on ha basis. The hea maps p esen ed in his s udy we e gene a ed using Cha GPT 4.1. All s a is ical analyses, including co espondence analysis (CA), we e pe o med using R ( e sion 4.4.0) wi hin he RS udio in eg a ed de elopmen en i onmen (IDE) ( e sion 1.1.456). 3. Resul s 3.1. Jou nalis s’ Pe cep ions o Disin o ma ion Associa ed wi h AI The majo i y o p o essionals pe cei e ha a i icial in elligence will signi ican ly inc ease he isks o disin o ma ion. Speci ically, 89.88% o he esponses all in o he Jou nal. Media 2025,6, 133 5 o 15 ca ego ies “conside ably” (n = 171) and “signi ican ly” (n = 282), while only 10.12% conside ha he impac o all unde he ca ego ies o “minimally” (n = 46) o “no a all” (n = 5). These esul s sugges ha jou nalis s iew a i icial in elligence as a signi ican challenge o he in eg i y o he in o ma ion ecosys em. When analyzing he dis ibu ion o esponses acco ding o he ype o media in which hey wo k (Figu e 1), a end simila o ha o he o e all sample is obse ed. Howe e , among jou nalis s wo king in digi al-na i e media, al hough he nega i e pe cep ion e- mains p edominan (81.81%, n = 54), he e is a highe pe cen age o esponses ha minimize he impac (“minimally”: 18.18%) compa ed o he a e age o o he ca ego ies, such as p in , adio, o ele ision (91.03%). In con as , p o essionals wo king in communica ion o ices and p ess agencies a e he mos likely o conside ha AI will ha e a high nega i e impac , wi h 96.95% (n = 84) in he highes -impac ca ego ies. Figu e 1. Pe cen age dis ibu ion o pe cei ed disin o ma ion isk by media ype (X-axis: pe cep ion a ibu es; Y-axis: media ype). The pe cep ion o he isk o AI-d i en disin o ma ion is e y simila ac oss gende s, wi h 90.22% (n = 249) o men and 89.82% (n = 196) o women in e ms o ela i e equency o he “signi ican ly” and “conside ably” op ions (Figu e 2). The “o he ” ca ego y includes only wo cases, so i s ep esen a i eness is e y limi ed. The esul s o Fishe ’s exac es , wi h a p- alue o 0.05419, a e no conclusi e, as hey a e sligh ly abo e he e e ence alue o 0.05. A e excluding he “o he ” ca ego y and g ouping he “minimally” and “no a all” ca ego ies o ensu e a su icien ly la ge sample size pe ca ego y ( ypically abo e 30 cases) o he Chi-squa e es , he p- alue inc eases o 0.5916. The e ec size, measu ed by C amé ’s V, is 0.046, which is conside ed e y small (Cohen,1988). Adjus ing o mul iple compa isons using he Holm and Benjamini– Hochbe g me hods did no al e his conclusion (adjus ed p- alues: 0.5916). The e o e, we can conclude ha he e is no su icien s a is ical e idence o a ela ionship be ween hese a iables. Wi h ega d o he ela ionship be ween in o ma ion diso de s and yea s o expe ience in jou nalism (Figu e 3), he sum o he pe cen ages o he “signi ican ly” and “consid- e ably” op ions o each o he g oups s udied sugges s a end wi h some nuances: as yea s o expe ience inc ease, so does he pe cep ion o AI- ela ed disin o ma ion isk (mo e han 20 yea s, 92.76%; 10 o 20 yea s, 94.44%; 5 o 10 yea s, 87.32%; and i e yea s o less, Jou nal. Media 2025,6, 133 6 o 15 80.77%). Jou nalis s wi h i e yea s o less o p o essional expe ience epo he lowes pe cen age in isk pe cep ion; howe e , his g oup s ill ep esen s a clea majo i y o e all. Figu e 2. Pe cen age dis ibu ion o pe cei ed disin o ma ion isk by gende (X-axis: gende ; Y-axis: pe cep ion a ibu es). Figu e 3. Pe cen age dis ibu ion o pe cei ed disin o ma ion isk by yea s o expe ience (X-axis: yea s o expe ience; Y-axis: pe cep ion a ibu es). In his case, he p- alue (0.001937) om Fishe ’s exac es clea ly indica es he exis ence o s a is ically signi ican e idence; hese da a suppo he conclusion ha he e is an associa ion be ween he a iables s