Assessing the accuracy and usability of artificial intelligence–based language models in responding to common periodontal patient questions

{"BACKGROUND:":[0],"Artificial":[1],"intelligence-powered":[2],"large":[3],"language":[4,288],"models":[5,450],"(LLMs)":[6],"are":[7,319,459,476],"increasingly":[8,382],"used":[9,48],"by":[10,89,96,259],"patients":[11,294,363,381],"seeking":[12],"quick":[13],"information":[14,361,401],"regarding":[15,26],"dental":[16,245,428],"and":[17,30,44,73,81,87,126,142,234,265,279,290,316,325,335,351,368,406,479],"medical":[18],"problems.":[19],"Despite":[20],"their":[21,427],"growing":[22],"popularity,":[23],"concerns":[24],"remain":[25,284],"the":[27,42,114,165,201,215,343,394,409,447,482],"accuracy,":[28],"clarity,":[29],"clinical":[31,344],"usefulness":[32],"of":[33,46],"LLM-generated":[34],"responses.":[35],"This":[36,430],"study":[37,431],"aimed":[38],"to":[39,52,75,321,364,397,421,485],"comparatively":[40],"evaluate":[41],"accuracy":[43],"usability":[45,119,207,292],"widely":[47],"LLMs":[49,77,219,248,299],"in":[50,78,252,286,291,437,468,471],"responding":[51],"frequently":[53],"asked":[54,64],"periodontal":[55,65,255,338],"questions.":[56,229,429],"METHODS:":[57],"In":[58],"this":[59],"analytical-comparative":[60],"study,":[61],"15":[62],"commonly":[63],"questions":[66,225,442],"were":[67,84,94],"selected":[68],"from":[69],"real":[70],"patient":[71,306],"encounters":[72],"administered":[74],"ten":[76],"both":[79],"English":[80,179,469],"Persian.":[82],"Responses":[83],"evaluated":[85,197],"independently":[86],"blindly":[88],"two":[90,460],"board-certified":[91],"periodontists;":[92],"disagreements":[93],"resolved":[95],"discussion":[97],"or":[98],"third-reviewer.":[99],"Each":[100],"response":[101],"was":[102,147],"rated":[103],"on":[104,223,227,384],"a":[105,375,492,499,503],"5-point":[106],"Likert":[107],"scale":[108],"across":[109,168],"six":[110],"criteria:":[111],"correlation":[112,199],"with":[113,122,139,160,176,210,240,295,362,408,426,502],"question,":[115],"adequacy,":[116],"comprehensiveness,":[117],"clarity/readability,":[118],"for":[120,208,293,305,347,386,424,481],"individuals":[121,209],"limited":[123,211,296],"scientific":[124,127,212,297],"literacy,":[125],"accuracy.":[128],"Statistical":[129],"analyses":[130],"included":[131],"descriptive":[132],"statistics,":[133],"Mann-Whitney":[134],"U":[135,190],"test,":[136],"Kruskal-Wallis":[137],"test":[138],"Bonferroni":[140],"correction,":[141],"univariate":[143],"ANOVA.":[144],"Significance":[145],"level":[146],"set":[148],"at":[149],"p":[150,156,193],"<":[151,157,194],"0.05.":[152],"RESULTS:":[153],"=":[154,191],"294.78,":[155],"0.001).":[158,195],"ChatGPT-4.5":[159,231,241,274],"deep":[161],"search":[162],"enabled":[163],"achieved":[164],"highest":[166,202],"scores":[167],"most":[169,448],"criteria,":[170,198],"while":[171,488],"ChatGPT-4o":[172],"generally":[173,336],"underperformed":[174],"compared":[175],"other":[177],"LLMs.":[178],"responses":[180],"scored":[181],"significantly":[182,221,466],"higher":[183],"than":[184,226,470],"Persian":[185,287],"(mean":[186],"4.25":[187],"vs.":[188],"3.96;":[189],"333,058,":[192],"Among":[196],"had":[200,214],"mean":[203],"score":[204,217],"(4.81),":[205],"whereas":[206],"literacy":[213],"lowest":[216],"(3.80).":[218],"performed":[220],"better":[222,467],"treatment/prevention":[224],"diagnosis/pathogenesis":[228],"Only":[230],"(deep":[232,236],"search)":[233,237],"Grok-3":[235],"provided":[238],"citations,":[239],"referencing":[242],"more":[243],"specialized":[244],"sources.":[246],"CONCLUSIONS:":[247],"show":[249],"variable":[250],"performance":[251,464],"answering":[253],"frequent":[254],"patients'":[256],"questions,":[257],"influenced":[258],"model":[260],"architecture,":[261],"language,":[262],"question":[263],"type,":[264],"evaluative":[266],"domain.":[267],"Although":[268],"advanced":[269,449],"configurations":[270],"such":[271],"as":[272,302,393,452],"deep-search":[273],"offer":[275],"highly":[276,333],"accurate,":[277,405,455],"comprehensive,":[278],"well-referenced":[280],"responses,":[281],"significant":[282],"limitations":[283],"particularly":[285],"output":[289],"literacy.":[298],"may":[300],"serve":[301],"complementary":[303],"tools":[304,331,435],"communication,":[307],"but":[308],"should":[309,356,496],"not":[310],"replace":[311,498],"professional":[312,500],"advice.":[313],"Further":[314],"development":[315],"language-specific":[317],"optimization":[318],"needed":[320],"improve":[322],"global":[323],"accessibility":[324],"reliability.":[326],"KEY":[327],"POINTS:":[328],"While":[329,446],"AI":[330,385,489],"provide":[332,454],"readable":[334],"accurate":[337],"information,":[339,457],"they":[340,440],"often":[341],"lack":[342],"nuance":[345],"required":[346],"personalized":[348],"risk":[349],"assessment":[350],"complex":[352],"treatment":[353,412],"planning.":[354],"Practitioners":[355],"proactively":[357],"discuss":[358],"AI-generated":[359],"health":[360,388],"address":[365],"potential":[366],"oversimplifications":[367],"ensure":[369],"online":[370,400],"advice":[371],"is":[372,465],"integrated":[373],"into":[374],"professional,":[376],"evidence-based":[377],"care":[378],"plan.":[379],"As":[380],"rely":[383],"initial":[387],"guidance,":[389],"practitioners":[390],"must":[391],"act":[392],"authoritative":[395],"\\"filter\\"":[396],"verify":[398],"that":[399,433],"remains":[402],"clinically":[403],"safe,":[404],"aligned":[407],"patient's":[410],"specific":[411],"goals.":[413],"PLAIN":[414],"LANGUAGE":[415],"SUMMARY:":[416],"Many":[417],"people":[418],"now":[419],"turn":[420],"artificial":[422],"intelligence":[423],"help":[425],"found":[432],"these":[434],"vary":[436],"how":[438],"well":[439],"answer":[441],"about":[443],"gum":[444],"disease.":[445],"(such":[451],"ChatGPT-4.5)":[453],"high-quality":[456],"there":[458],"main":[461],"limitations:":[462],"first,":[463],"Persian;":[472],"second,":[473],"many":[474],"explanations":[475],"too":[477],"technical":[478],"hard":[480],"average":[483],"person":[484],"understand.":[486],"Overall,":[487],"can":[490],"be":[491],"helpful":[493],"guide,":[494],"it":[495],"never":[497],"consultation":[501],"dentist.":[504]}