PSYCHOLINGUISTIC ANALYSIS OF TEXTS GENERATED BY ARTIFICIAL INTELLIGENCE (USING THE EXAMPLE OF ARTICLES ON THE ZEN PLATFORM)

Gritsai L.A.

Yaroslavl State Pedagogical
K.D. Ushinsky University
Yaroslavl, Russia
e-mail: usan82@gmail.com

Introduction

The rapid development of generative language models and their integration into media platforms are creating a fundamentally new communicative reality, in which text is increasingly the product of algorithmic processing of large data sets rather than the result of individual speech activity. In this context, the Zen platform represents a illustrative environment where one can observe the transition from author-driven blogging to the mass distribution of generated publications devoid of any distinct subjective element. The relevance of this study is determined by the fact that, with the increasing share of algorithmically generated materials, it is necessary to understand their impact on readers' consciousness, the formation of cognitive schemas, the depth of information processing, and the specifics of emotional responses, as these processes are directly linked to the transformation of information culture.

The aim of the study is to identify psycholinguistic parameters characterizing the generated texts and to determine their differences from the author's publications in terms of coherence indicators,
The theme is rheme dynamics, lexical variability, and pragmatic focus; the scientific novelty of the work lies in the integration of psycholinguistic analysis methods with a media discourse approach as applied to a specific social platform, and its significance is determined by the need to develop theoretical and applied foundations for assessing the impact of algorithmic content on the cognitive sphere of an individual.

 

Methodology and methods

The methodological basis of the work was formed by the principles of textual psycholinguistics, cognitive linguistics, and media discourse theory, as well as ideas of the algorithmic public sphere developed in interdisciplinary studies of digital media. The methods used included comparative psycholinguistic analysis, content analysis, structural-semantic modeling, elements of discourse analysis, and interpretive analysis. This allowed us to compare original and generated publications on the Zen platform, identify their systemic differences, and describe the characteristics of their cognitive impact.

The degree of development of the problem in modern science should be characterized as being in the stage of active formation, since the foundations of the analysis of textual coherence and discursive understanding were laid in the works of T. A. van Dijk [7], M. A. K. Halliday [31],
R. de Bogrand and V. Dressler [30], and the cognitive mechanisms of text interpretation were covered in the works of A.A. Leontiev [15], T.M. Dridze [8],
E.S. Kubryakova [12], U. Kinch [32], however, direct understanding of the impact of generated texts on the reader has become the subject of systematic analysis only in recent years in the context of studies of algorithmic communication and digital rhetoric, while
A.R. Aidagulova [2], R.E. Telpov, S.V. Lartsina [25] record repetitions and a decrease in coherence, I.A. Turkulets [26], L.S. Kaladze, Ya.M. Vorsunov [10] point to the compositional standardization of machine texts,
E.S. Studenikina [22], M.A. Bagrova [4] emphasize the limitations of neural networks in reproducing complex syntactic structures, whereas
V.S. Pavlov [20], A.A. Spiridanchuk, T.V. Mamaeva [21], M.V. Kuzmenko [13],
E.V. Figura [27], E.V. Osetrova, A.V. Sedova [19] analyze the risks, limitations and communicative features of generation, which indicates a significant development of the formal linguistic aspect with a persistent deficit of complex psycholinguistic studies of the cognitive impact of algorithmically created text.

At the same time, the psychological dimension of the problem is revealed in the works of B.M. Velichkovsky [5], who considers artificial intelligence as a factor in the transformation of cognitive processes, E.S. Sukhareva [24], who analyzes the change in cognitive operations when interacting with algorithmic agents, M.K. Akimova and S.V. Persiyantseva [3], who emphasize the difference in the nature of creativity and intentionality of humans and machines, as well as A.K. Suyundukova and L.N. Uvarova [23],
A.N. Lebedeva [14], E.N. Davydenko and E.S. Okuneva [6], B.A. Adambekova and
G.M. Sydykova [1], M.O. Krukovsky [11], who consider artificial intelligence as a factor in changing the structure of cognitive activity, the mechanisms of attribution of intentions, the motivational sphere and even anthropogenesis, which emphasizes the need to integrate linguistic and psychological analysis in the study of generated texts.

 

Literature review

Texts generated by artificial intelligence offer specific linguistic models and possess specific stylistic features that manifest themselves in scientific, artistic, and journalistic genres, forming a special type of speech activity that differs in its structural, semantic, and pragmatic parameters from human-created texts. It is precisely in psycholinguistic terms that it is fundamentally important to establish what these differences consist of, how they are reflected in the mechanisms of perception, interpretation, and understanding of information by readers, and how they influence their cognitive strategies, thinking, consciousness, and the formation of mental representations of the text.

In psycholinguistics, text is traditionally interpreted as a product of speech activity, reflecting the interaction of the linguistic system and the cognitive structures of the individual, while, as G.M. Nurullina and L.A. Usmanova emphasize, its perception is an act of cognition and creativity, correlated with the mental projection of the content in the consciousness of the addressee
[17, p. 73], therefore, the artificially generated text turns out to be a specific object of analysis, since it is formed in the absence of anthropological intention and a different nature of the text generation mechanism.

A.R. Aydagulova, studying ChatGPT texts, notes repetitions, duplication of thoughts, fragmentary coherence, superficiality of examples and the absence of correct citations [2, p. 155], which, from a psycholinguistic point of view, disrupts coherence and hinders the formation of a holistic cognitive scheme, while R.E. Telpov and S.V. Lartsina identify an increased frequency of heading vocabulary and a specific distribution of thematic groups [25, p. 48], creating the effect of thematic concentration with limited conceptual novelty. I.A. Turkulets describes a compositional model with an introductory paragraph, a numbered list and a brief summary [26, p. 61], which facilitates initial perception, but reduces cognitive strain and the depth of material processing.

