**2. Methods**

For the purposes of this narrative review, we performed an exhaustive literature search, with primary source platforms being Google™ Scholar and PubMed. The primary search term was "surgery" with the following secondary terms—"artificial intelligence," "machine learning," "technology," and "subspecialty." Specific names of surgical specialties (e.g., orthopedics, neurosurgery, and vascular surgery) were also employed. The primary search term "surgery" in combination with each of the other keywords, in various iterations, resulted in more than 875,000 potential listings. Literature screening focused on sources with "full text" availability, limited to English language. In addition, various correspondences (e.g., Letters to Editor and Brief Communications) were excluded. This resulted in approximately 142,000 secondary literature results. The search was limited to original research and reviews within this group, with at least five citations (using Google™ Scholar). With these criteria, our final list of potentially suitable articles was fewer than 2000. A more intensive review of the tertiary phase of our article screening resulted in 96 articles with relevance to this review. After this, secondary sources (derived during in-depth review of our 96 most relevant articles and examining their respective reference lists) were added. Utilizing the above methodology, the resultant reference list includes 158 citations (**Figure 1**).

In the primary search, only studies with five or more citations were considered. Because newer studies tend to have fewer citations, this may introduce selection bias against newer studies that either address aspects of these concerns or bring up new ones. Given the rapidly evolving field of AI, future reviews could evaluate more novel studies for potential innovations.

**Figure 1.** *Flowchart of the selection process for review articles.*