udied. The e ec size (C amé ’s V = 0.144) is small o medium, and he esul emains signi ican a e adjus ing o mul iple compa isons (adjus ed p- alue: 0.0058, Holm; 0.0058, Benjamini–Hochbe g). Jou nal. Media 2025,6, 133 7 o 15 The co espondence analysis isualiza ion (Figu e 4) co obo a es he p e iously no ed end. As p o essional expe ience in he ield inc eases, jou nalis s a e mo e likely o pe cei e a g ea e in luence o AI on disin o ma ion. Ne e heless, all jou nalis s, ega dless o yea s o expe ience, a ibu e some impac o AI, as can be in e ed om he dis ance be ween he “no a all” ca ego y ( ed iangle) and i s coun e pa s ep esen ing yea s o expe ience (blue do s). Rega ding he co espondence analysis, he i s dimension explained 85.7% o he o al ine ia, while he second dimension accoun ed o 13.3%. Toge he , he i s wo dimensions explained 99.0% o he o al a iance, indica ing ha he wo-dimensional solu ion p o ides a good summa y o he associa ion s uc u e in he da a. Figu e 4. Co espondence analysis o he associa ion be ween jou nalis ic expe ience and pe cei ed in luence o a i icial in elligence on disin o ma ion. Wi h ega d o how he use o a i icial in elligence in he wo kplace in luences he pe cep ion o disin o ma ion isks, Fishe ’s exac es yielded a p- alue o 0.0056, indica ing a s a is ically signi ican associa ion be ween he a iables. The e ec size, measu ed by C amé ’s V, was 0.120, which is conside ed small o medium acco ding o Cohen’s (1988) c i e ia. This esul emained signi ican a e adjus ing o mul iple compa isons using bo h he Holm (adjus ed p= 0.0112) and Benjamini–Hochbe g (adjus ed p= 0.0084) me hods. As shown in he co espondence analysis (Figu e 5), he e is an in e se ela ionship be ween jou nalis s who do no use a i icial in elligence a all and hei pe cep ion o i s in luence on disin o ma ion, as hey end o choose he “signi ican ly” ca ego y. In con as , in o ma ion p o essionals who make in ensi e use o his echnology ha e a mo e mode a e pe cep ion o i s impac . The co espondence analysis e ealed ha Dimension 1 accoun ed o 70.3% o he o al ine ia, and Dimension 2 con ibu ed an addi ional 27.6%. Combined, hese wo dimensions cap u ed 97.9% o he o e all a iance, demon- s a ing ha he ela ionship be ween he a iables is e ec i ely ep esen ed wi hin a wo- dimensional space. Jou nal. Media 2025,6, 133 8 o 15 Figu e 5. Co espondence analysis o a i icial-in elligence usage in he wo kplace and jou nalis s’ pe cep ions o disin o ma ion isk. 3.2. Pe cei ed Risks o AI Use Among Basque Jou nalis s Rega ding he speci ic isks associa ed wi h he di ec use o a i icial in elligence, Basque jou nalis s mainly highligh he di icul y in iden i ying alse con en and deep akes, as well as he isk o ob aining inaccu a e o e oneous da a. Di icul y de ec ing alse con en and deep akes is he mos equen ly ci ed isk, accoun ing o 37.9% o esponses (n = 382), ollowed by conce ns abou con en o da a inaccu acy, men ioned by 33.33% (n = 336). O he isks include becoming a ic im o c iminal uses, such as scams o skimming (11.61%, n = 117), and biases s emming om da a o igin (gende , social class, e c.) (10.81%, n = 109). Only 6.35% (n = 64) (Table 1) epo no addi ional isks. I is impo an o no e ha he o al numbe o esponses (1008) exceeds he numbe o esponden s (504), as each pa icipan was allowed o selec up o wo isks. This design pe mi ed esponden s o indica e mo e han one pe cei ed isk, esul ing in a highe o e all esponse coun . Table 1. Main isks iden i ied by jou nalis s in hei use o AI. Risk Ca ego y Coun Pe cen age Di icul y iden i ying alse con en and deep akes 382 37.9% Ob aining inaccu a e o e oneous con en /da a 336 33.33% Being a ic im o c iminal uses (scams and skimming) 