E.V. Osetrova and A.V. Sedova, relying on the communicative approach, note the circularity of the theme, the multiplicity of rhemes without further thematization and insufficient consideration of the factors of the author and the addressee [19, pp. 49–50], which complicates the construction of a hierarchy of meanings, and M.N. Cherkasova and A.V. Taktarova interpret machine text as an «artificial speech act» devoid of intentionality and empathy [29, p. 2554], as a result of which the center of meaning generation shifts to the interpretative activity of the reader.
E.V. Figura emphasizes the stereotyped nature and limitations of syntactic and connotative means [27, p. 67], which leads to the reduction of the figurative-associative component, while N.D. Maltsev points out the dependence of the result on the generation scenario and query parameters [16, p. 458], and N.V. Spiridanchuk and
N.A. Mamaeva connects the quality of the text with the algorithm for forming a prompt [21, p. 446], thereby actualizing the problem of distributing cognitive responsibility.

A comparative analysis by M.V. Khokhlova and M.V. Koryshev demonstrates formal structuring according to the IMRAD model with superficiality and repetitions [28, p. 1002], which is consistent with the conclusion
M.V. Dudorova on the difficulties of neural networks at the level of coherence and completeness with the successful implementation of cohesion [9, p. 766], while
N.N. Olomskaya and E.A. Yurova identify markers of generation in media discourse [18], and M.V. Kuzmenko emphasizes the risks of hallucinations and plagiarism [13], which is directly related to the psycholinguistic problem of trust and epistemic evaluation of algorithmic text.

At the same time, turning to psychological research allows us to deepen the analysis and go beyond a purely linguistic description. Thus, B.M. Velichkovsky pointed out that artificial intelligence should be considered in the context of the general theory of cognitive processes, since human interaction with intelligent systems transforms the structure of attention, memory, and thinking [5, pp. 12–88]; E.S. Sukhareva emphasizes that the intersection of artificial intelligence and human thinking leads to a redistribution of cognitive functions and a change in problem-solving strategies [24, pp. 231–234], which is directly related to the peculiarities of perception of algorithmically created text. M.K. Akimova and S.V. Persiyantseva note the difference in the nature of human intellectual abilities and artificial intelligence, emphasizing the latter's lack of reflexivity and genuine creativity [3, pp. 281–283], which explains the formal imitation of semantic depth in the absence of personal experience. A.K. Suyundukova and L.N. Uvarova analyzes the psychological aspects of the functioning of artificial intelligence, including the problem of trust in the algorithmic source and the peculiarities of attributing intentions to a machine [23, pp. 24–29], and A.N. Lebedev considers artificial intelligence as a factor in changing the cognitive practices of modern humans [14, pp. 6–22]. E.N. Davydenko and E.S. Okuneva study human-machine communication, noting the phenomenon of attributing subjectivity to an algorithmic agent [6, pp. 133–141], which enhances the effect of pseudo-intentionality of the text. B.A. Adambekova and G.M. Sydykova emphasize the influence of artificial intelligence on the motivational and value sphere of an individual [1, pp. 103–109], and M.O. Krukovsky considers AI as a factor of anthropogenesis, capable of transforming the cognitive and communicative structures of society [11, pp. 66–71].

Thus, the theoretical review shows that, despite the development of the structural-linguistic analysis of texts generated by artificial intelligence, it is the psycholinguistic dimension, including the study of cognitive mechanisms of perception, the characteristics of the formation of mental models, the processes of attribution of authorship and trust, as well as the influence of algorithmic speech on the thinking and consciousness of the addressee, that requires further systematic development, especially in the context of the analysis of texts on the Zen platform, where machine generation is increasingly becoming a part of everyday media communication.

 

Research results

Psycholinguistic text analysis differs fundamentally from linguistic analysis proper in that its focus is not only on the structure of verbal material as a system of linguistic units and the rules for their organization, but primarily on the processes of text generation, perception, interpretation, and assimilation in the human mind, as well as the cognitive, motivational, and emotional mechanisms that mediate the interaction between the verbal message and the recipient's mental structures. While linguistic analysis captures the categorical characteristics of a text, such as cohesion, composition, lexical organization, and syntactic structure, psycholinguistic analysis shifts the emphasis to how these characteristics function in the reader's cognitive system, what mental models they initiate, what cognitive load they create, and how they influence the processes of comprehension, memorization, and evaluation of information.

This is precisely why traditional linguistic parameters of text analysis require rethinking into psycholinguistic indicators, since their significance is determined not so much by the description of the linguistic form, but by identifying the nature of the impact on the cognitive system of the addressee, as a result of which thematic-semantic coherence is interpreted as a condition for the formation of a holistic mental model of the situation, lexical variability as a factor in the activation of semantic memory networks and an indicator of trust in the message, syntactic complexity as an indicator of the distribution of cognitive load and the involvement of working memory, and genre structure as a mechanism for regulating reader expectations and strategies for processing information in the digital environment.

The development of parameters for the psycholinguistic analysis of texts generated by artificial intelligence on digital platforms presupposes a distinction between the purely linguistic and psycholinguistic approaches, since the former focuses on the description of the lexical composition, syntactic organization, and composition, while the latter encompasses the processes of generation, interpretation, and emotional experience of the text in the consciousness of the addressee. In this regard, the theme-rheme dynamics is considered as an indicator of the progressive development of meaning, where excessive repetitions cause cognitive inhibition in the context of fragmented digital reading, and lexical and syntactic patterns can create the effect of primitivization of content, or, conversely, excessive complexity hinders the integration of information and reduces the rate of perception.

Discursive-pragmatic characteristics, including markers of addressability and engagement strategies, are understood as factors in the formation of a communicative framework and a sense of interactivity in platform discourse, while emotional-evaluative means determine the degree of the reader's affective participation and the level of trust in the message, while the genre-compositional organization controls the rhythm of information processing and the distribution of attention, and markers of algorithmic generation, such as template transitions and circularity of conclusions, affect the cognitive assessment of credibility, prompting the addressee to adjust the interpretation strategy depending on the recognizable signs of the machine origin of the text.

Thus, the identified psycholinguistic parameters allow us to consider the artificial intelligence-generated texts of the Zen platform as a special type of media speech, in which formal coherence can be combined with semantic superficiality, and genre ordering does not guarantee cognitive depth, which necessitates a comprehensive study not only of the text structure, but also of the cognitive and emotional processes that it initiates in the consciousness of the mass addressee.

Taking into account the theoretical foundations outlined, the parameters of the psycholinguistic analysis of AI-generated texts published on Zen platforms can be presented in Table 1.

