This retrospective, single-platform study asked whether deep learning and machine-learning models can accurately detect COVID-19 misinformation in Persian-language Instagram posts, using transformer-based NLP, LSTM sequence models, CNN text classifiers, and traditional ML algorithms to perform binary misinformation classification. Researchers analyzed Instagram posts and comments (collected 5 September – 11 December 2022) comprising Persian COVID-19 content sourced through hashtag-based searches and health-organization accounts. Text preprocessing included Persian-specific normalisation, tokenisation using the Hazm library, and manual annotation by a multidisciplinary team of clinicians, with verification via 30 peer-reviewed COVID-19 studies. The models tested included XGBoost, KNN, CNN, LSTM, and ParsBERT, compared against each other as methodological baselines. The best-performing model, LSTM, achieved 97% accuracy, precision 0.91, recall 0.85, and an F1-score of 0.85. The analysis showed that misinformation constituted a minority class, contributing to imbalanced performance patterns in which LSTM exhibited high precision but lower recall. Secondary analyses included training-curve comparisons, model-specific scalability assessments, and evaluation of runtime efficiency, where LSTM outperformed CNN and BERT. Additional results demonstrated that LSTM and BERT achieved the strongest recall among all models tested, whereas XGBoost and KNN underperformed on semantic nuance and scalability, particularly given the complexity of Persian linguistic structure. Limitations include potential selection bias due to Instagram-only sampling, imbalanced class distributions, and subjectivity in manual annotations despite expert review. External validation was not performed, and subgroup fairness analyses were absent. Findings reflect patterns in Persian online discourse and model performance on annotated datasets rather than real-world behavioral or health outcomes and therefore do not imply clinical or population-level efficacy.