117 11.61% Biases due o da a o igin (gende , social class, e c.) 109 10.81% None o he 64 6.35% No e: Responden s could selec up o wo isks. To examine how he di e en isks a e associa ed in esponden s’ minds, we buil a 4 × 4 co-occu ence ma ix ha combines he wo op ions each jou nalis could selec . Table 2 epo s he en dis inc isk pai s and hei ela i e equencies; he h ee mos equen pai s accoun o mo e han h ee qua e s o all co-occu ences. Jou nal. Media 2025,6, 133 9 o 15 Table 2. Co-occu ence o pe cei ed isk pai s in jou nalis s’ use o AI. Risk Pai Coun Pe cen age Di icul y iden i ying alse con en and deep akes pai ed wi h ob aining inaccu a e o e oneous con en /da a 253 50.5% Di icul y iden i ying alse con en and deep akes pai ed wi h being a ic im o c iminal uses (scams, skimming) 69 13.77% Ob aining inaccu a e o e oneous con en /da a pai ed wi h biases due o da a o igin (gende , social class, e c.) 58 11.58% Di icul y iden i ying alse con en and deep akes pai ed wi h biases due o da a o igin (gende , social class, e c.) 40 7.99% Ob aining inaccu a e o e oneous con en /da a pai ed wi h being a ic im o c iminal uses (scams, skimming) 17 3.99% Di icul y iden i ying alse con en and deep akes pai ed wi h none o he 20 3.99% Being a ic im o c iminal uses (scams, skimming) pai ed wi h none o he 25 4.99% Ob aining inaccu a e o e oneous con en /da a pai ed wi h none o he 8 1.6% Biases due o da a o igin (gende , social class, e c.) pai ed wi h0 none o he 5 1% Biases due o da a o igin (gende , social class, e c.) pai ed wi h being a ic im o c iminal uses (scams, skimming) 6 1.2% No e: Because each esponden could selec up o wo isks, he co-occu ence coun s a e based on 504 esponden s. Among he jou nalis s who selec ed wo isks, 50.5% (n = 253) combined conce ns abou di icul y iden i ying alse con en and deep akes wi h he isk o ob aining inaccu a e o e oneous con en o da a. This inding sugges s ha , o a signi ican po ion o esponden s, he challenges o de ec ing alse con en , such as deep akes, manipula ed images, o ab ica ed news, a e closely linked o he isk o ecei ing inaccu a e o e oneous in o ma ion gene a ed by AI sys ems. In o he wo ds, jou nalis s do no iew hese issues in isola ion; a he , hey pe cei e hem as pa o a b oade , in e connec ed h ea o he eliabili y o in o ma ion. The second mos common pai ing, selec ed by 13.8% (n = 69) o esponden s, links he di icul y o iden i ying alse con en and deep akes wi h he isk o being a ic im o c iminal uses, such as scams o skimming. The hi d mos equen pai ing, iden i ied by 11.6% (n = 58) o esponden s, combines conce ns abou ob aining inaccu a e o e oneous con en o da a, wi h wo ies abou biases due o da a o igin, such as gende , social class, o o he demog aphic ac o s. This co-occu ence indica es ha , o a segmen o jou nalis s, he issue o ac ual co ec ness is closely in e wined wi h he ques ion o ai ness in AI-gene a ed in o ma ion. A ela ed inding is ha 7.99% (n = 40) o esponden s also pai ed di icul y iden i ying alse con en and deep akes wi h conce ns abou da a biases. The emaining pai s ha e ma kedly lowe equencies. Two in ol e “none o he ”, sugges ing a subse o esponden s who pe cei e only a single addi ional isk beyond hei p ima y choice. Finally, co-occu ences be ween biases and c iminal uses (1.2%) o be ween inaccu acy and c iminal uses (3.99%) appea only spo adically. 4. Discussion In line wi h he indings o p e ious s udies, he esul s o his esea ch show a b oad consensus among jou nalis s ega ding he pe cei ed isk ha he implemen a ion o a i icial in elligence may pose in ampli ying disin o ma ion (RQ1). This iew gi es ise o p edominan ly cau ious a i udes owa d he applica ion o his echnology in jou nalism