Table 1. Parameters of psycholinguistic analysis of AI-generated texts published on Zen platforms«

Analysis parameter	Substantive characteristics	Psycholinguistic justification
Cognitive-semantic coherence	Logical coherence of fragments, sequence of disclosure of the topic, presence of a single semantic line	Ensures the formation of a holistic mental model and influences the depth of understanding and memorization
Theme-rhematic dynamics	The ratio of topic and new information, the absence of excessive repetitions	Determines the progressive movement of meaning and prevents cognitive inhibition
Lexical activation variability	Frequency of repetitions, use of synonyms, degree of cliché	Associated with the activation of semantic networks in memory and the feeling of naturalness of speech
Syntactic cognitive load	Types of sentences, variety of constructions, ratio of simple and complex structures	Affects the processing of information in working memory and the level of mental stress
Pragmatic attribution of intention	The presence of address markers, imitation of dialogue, engaging constructions	Creates a sense of communicative interaction and influences trust in the text
Emotional-evaluative involvement	Use of evaluative vocabulary and expressive means	Determines the degree of empathic response and emotional activation
Genre-compositional regulation of expectations	Organization of text, presence of standard blocks	Forms a reading strategy and manages audience expectations
Indicators of algorithmic predictability	Stereotyped formulations, circular conclusions, declarativeness	Affect the recognition of machine generation and epistemic assessment of reliability
Level of conceptual depth	The presence of cause-and-effect relationships, argumentation, and analytical nature	Determines the possibility of constructing complex cognitive structures
Receptive adaptability to the media environment	Format compliance with digital reading dynamics	Affects the speed of perception, attention retention and satisfaction with reading

Rethinking linguistic parameters in a psycholinguistic manner allows us to go beyond the description of formal characteristics and focus on their functional role in the addressee's cognitive system, which is especially significant in the digital media environment, where text not only informs but also shapes ways of thinking, evaluating, and interpreting reality.

As part of the empirical stage of this study, the texts posted on the Zen platform were preliminarily verified using the service’s AI detector. ETXT.RU, which is considered one of the most authoritative systems for automatically identifying algorithmically generated materials in the professional community of digital editors and content specialists. It ranks second only to Antiplagiat.ru in terms of prevalence and trust, making its diagnostic indicators sufficiently reliable for inclusion in scientific analysis tools. This service operates by comparing probabilistic models of the distribution of lexical and syntactic patterns, repetition frequencies, and typical semantic relationships characteristic of generative language models. The resulting indicator reflects the probability of the text's algorithmic origin. While not definitive proof in research logic, it does create a statistically sound basis for further psycholinguistic study of the material as a product of machine generation.

Following the initial review, the article "Why Do We Love to Suffer? How 'Office Romance' and 'The Irony of Fate' Taught Us That Happiness Must Be Earned Through Pain"«[1], published on the Zen channel "Notes of an Actor", was identified by the system as a text with a high probability of algorithmic generation, and only after this status was recorded, its detailed psycholinguistic analysis was carried out in comparison with the author's publication of the same channel "Why go to the theater?"« [2], created several years earlier and showing no signs of machine origin.

As part of the study, the analysis of two texts posted on the Zen platform was carried out based on a system of psycholinguistic description parameters, focused not on recording formal linguistic characteristics, but on reconstructing those cognitive, affective and communicative processes that are initiated in the recipient's consciousness at the moment of interaction with the media text, since in the conditions of the digital environment, it is precisely receptive mechanisms that largely determine the effectiveness and depth of impact.

Looking at the generated article through the prism
cognitive-semantic coherence, it should be noted that the text builds a relatively holistic mental model of the interpretation of Soviet cinema as a source of the passive model of love, however, this integrity is ensured mainly by the repetition of one dominant thesis in various formulations, whereas
The cause-and-effect relationships between the cultural context, historical conditions and the individual psychology of the viewer remain weakly explicated, as a result of which the resulting cognitive scheme turns out to be linear and simplified.

The theme-rheme dynamics in the analyzed material demonstrates the cyclical nature of semantic development, in which rheme positions do not expand the conceptual field, but vary the initial statement about the connection between love and suffering, which, from a psycholinguistic point of view, creates the effect of movement in the absence of a real increase in semantic complexity and can lead to a decrease in cognitive motivation upon repeated encounter with similar semantic structures.

Lexical activation variability in this text is aimed at triggering stable cultural scenarios and collective representations, since recognizable names of characters, well-known film images and clichéd journalistic formulas are used, activating already formed semantic networks in the long-term memory of the addressee, however, an individualized speech profile is practically not manifested, which limits the depth of personal identification and reduces the effect of the presence of a specific subject of the utterance.

From the point of view of syntactic cognitive load, the text is characterized by fragmentation and the predominance of short segmented constructions, which facilitates the processing of information in the conditions of on-screen reading and reduces the demands on working memory, but at the same time hinders the formation of expanded argumentative chains capable of initiating more complex processes of analytical thinking and conceptual integration.

The pragmatic attribution of intention in the generated material is formal in nature, since the appeal to the reader, rhetorical questions, and invitation to discussion create the appearance of dialogicity, but behind them no stable personal position is found, as a result of which the addressee attributes to the text a media platform intention rather than an individual authorial intention, which affects the level of trust and the strategy of interpretation.

Emotional-evaluative involvement is achieved through declarative generalizations and contrasting formulations; however, the affective energy of the text is distributed evenly and is not supported by the personal experience of the subject of speech, as a result of which the emotional response is of the nature of short-term reactivity, rather than stable empathic involvement.

In this case, the genre-compositional regulation of expectations is built on the model of digital journalistic material with a clear division into semantic blocks, insertions of examples, and a final call to interaction, which corresponds to the media habits of the audience and ensures high receptive accessibility, but simultaneously enhances the feeling of algorithmic assembly from functionally heterogeneous modules.

Indicators of algorithmic predictability are manifested in the repetition of structures, in the symmetrical distribution of evaluative formulas, and in the use of typical transitions between semantic fragments, which creates a feeling of structural neatness, but reduces the likelihood of an unexpected conceptual turn that can transform the reader’s initial cognitive frame.

