<h3>Importance</h3> There are limitations in current diagnostic testing approaches for Alzheimer disease (AD). <h3>Objective</h3> To examine plasma tau phosphorylated at threonine 217 (P-tau217) as a diagnostic biomarker for AD. <h3>Design, Setting, and Participants</h3> Three cross-sectional cohorts: an Arizona-based neuropathology cohort (cohort 1), including 34 participants with AD and 47 without AD (dates of enrollment, May 2007-January 2019); the Swedish BioFINDER-2 cohort (cohort 2), including cognitively unimpaired participants (n = 301) and clinically diagnosed patients with mild cognitive impairment (MCI) (n = 178), AD dementia (n = 121), and other neurodegenerative diseases (n = 99) (April 2017-September 2019); and a Colombian autosomal-dominant AD kindred (cohort 3), including 365<i>PSEN1</i>E280A mutation carriers and 257 mutation noncarriers (December 2013-February 2017). <h3>Exposures</h3> Plasma P-tau217. <h3>Main Outcomes and Measures</h3> Primary outcome was the discriminative accuracy of plasma P-tau217 for AD (clinical or neuropathological diagnosis). Secondary outcome was the association with tau pathology (determined using neuropathology or positron emission tomography [PET]). <h3>Results</h3> Mean age was 83.5 (SD, 8.5) years in cohort 1, 69.1 (SD, 10.3) years in cohort 2, and 35.8 (SD, 10.7) years in cohort 3; 38% were women in cohort 1, 51% in cohort 2, and 57% in cohort 3. In cohort 1, antemortem plasma P-tau217 differentiated neuropathologically defined AD from non-AD (area under the curve [AUC], 0.89 [95% CI, 0.81-0.97]) with significantly higher accuracy than plasma P-tau181 and neurofilament light chain (NfL) (AUC range, 0.50-0.72;<i>P</i> &lt; .05). The discriminative accuracy of plasma P-tau217 in cohort 2 for clinical AD dementia vs other neurodegenerative diseases (AUC, 0.96 [95% CI, 0.93-0.98]) was significantly higher than plasma P-tau181, plasma NfL, and MRI measures (AUC range, 0.50-0.81;<i>P</i> &lt; .001) but not significantly different compared with cerebrospinal fluid (CSF) P-tau217, CSF P-tau181, and tau-PET (AUC range, 0.90-0.99;<i>P</i> &gt; .15). In cohort 3, plasma P-tau217 levels were significantly greater among<i>PSEN1</i>mutation carriers, compared with noncarriers, from approximately 25 years and older, which is 20 years prior to estimated onset of MCI among mutation carriers. Plasma P-tau217 levels correlated with tau tangles in participants with (Spearman ρ = 0.64;<i>P</i> &lt; .001), but not without (Spearman ρ = 0.15;<i>P</i> = .33), β-amyloid plaques in cohort 1. In cohort 2, plasma P-tau217 discriminated abnormal vs normal tau-PET scans (AUC, 0.93 [95% CI, 0.91-0.96]) with significantly higher accuracy than plasma P-tau181, plasma NfL, CSF P-tau181, CSF Aβ42:Aβ40 ratio, and MRI measures (AUC range, 0.67-0.90;<i>P</i> &lt; .05), but its performance was not significantly different compared with CSF P-tau217 (AUC, 0.96;<i>P</i> = .22). <h3>Conclusions and Relevance</h3> Among 1402 participants from 3 selected cohorts, plasma P-tau217 discriminated AD from other neurodegenerative diseases, with significantly higher accuracy than established plasma- and MRI-based biomarkers, and its performance was not significantly different from key CSF- or PET-based measures. Further research is needed to optimize the assay, validate the findings in unselected and diverse populations, and determine its potential role in clinical care.