Traditional literary criticism is shaped by critics’ personal interests, research orientations, and inevitably limited time, and thus often focuses on a small number of phenomenal “masterpieces” while overlooking many other works across the vast literary landscape. In the face of the massive volume of online literature, traditional literary criticism falls short. Today, with the help of AI, it has become possible to include a greater number of works into the scope of criticism. While AI criticism opens up new frontiers in critical analysis, it also has its limitations, underscoring the necessity of human–machine collaboration to overcome these shortcomings.

Resurrecting literary ‘ghosts’

AI technology not only catalyzes literary production but also empowers literary criticism. By introducing fresh analytical perspectives, AI makes it possible to rediscover vast numbers of literary “ghosts,” many of them found in online literature, and bring them back into critical view.

First, through natural language processing, AI can rapidly analyze words, sentences, and paragraphs across large bodies of text. It can identify details that traditional literary criticism may overlook—such as the frequency of specific terms across different works—thereby generating new critical insights. Constrained by limited samples and the time required for extensive reading, traditional criticism struggles to achieve this kind of “distant reading.”

Second, AI’s pattern-recognition capabilities can detect structural similarities across massive volumes of texts. This makes it possible to trace the development and evolution of particular literary genres, clarify patterns of inheritance and transformation between genres, and reassess their place in literary history.

Finally, AI’s generative and inferential technologies allow for the deconstruction and recombination of textual language patterns and semantic networks through algorithmic inference and simulation. This opens up the internal spaces of texts and creates new possibilities for multiple interpretations.

Struggling to capture ‘deep ghosts’

While the “distant reading” enabled by AI offers new perspectives for literary analysis, this does not imply that AI criticism can fully replace traditional criticism. The “deep ghosts” manifest as certain “gaps” within the text. In literary works, such gaps are almost ubiquitous—whether in abruptly truncated dialogue, unfinished paragraphs, or sudden interruptions in narrative development. Regardless of their form, these gaps are deliberately crafted by the author. They represent suppressed emotions or unspoken philosophical truths and, by remaining incomplete, invite readers to imagine and feel their meaning. These gaps are precisely where the “deep ghosts” of a text reside: elusive, formless, and beyond the reach of algorithmic analysis, existing only within the aesthetic imagination of human critics.

“Deep ghosts” also lurk in the finer details of the text. The actions of literary characters are not driven solely by personal will but are shaped by sociocultural norms, including cultural expectations, ethical constraints, and everyday etiquette. This influence is often reflected in subtle details of daily life: a character’s speech patterns, demeanor, attire, or even a fleeting glance or expression may reveal the shadowy presence of these social norms.

Moreover, the “deep ghosts” emerge through elusive symbolic metaphors. In many works, the symbolic meanings underlying key themes are not explicitly articulated in the text. Readers must instead draw on their own life experiences and aesthetic imagination to perceive them.

Human–machine collaboration

The complementary tension between traditional criticism and AI criticism highlights the unique value of the human–machine collaboration paradigm in literary analysis.

First, human–machine collaboration can expand both the breadth and depth of literary criticism. AI can rapidly scan massive volumes of literary works, using “distant reading” to identify macro-level features such as literary patterns and structures, thereby mapping out a broad literary landscape. Human critics, meanwhile, can delve into the intricate textures of individual works through “close reading,” engaging with the emotional density of language and resonating with the subtle sentiments embedded in the text.

The organic integration of these two approaches avoids the narrowness that may result from an exclusive reliance on traditional “close reading” while also mitigating the bias toward instrumental rationality in AI criticism. This synthesis achieves a dialectical unity of breadth and depth, making literary criticism more comprehensive, more profound, and better equipped to reflect the complexity and variability of literary phenomena.

Second, human–machine collaboration can harmonize the subjectivity and objectivity of literary criticism. Supported by algorithms and data, AI criticism can adopt a relatively impersonal stance, enabling a macro-level grasp and evaluation of literary works. However, it also risks slipping into excessive instrumental rationality and losing sensitivity to the nuanced meanings embedded in texts.

Every literary work embodies the author’s lived experiences and aesthetic sensibilities, giving rise to distinctive artistic styles that often elude large-scale data analysis. In this regard, the “close reading” conducted by human critics becomes particularly vital. This synergy thus ensures that literary criticism avoids forced interpretations resulting from subjective arbitrariness while preventing overly simplistic conclusions, striking a delicate balance between the subjectivity and objectivity of literary criticism.

Third, human–machine collaboration enables the reconciliation of universality and particularity in literary criticism. AI can identify commonly recurring structural patterns across vast volumes of texts, but it struggles to engage deeply with individual works or grasp the author’s most profound lived experiences and aesthetic sensibilities. As a result, it may resort to imposing universal structural patterns to interpret the historical evolution of literary phenomena.

By contrast, human literary critics can offer creative interpretations grounded in their distinctive perspectives, not only uncovering the true value of literary works but also generating new ideas and theories from them. The human–machine collaborative model of criticism thus represents a promising transformation in the current paradigm of literary criticism.

Nonetheless, as human–machine collaboration continues to deepen, vigilance is required against the possibility that the authority to define standards of literary criticism could shift from humans to AI, giving rise to a form of discourse legitimated in the name of “objectivity.” It is therefore essential to uphold the conviction that AI should always remain a tool of literary criticism, while humans are—and will continue to be—the ultimate agents of critical judgment.

Yu Jianxiang is a professor from the School of Humanities at Central South University.