The level of conceptual depth in the generated text remains moderate, since the analysis is limited to an illustrative comparison of film images without reaching a systemic understanding of the psychological mechanisms of the internalization of cultural models of love, and therefore the resulting mental construct turns out to be relatively flat and does not require significant intellectual processing.

Receptive adaptability to the media environment, on the contrary, is expressed quite clearly, since the structure of the material, its rhythm and linguistic organization are oriented towards short sessions of attention, characteristic of digital content consumption, which ensures ease of perception and rapid integration of information into existing cognitive schemes.

An analysis of the author's article using the same parameters reveals a different configuration of psycholinguistic characteristics, since
Cognitive-semantic coherence is formed through the consistent inclusion of personal experience and professional reflection, thanks to which a multi-layered mental model of theatrical communication is built in the consciousness of the addressee, including emotional, aesthetic and social levels.

The theme-rheme dynamic here is progressive, as each new fragment does not repeat the original theme but concretizes and deepens it, expanding the conceptual field by introducing new aspects. This activates processes of semantic integration and maintains cognitive tension throughout the text. Lexical activation variability is expressed through individual intonations, elements of professional discourse, and explicit self-identification markers, thereby activating the addressee's semantic networks not only at the level of cultural stereotypes but also at the level of personal experience of interaction with a specific speaker.

The syntactic cognitive load in the author's text is higher, as sentences often develop into argumentative constructions with clarifications and explanations. This requires more intensive working memory, but it facilitates the formation of stable conceptual connections and a deeper understanding of the content. The pragmatic attribution of intention in this case is clearly related to the personal position of the author, who openly expresses their professional role and emotional attitude toward the subject. Consequently, the addressee perceives the text as the authentic utterance of a specific person, rather than as a neutral media product.

Emotional evaluative engagement is reinforced by personal confessions and descriptions of individual experiences, which strengthens empathic identification and creates a lasting affective connection between the text and the reader. The genre-compositional regulation of expectations in the author's article is less standardized; transitions between semantic blocks are motivated by the logic of personal reasoning, not solely by the requirements of the platform format, which reduces the feeling of predictability and enhances the impression of lively speech improvisation. Indicators of algorithmic predictability are practically absent, as the text contains subjective assessments, individual emphases, and semantic deviations that do not fit into the typical media publication template.

The level of conceptual depth is higher here, as the discussion of theater extends beyond the description of external effects and addresses issues of actor-audience interaction, the social function of art, and the creator's personal responsibility, which requires more complex cognitive processing from the recipient. Receptive adaptability to the media environment is maintained, but it is not achieved through a simplification of content, but rather through a combination of personal sincerity and structural clarity, ensuring the text remains accessible without losing its analytical richness.

Thus, the application of the identified parameters allows us to identify fundamental differences in the psycholinguistic organization of algorithmically generated and author-generated text, confirming the effectiveness of the chosen analytical model, which focuses on studying the cognitive and affective mechanisms of media speech perception in the digital space. A comparative analysis of these two articles is presented in Table 2.

Table 2. Comparative analysis of the generated and author's article of the Zen channel "Notes of an Actor"«

Criterion	Generated article "Why do we love to suffer?"«	Author's article "Why go to the theater?"«
Cognitive-semantic coherence	A formally coherent structure, ensured by the repetition of the dominant thesis; the mental model is linear, without significant complication of cause-and-effect relationships	A holistic and multi-level structure formed through personal experience and reflection; the mental model includes professional, emotional, and social levels
Theme-rhematic dynamics	Cyclical, rhematic positions vary the original theme without significantly expanding the conceptual field	Progressively, each rhematic position deepens and concretizes the original problematic
Lexical activation variability	Activation of cultural stereotypes and collective memory; prevalence of clichéd media patterns; weak individualization of the subject of speech	Individualized vocabulary, elements of professional discourse and personal self-identification; activation of personalized semantic networks
Syntactic cognitive load	Predominance of short, segmented constructions; reduced load on working memory; limited opportunities for complex argumentation	Expanded argumentative constructions; increased cognitive load, promoting deeper processing of information
Pragmatic attribution of intention	Formal imitation of dialogicity; the intention is perceived as media-platform, not personal	Clearly attributed personal intention; expressed professional position and genuine communicative contact
Emotional-evaluative involvement	Declarative, based on cultural generalizations; limited empathic potential	Personally colored, supported by experience; high degree of empathic identification
Genre-compositional regulation of expectations	Standardized modular structure; compliance with digital consumption algorithms; a sense of template assembly	The logic of the composition is motivated by the internal development of thought; reduced predictability of structural transitions
Indicators of algorithmic predictability	Repetition of formulas, symmetry of evaluative structures, typical transitions; high structural predictability	The presence of subjective accents, semantic deviations and individual intonations; low degree of predictability
Level of conceptual depth	Moderate; illustrative analysis without reaching a systemic psychological understanding	High; formation of a complex conceptual model of theatrical communication
Receptive adaptability to the media environment	High adaptation to screen reading; orientation to short attention cycles	Balanced adaptation: accessibility while maintaining analytical richness

The table demonstrates that the differences between the texts are manifested not so much at the level of formal linguistic characteristics, but in the nature of the cognitive organization of the content, the degree of personalization of the subject of speech and the intensity of the psychological impact on the addressee, which confirms the feasibility of using a psycholinguistic model of analysis in the study of algorithmically and anthropogenically created media texts.

In order to verify the previously obtained results and clarify the specifics of the psycholinguistic organization of the algorithmically generated media text, we conducted an additional analysis of publications posted on the Zen platform on the Kulturologiya channel, which has an audience of 823 thousand subscribers. The preliminary stage of the study included checking materials from recent months using the service's AI detector. ETXT.RU, which resulted in the article «The wound is bleeding, I wouldn’t wish it on my enemy»: Why is Uspenskaya worried about her 36-year-old daughter?» [3] was identified as a text with a high probability of machine generation, which allowed it to be considered as a representative example of algorithmically constructed discourse and to carry out its subsequent analysis from a strictly psycholinguistic perspective.

