Artificial Intelligence in Accounting: A Systematic Literature Review

In the world of modern finance, numbers have always told a story, but the way those stories are read is changing. For decades, accountants and auditors have relied on human judgment to check ledgers, spot errors, and ensure that financial reports are honest. Today, a new kind of tool is entering the office: artificial intelligence. This is not a single machine, but a collection of smart computer programs capable of learning from data. These tools can scan thousands of documents in seconds, looking for patterns that a human eye might miss. They are being used to automate routine tasks, predict financial risks, and hunt for fraud. Yet, as these tools become more common, a question remains: how much do we truly understand about their impact on the profession? Are they replacing human decision-makers, or simply acting as powerful assistants? The answers are scattered across hundreds of new studies, creating a landscape that is rich with information but difficult to navigate.

To bring clarity to this growing field, a researcher named Jinwoo Chae from GlobalNxt University set out to map the current state of knowledge. The goal was not to test a specific software or run a new experiment, but to gather and organize the existing research published between 2023 and 2026. The researcher started with a list of 49 records, which included journal articles, book chapters, and data reports. After carefully removing duplicates and studies that did not fit the scope of accounting, 47 publications remained to be examined. The process involved reading the titles and types of these works to categorize them by year, topic, and method. It was a systematic effort to see where the research was heading, what questions were being asked, and where the gaps in knowledge still lay.

The first thing the review revealed was the sheer speed of growth in this field. The number of studies published has exploded in just a few years. In 2023, there were only four studies on the list. By 2024, that number had more than doubled. The momentum continued into 2025 and 2026, with 16 and 18 studies respectively. In total, 34 of the 47 studies, which is about 72 percent, were published in the last two years alone. This rapid increase suggests that the conversation around artificial intelligence in accounting is moving from a niche interest to a central topic of discussion. However, the review also noted that this list represents a specific snapshot of available literature and should not be taken as a complete count of every study ever written on the subject.

When looking at what these studies were actually about, a clear pattern emerged. The research is heavily focused on two main areas: general overviews of how artificial intelligence fits into accounting and auditing, and the specific application of these tools to detect fraud. Nearly two-thirds of the studies fall into these categories. Fraud detection, in particular, is a concentrated area of interest because computers are exceptionally good at sifting through massive amounts of data to find unusual patterns that might indicate irregularities or error. Auditing, which involves checking the accuracy of financial records, is the second most common function being studied. In contrast, other important parts of the accounting world, such as management accounting—which helps businesses plan and control their operations—appear almost entirely absent from the current research, with only one study dedicated to it. Taxation is also rarely the main focus, usually appearing only as a small part of broader titles.

The methods used in these studies tell another part of the story. A large portion of the work consists of reviews, where researchers summarize what others have already found. About 13 percent of the studies are systematic literature reviews, and another 15 percent are narrative or conceptual reviews. Only a very small number of the studies, just one that could be clearly identified from its title, presented new field evidence or data collected from real-world experiments. This means that the current body of knowledge is built more on summarizing ideas and discussing possibilities than on testing them with hard data from actual accounting firms. While these reviews are useful for getting a broad picture of a fast-moving field, the researcher suggests that relying too heavily on them limits the ability to draw firm, evidence-based conclusions.

The review also highlighted a need for greater precision in how these technologies are described. Some studies treat all forms of artificial intelligence as a single, uniform tool. However, the researcher argues that this is misleading. There are important differences between simple rule-based automation, which follows strict instructions, and advanced machine learning, which can learn and adapt. There is also a distinction between these and generative models that can create new content. Treating them all as the same "artificial intelligence" can hide important differences in what they can do, what data they need, and the risks they pose. Future research needs to be more specific about which technology is being used and how much human oversight is involved.

Looking ahead, the researcher proposes a new direction for the field. The next stage of study should move beyond general discussions and conceptual arguments toward more rigorous, real-world testing. This means using established theories to explain why accountants adopt these tools and how they change professional judgment. It also means expanding the scope of research to include different countries, smaller businesses, and under-studied areas like sustainability reporting and tax compliance. Crucially, the researcher calls for more studies that examine the governance of these systems: who is responsible when a computer makes a mistake, and how can we ensure the decisions made by these tools are reliable and fair?

The review concludes that while the field is growing rapidly, it is still in an early stage. The current literature is dominated by broad discussions and reviews, with a shortage of deep, empirical evidence from real accounting environments. The concentration of research on fraud detection and auditing suggests that the profession has naturally gravitated toward areas where the benefits of automation are most obvious. To truly understand the future of accounting, researchers must now broaden their lens. They need to investigate how these tools affect the daily work of accountants in various settings, how they change the way professionals think, and what rules are needed to keep the system honest. The path forward requires moving from asking what artificial intelligence might do, to measuring exactly what it is doing, and how it is reshaping the profession from the inside out.