It is important to emphasize that the Kulturologiya channel was initially positioned as an educational media project focused on cultural and historical issues, but over time it transformed into a commercially oriented platform with pronounced elements of sensationalism, which manifested itself not only in a thematic shift to the private lives of public figures, but also in a change in communication strategy, and therefore a comparison with an earlier publication of the same channel, «The History of the Most Poignant Monument in the World: Shoes Remaining on the Banks of the Danube»[4], created during the period of absence of signs of algorithmic generation, seems methodologically justified.

In the article about Tatyana Plaksina, the parameter of cognitive-semantic coherence is realized mainly at the level of formal chronological coherence, but the semantic core remains fragmented due to the constant switching between facts, emotionally charged quotations and sensational details, as a result of which the theme-rhematic dynamics is built on the principle of increasing affective tension without expanding the analytical field, lexical variability is characterized by the dominance of expressive units such as «nightmare» and «terrible details», activating threat scenarios and reducing the critical distance, the syntactic load is minimized by short sentences and paragraphs, which facilitates screen reading, but prevents the formation of hierarchical connections, the pragmatic intention is focused on maintaining attention to the dramatic narrative, emotional-evaluative involvement prevails over reflection, the genre-compositional organization is realized according to the model of clip montage with subtitles, and the repetition of structures and standardized transitions indicate algorithmic predictability, ensuring ease of perception and clickability with limited conceptual depth.

A different psycholinguistic configuration is found in the author's article about the memorial on the banks of the Danube, where cognitive-semantic coherence is built around a specific historical event and its symbolic representation with the inclusion of the political context and the actions of the Arrow Cross party; the theme-rhematic dynamics are progressive in nature and form a cause-and-effect model; lexical expressiveness is combined with a factual dominant without sensationalism; the syntactic structure is complicated by expanded sentences with numerical and historical clarifications, which increases the cognitive load and promotes long-term memorization; the pragmatic attitude is of an educational nature; emotional involvement is formed through the symbol of collective memory; the genre organization follows the logic of historical narrative; indicators of algorithmic predictability are practically absent; and the conceptual depth of the text stimulates reflection on the mechanisms of violence and memory.

Thus, a comparative psycholinguistic analysis demonstrates that algorithmically generated text predominantly activates affective mechanisms and is focused on a superficial emotional reaction, whereas the author's publication forms a complex cognitive model of a historical event, ensuring a deeper and more lasting impact on the consciousness of the addressee in the context of digital media communication.

Table 3. Comparative analysis of the generated and original article of the Zen channel "Cultural Studies"«

Parameter	Generated article about Plaksina	An original article about the memorial on the Danube
Cognitive-semantic coherence	Formal coherence is supported by chronology and repetition of the central motif, but the cause-and-effect relationships are fragmented, making it difficult to construct a coherent mental model.	The semantic structure is hierarchical and conceptually organized, cause-and-effect relationships are explicated, which ensures the formation of a stable mental model of the event
Theme-rhematic dynamics	The development of the theme is carried out through affective build-up; rhematic elements enhance the emotional background without significantly expanding the analytical field.	Rhematic positions consistently clarify and deepen the content, forming a progressive semantic development
Lexical activation variability	Predominance of expressive and clichéd units that activate stereotypical scenarios of suffering and sensation; limited individualization	A combination of factual and evaluative vocabulary, a variety of nominations, the absence of excessive clichés; a clear authorial position
Syntactic cognitive load of information	Short sentences and segmented paragraphs reduce the load on working memory, promoting fast screen reading.	Expanded syntactic constructions increase cognitive load and stimulate analytical processing
Pragmatic attribution of intention	The intention is to hold attention and enhance dramatization; the communicative attitude is focused on the consumption of sensational content.	The intention is educational, aimed at a joint understanding of historical material; the author's subjective position is expressed
Emotional-evaluative involvement	The affective tone dominates, a stable emotional background is formed without deep reflection	Emotionality is combined with analytical skills, empathy is supported by factual argumentation
Cognitive depth	Limited, emotional impact prevails	High, a cause-and-effect model is formed
Genre-compositional regulation of expectations	Modular structure with subheadings and clip-based organization of material; a sense of template assembly	The composition is subordinated to the logic of the historical narrative, the structural transitions are motivated by the development of thought
Indicators of algorithmic predictability	Repetition of formulas, symmetry of emotional accents, standardized transitions between blocks	Individualized intonation shifts, lack of template symmetry, variability of compositional solutions
Level of conceptual depth	Limited to the description of individual episodes without integration into a broad socio-psychological framework	High level of conceptualization, inclusion of historical, social and symbolic context
Receptive adaptability to the media environment	Maximum adaptation to the digital format, focusing on short attention spans and clickability	A balanced adaptation that combines accessible presentation with analytical richness

The presented table allows us to clearly demonstrate that the differences between the texts are manifested not only in style and subject matter, but above all in the nature of the cognitive organization of the content, the degree of conceptual depth and the model of psychological impact on the addressee, which confirms the productivity of applying psycholinguistic optics to the analysis of media texts of various origins.

 

Conclusion

The conducted research allowed us to establish that texts generated by artificial intelligence and posted on the Zen platform exhibit a stable set of psycholinguistic characteristics, among which the most prominent are a high degree of structural standardization, repetition of thematic-rheme patterns, limited lexical variability, and a pronounced focus on the affective involvement of the addressee. As a result, a specific model of reception is formed, in which superficial cognitive processing of information dominates, based on the rapid recognition of template semantic blocks and the activation of stereotypical semantic scenarios, while deep integration of content, which involves the construction of a complex mental model and the establishment of extensive cause-and-effect relationships, is reduced.

Consequently, the cognitive-semantic coherence of such texts is supported primarily by formal means of coherence, but conceptual depth often remains limited, since the development of the theme is carried out according to the principle of increasing emotional tension, rather than analytical complexity, which is reflected in the cyclical theme-rheme dynamics, reproducing the original thesis in new formulations without a significant expansion of the interpretative field, while lexical activation variability is often reduced to the reproduction of clichéd evaluative units that initiate a quick emotional response in the reader and create the effect of recognition, but at the same time reduce the likelihood of critical reflection.

The syntactic organization of generated publications is generally aimed at reducing cognitive load through the predominance of short sentences and segmented composition, which facilitates the processing of material in the conditions of screen reading and corresponds to the dynamics of digital information consumption. However, such a structural strategy limits the possibilities for expanded argumentation and the formation of hierarchically organized conceptual structures in the consciousness of the addressee, and indicators of algorithmic predictability, manifested in the symmetry of compositional blocks and the repetition of transition formulas, enhance the feeling of a stereotyped and standardized speech behavior.

In contrast, the author's publications demonstrate a more complex argumentative organization, individualized intonation, and a higher level of conceptual richness, which is expressed in an expanded theme-rheme dynamic, variable lexical structure, and syntactic multi-levelness, stimulating active processing of information, the formation of stable semantic connections, and the inclusion of the reader's personal experience in the interpretation process, due to which the receptive strategy shifts from a quick emotional response to analytical comprehension and long-term memorization.

The obtained results indicate a fundamentally different direction of the psycholinguistic impact of these types of texts on the consciousness of the audience, since algorithmically created materials predominantly activate the mechanisms of affective attribution and accelerated recognition of familiar media patterns, while the author's texts initiate more complex cognitive operations associated with the reconstruction of the author's intention, the evaluation of argumentation and the formation of a critical position, which makes further experimental study of the influence of generative content on the processes of understanding, memorization and the development of critical thinking relevant.

Literature

1. Adambekova B. A., Sydykova G. M. Modern problems of psychology: artificial intelligence in light of psychological aspects // Bulletin of the Kyrgyz National University named after Jusup Balasagyn. 2024. No. 1 (117). P. 103–109. DOI 10.58649/1694–8033-2024-1 (117) -103–109.
2. Aidagulova A.R. Features of texts generated by artificial intelligence // Bulletin of the Bashkir State Pedagogical University named after M. Akmulla. 2023. No. 4(72). P. 154–156.
3. Akimova MK, Persiantseva S.V. Analysis of the capabilities of the human intellectual sphere and artificial intelligence // Artificial intelligence and social development: new opportunities and obstacles: collection of materials, Moscow, February 21, 2022. Moscow: OOO MAX Press, 2022. Pp. 281–283.
4. Bagrova N.A. A Plain and Simple: Idiomatic Expressions of the Russian Language Through the Eyes of a Neural Network // Barkhudarov Readings: Coll. of Scientific Articles of Participants of the International Scientific and Methodological Conference, Dedicated to the 130th Anniversary of the Birth of S.G. Barkhudarov (May 3–5, 2024). Yerevan: RAU, 2024. Pp. 25–28.
5. Velichkovsky B.M. Psychological aspects of artificial intelligence // News of artificial intelligence. 1995. No. 1. P. 12–88.
6. Davydenko E.N., Okuneva E.S. Man and Machine: Psychological Aspects of Communication with Artificial Intelligence // Actual Problems of Humanitarian Knowledge: Past and Present: Proceedings of the VI International Scientific Conference, Moscow, November 24, 2024. Moscow: Research Institute of History, Economics and Law, 2024. Pp. 133–141.
7. Dijk T.A. Language. Cognition. Communication: translation from English / T. A. Dijk; ed. V. I. Gerasimov; compiled by V. V. Petrov; introduction by Yu. N. Karaulova, V. V. Petrova. 2nd ed. Moscow: URSS: LENAND, 2015. 310 p.
8. Dridze T.M. Textual activity in the structure of social communication: problems of semiosociopsychology / T. M. Dridze; ed. I. T. Levykin. Moscow: Nauka, 1984. 267 p.
9. Dudorova M.V. The Specifics of Manifestation of Text Categories in AI Essays // Communicative Research. 2025. Vol. 12, No. 4. Pp. 758–773. DOI 10.24147/2413-6182.2025.12(4).758-773.
10. Kaladze V.A., Vorsunov A.A. A study of the differences between texts created by a generative network and by humans // Problems of the theory and practice of innovative development and integration of modern science and education: Proceedings of the V International Scientific and Practical Conference, Moscow, February 14, 2024. Moscow: PRINTIKA LLC, 2024. Pp. 124–128.
11. Krukovsky M.O. Artificial intelligence as a factor of anthropogenesis: psychological aspect // Scientific discussions of modern Russian society: social and humanitarian paradigms: collection of scientific articles. Moscow: OOO NPK Nauka, 2025. Pp. 66–71.
12. Kubryakova E.S. Language and knowledge: on the way to gaining knowledge about language: parts of speech from a cognitive point of view. The role of language in understanding the world / E. S. Kubryakova; Institute of Linguistics, Russian Academy of Sciences. Moscow: Languages of Slavic Culture, 2004. 555 p.
13. Kuzmenko M.V. Artificial Intelligence in Research and Academic Writing: An Analytical Review of Application Areas and Limitations // Continuous Education: 21st Century. 2025. Vol. 13, No. 3. DOI 10.15393/j5.art.2025.10846
14. Lebedev A.N. Artificial Intelligence and Psychology // Scientific Notes of the Institute of Psychology of the Russian Academy of Sciences. 2023. Vol. 3, No. 2(8). P. 6–22.
15. Leontiev A.A. Psychology of Communication: A Textbook for University Students Majoring in Psychology / A. A. Leontiev. 3rd ed. Moscow: Smysl: Academia, 2005. 365 p.
16. Maltsev N.D. Structural and philological features of text generative neural networks // Neofilologiya. 2024. Vol. 10, No. 2. Pp. 452–464. DOI 10.20310/2587-6953-2024-10-2-452-464.
17. Nurullina G.M., Usmanova L.A. Psycholinguistic aspect of studying text as a product of speech activity // Philology and Culture. 2018. No. 3(53). P. 72–76
18. Olomskaya N.N., Yurova E.A. Linguopragmatic features of generated text in the media discourse of social networks (using the example of texts about the 2023 gubernatorial elections in the Nizhny Novgorod region) // Russian Social and Humanitarian Journal. 2025. No. 2. DOI 10.18384/2224-0209-2025-2-1649.
19. Osetrova E.V., Sedova A.V. Characteristics of the generated text: linguistic and socio-communicative analysis // Siberian Philological Forum. 2025. No. 2(31). P. 45–55. DOI 10.24412/2587-7844-2025-2-45-55.
20. Pavlov A.A. «Scientific» texts created using artificial intelligence: current state and development prospects // Methodological problems of civilistic studies. 2024. No. 6. pp. 111–127. DOI 10.33397/2619-0559-2024-6-6-111-127.
21. Spiridanchuk N.V., Mamaeva N.A. Features of generating text content for articles using neural networks using ChatGPT // Informatization of education and e-learning methods: digital technologies in education: Proceedings of the VIII International Scientific Conference, Krasnoyarsk, September 24–27, 2024. Krasnoyarsk: Krasnoyarsk State Pedagogical University named after V.P. Astafiev, 2024. pp. 444–451.
22. Studenikina K.A. Evaluation of the language ability of neural models using the material of predicative agreement in the Russian language // Proceedings of the Institute for System Programming of the Russian Academy of Sciences. 2022. No. 34(6). pp. 179–184. DOI15514/ispras-2022-34(6)-14.
23. Suyundukova A.K., Uvarova L.N. Psychological aspects of artificial intelligence // Kazakhstan Science Journal. 2018. Vol. 1, No. 1(1). P. 24–29.
24. Sukhareva E.S. Intersections of Artificial Intelligence and Human Thinking // Language and Personality: Sociocultural and Psychological Transformations: Proceedings of the International Scientific and Practical Conference, Moscow, December 15–16, 2023. Moscow: Agency for Social and Humanitarian Technologies, LLC, 2023. Pp. 231–234.
25. Telpov R.E., Lartsina S.V. Typical differences between natural and neural network-generated texts in a quantitative aspect // Scientific Dialogue. 2023. Vol. 12, No. 7. Pp. 47–65
26. Turkulets I.A. Compositional Features of Texts Generated by ChatGPT as a Marker of Dependence on Students' Work Performance // Legal Reality in the Context of Digitalization of Society: Proceedings of the All-Russian Scientific and Practical Conference, Khabarovsk, November 9–11, 2023. Khabarovsk: Far Eastern State Transport University, 2023. Pp. 59–68
27. Figure of E.V. Text generation as an object of linguistic research // Bulletin of the Russian New University. Series: Man in the Modern World. 2025. No. 3. pp. 64–70. DOI 10.18137/RNU.V925X.25.03.P.064.
28. Khokhlova M.V., Koryshev M.V. Comparative analysis of generated and original abstracts of scientific articles on philological topics // Scientific and Technical Bulletin of Information Technologies, Mechanics and Optics. 2024. Vol. 24, No. 6. Pp. 999–1006. DOI 10.17586/2226-1494-2024-24-6-999-1006.
29. Cherkasova M.N., Taktarova A.V. Artificially generated academic text (linguopragmatic aspect) // Philological sciences. Theoretical and Practical Issues. 2024. Vol. 17, No. 7. Pp. 2551–2557. DOI 10.30853/phil20240363.
30. Beaugrande R.-A. de, Dressler WU. Konzepte der Sprach- und Literaturwiss. Einführung in die Textlinguistik / R.-A. de Beaugrande, WU Dressler. Tübingen: Niemeyer, 1981. XIII, 290 S
31. Halliday MAK, Hasan R. Cohesion in English: A Study Guide / MAK Halliday, R. Hasan. Harlow: Longman, 1997. XV, 374 p.: tabl. (English language series / Gen. ed. R. Quirk; 9). Bibliography: pp. 357–366
32. Kintsch W. The role of knowledge in discourse comprehension: a construction-integration model / W. Kintsch // Psychological Review. Vol. 95, No. 2. P. 163–182.

 

References

1. Adambekova BA, Sydykova GM. (2024) Modern problems of psychology: artificial intelligence in the light of psychological aspects // Bulletin of the Kyrgyz National University named after Zhusup Balasagyn. No. 1(117). P. 103–109. DOI 10.58649/1694-8033-2024-1(117)-103-109
2. Aidagulova AR. (2023) Features of texts generated by artificial intelligence // Bulletin of the Bashkir State Pedagogical University named after M. Akmully.
   No. 4(72). S. 154–156.
3. Akimova MK, Persjanceva SV. (2022) Analysis of the abilities of the human intellectual sphere and artificial intelligence // Artificial intelligence and social development: new opportunities and obstacles: collection of materials, Moscow, February 21, 2022. Moscow: OOO MAKS Press. P. 281–283
4. Bagrova NA. (2024) Easy to understand: idiomatic expressions of Russian language through the eyes of neural networks // Barkhudarovskie chtenija: sb. nauch. st. pakistnikov mezhdunar. nauch.-metod. konf., prioch. k k 130-letiju so dnja zhdenija SG Barkhudarova (3–5 May 2024). Yerevan: RAU. pp. 25–28.
5. Beaugrande R.-A. de, Dressler WU. (1981) Konzepte der Sprach- und Literaturwiss. Einfuhrung in die Textlinguistik [Concepts of linguistics and literary studies. Introduction to text linguistics]. Tubingen: Niemeyer. XIII, 290 p.
6. Cherkasova MN, Taktarova AV. (2024) Artificially generated academic text: linguistic and pragmatic aspect // Philological sciences. Issues of theory and practice. Vol. 17, No. 7, pp. 2551–2557. DOI 10.30853/phil20240363
7. Davydjenko EN, Okuhneva ES. (2024) Human and machine: psychological aspects of communication with artificial intelligence // Actual problems of humanitarian knowledge: past and present: Proceedings of the VI International Scientific Conference, Moscow, November 24, 2024. Moscow: Research Institute of History, Economics and Law. pp. 133–141
8. Dridze TM. (1984) Tekstovaja dejatel'nost' v strukture social'noj kommunikacii: problemy semiosotsiopsihologii [Text activity in the structure of social communication: problems of semiosocio-psychology] / ed. IT Levykin. Moscow: Nauka. 267 p.
9. Dudorova MV. (2025) Specifika projavlenija tekstovyh kategorij v II-sochinenijah [Specific features of text categories in AI-written works] // Kommunikativnye issledovanija. T. 12, No. 4. S. 758–773. DOI 10.24147/2413-6182.2025.12(4).758-773.
10. Figura EV. (2025) Text generation as an object of linguistic research // Vestnik of the Russian New University. Series: Man in the Modern World. No. 3. pp. 64–70. DOI 10.18137/RNU.V925X.25.03.P.064
11. Halliday MAK, Hasan R. (1997) Cohesion in English: uchebnoe posobie [Cohesion in English: textbook]. Harlow: Longman. XV, 374 p. : tabl. (English language series / Gen. ed. R. Quirk; 9). Bible: p. 357–366
12. Hohlova MV, Koryshev MV. (2024) Comparative analysis of generated and original scientific article abstracts in philology // Scientific and technical bulletin of information technologies, mechanics and optics. Vol. 24, No. 6, pp. 999–1006. DOI 10.17586/2226-1494-2024-24-6-999-1006
13. Kaladze VA, Vorsunov AA. (2024) Study of differences between texts created by the generative network and humans // Problems of theory and practice of innovative development and integration of modern science and education: Proc. B Intern. scientific-practical. conf., Moscow, February 14, 2024. Moscow: OOO PRINTIKA. Pp. 124–128.
14. Kintsch W. (1988) The role of knowledge in discourse comprehension: a construction-integration model // Psychological Review. Vol. 95, No. 2. P. 163–182.
15. Krukovskij MO. (2025) Artificial intelligence as a factor of anthropogenesis: psychological aspect // Scientific discussions of modern Russian society: socio-humanitarian paradigms: collection of scientific states. Moscow: OOO NPK Nauka. P. 66–71.
16. Kubrjakova E.S. (2004) Language and knowledge: on the way to gaining knowledge about language: parts of speech from a cognitive perspective. Role of language in cognition / Institute of Language Studies RAS. Moscow: Slavic Languages of Culture. 555 p.
17. Kuz'menko MV. (2025) Artificial intelligence in research activity and academic writing: analytical review of directions and limitations // Continuous education: XXI century. Vol. 13, No. 3. DOI 10.15393/j5.art.2025.10846
18. Lebedev AN. (2023) Artificial intelligence and psychology // Scientific notes of the Institute of Psychology of the Russian Academy of Sciences. Vol. 3, No. 2(8). P. 6–22
19. Leont'ev AA. (2005) Psychology of communication: textbook for university students in Psychology / 3rd ed. Moscow: Smysl: Academia. 365 p.
20. Mal'cev ND. (2024) Structural and philological features of text-generative neural networks // Neofilologija. Vol. 10, No. 2. Pp. 452–464. DOI 10.20310/2587-6953-2024-10-2-452-464
21. Nurullina GM, Usmanova LA. (2018) Psycholinguistic aspect of text as a product of speech activity // Philology and Culture. No. 3(53). P. 72–76
22. Olomskaja NN, Jurova EA. (2025) Linguistic and pragmatic features of AI-generated texts in social media discourse (example: gubernatorial elections in Nizhny Novgorod region, 2023) // Russian Social and Humanitarian Journal. No. 2. DOI 10.18384/2224-0209-2025-2-1649
23. Osetrova EV, Sedova AV. (2025) Characteristics of AI-generated texts: linguistic and socio-communicative analysis] // Sibirskij filologicheskij forum. No. 2(31). S. 45–55. DOI 10.24412/2587-7844-2025-2-45-55
24. Pavlov AA. (2024) AI-generated «scientific» texts: current state and development prospects // Methodological problems of civilistic research. No. 6. pp. 111–127. DOI 10.33397/2619-0559-2024-6-6-111-127
25. Spiridanchuk NV, Mamaeva NA. (2024) Features of article content generation using neural networks: example of ChatGPT // Informatization of education and methods of e-learning: digital technologies in education: Proceedings of the VIII International scientific conference, Krasnojarsk, September 24–27, 2024. Krasnojarsk: Krasnojarsk State Pedagogical University named after V. P. Astafev. pp. 444–451
26. Studenikina KA. (2022) Evaluation of neural models' linguistic ability on predictive agreement in Russian // Trudy Instituta sistemnogo programmingovanija RAS. No. 34(6). pp. 179–184. DOI 10.15514/ispras-2022-34(6)-14
27. Suhareva ES (2023) Intersections of artificial intelligence and human thinking // Language and personality: socio-cultural and psychological transformations: Proc. of the international scientific-practical conference, Moscow, December 15–16, 2023. Moscow: OOO ’Agency for Social and Humanitarian Technologies«. Pp. 231–234
28. Sujundukova AK, Uvarova LN. (2018) Psychological aspects of artificial intelligence // Kazakhstan Science Journal. Vol. 1, No. 1(1). P. 24–29
29. Tel'pov RE, Larcina SV. (2023) Typical differences between natural and neural network-generated texts in quantitative aspect // Scientific dialog. Vol. 12, No. 7. Pp. 47–65
30. Turkulets IA. (2023) Compositional features of ChatGPT-generated texts as a marker of students' non-independent work // Legal reality in the conditions of the digitalization of society: Proc. All-Russian scientific and practical conference, Khabarovsk, November 9–11, 2023. Khabarovsk: Far East State University of Communications. pp. 59–68
31. Van Dijk TA. (2015) Language. Cognition. Communication / transl. from English; ed. VI Gerasimov; comp. VV Petrov; introd. YN Karaulova, VV Petrov. 2nd ed. Moscow: URSS : LENAND. 310 p.
32. Velichkovskyj BM. (1995) Psychological aspects of artificial intelligence // News of artificial intelligence. No. 1. P. 12–88

[1] Why do we love to suffer? How "Office Romance" and "The Irony of Fate" taught us that happiness must be earned through pain. https://dzen.ru/a/aYxzVR5z_QdrocpM  (date accessed: 10.02.2026)
[2] Why go to the theater? URL: https://dzen.ru/a/W3qdkSk5vACpGTEv (date accessed: 10.02.2026)

[3] «"The wound is bleeding, I wouldn't wish it on my worst enemy": Why is Uspenskaya worried about her 36-year-old daughter? URL: https://dzen.ru/a/aZsDVVyVY149K4EG (date accessed: 10.02.2026)

[4] The story of the "world's most poignant monument": shoes left on the banks of the Danube URL: https://dzen.ru/a/XiQZIi_ahgCxz5J1 (date accessed: 10.02.2026)