This article uses an integrative theory-synthesis design rather than claiming a systematic review or reporting empirical evidence. Conceptual articles create value by connecting established theoretical lenses, specifying mechanisms, and identifying boundary conditions that are not visible within isolated research streams [13], [14]. Consistent with guidance for rigorous literature synthesis, the review prioritizes peer-reviewed research that directly informs the focal constructs and relationships rather than attempting an exhaustive count of publications [15]. The synthesis integrates five streams: Consumer co-creation in NPD, human–artificial intelligence augmentation, perceived control and creative self-efficacy, psychological ownership, and consumer authenticity. The objective is to explain a new phenomenon, the distribution of creative agency in GenAI-enabled product co-creation, using complementary theories with clearly differentiated roles.

NPD has traditionally relied on firms to identify needs, translate customer information into specifications, and move concepts through structured development stages [16–18]. Co-creation changes the consumer’s role from information source to active contributor. Consumers may generate ideas, evaluate alternatives, customize attributes, test prototypes, or participate in online innovation communities [5], [18], [19]. Toolkits and open-innovation arrangements can further distribute problem solving to users who possess need-related knowledge that is difficult for firms to observe directly [20], [21].

The marketing value of participation is not limited to improved need fit. Participation can create empowerment, increase identification with the offering, and alter evaluations of both the process and the outcome. Empowerment strategies increase product demand when consumers perceive that their decisions meaningfully influence what is offered [7]. Customization can produce an “I designed it myself” effect, while invested effort can increase valuation even when the objective quality of the result is held constant [8], [22]. Similar effects appear when consumers customize for themselves, participate in production, or evaluate products developed by other users [23–26].

These findings imply that co-creation contains at least three value layers. The first is instrumental: participation can improve preference fit. The second is experiential: the process can be enjoyable, competent, and autonomy-supportive. The third is symbolic: The output can express the consumer’s identity and become incorporated into the extended self. GenAI may enhance the instrumental layer while weakening the experiential or symbolic layers if the system performs the decisions that make the outcome personally meaningful. Therefore, participation frequency or interface activity alone is an inadequate measure of co-creation quality in GenAI settings.

Artificial intelligence can automate tasks, augment human judgment, or create hybrid configurations in which human and machine capabilities are combined [27], [28]. In NPD, GenAI can support problem framing, concept generation, visual prototyping, copy development, preference translation, and rapid iteration [29]. These capabilities reduce search costs and make sophisticated design activities available to consumers without professional training. They also allow firms to scale individualized co-creation beyond the resource limits of human design assistance.

The marketing implications depend on the system role. As an assistant, GenAI responds to consumer-defined goals and performs bounded transformations, such as refining language or visualizing an explicitly described concept. As a collaborator, it proposes alternatives while the consumer iteratively evaluates, combines, and redirects them. As an autonomous creator, it generates the core concept and asks the consumer mainly to approve, rank, or lightly edit outputs. These configurations may produce similar final artifacts but fundamentally different psychological experiences.

Creativity evidence reinforces the importance of role design. GenAI assistance can raise individual creative performance and benefit less-skilled users, but it can also anchor users on machine-generated ideas and reduce diversity across outputs [9], [10], [12]. Moreover, creative self-efficacy is higher when users are treated as co-creators rather than editors of machine-generated work [11]. This evidence suggests that the critical variable is not access to GenAI itself, but whether the interaction architecture preserves a meaningful human contribution.

Perceived control refers to the consumer’s experienced capacity to influence a process and its outcomes. In service and choice contexts, perceived control improves evaluations because it reduces helplessness and allows consumers to align the experience with their preferences [30]. Autonomy is likewise a basic psychological need: People are more intrinsically motivated when they experience their actions as self-endorsed rather than externally controlled [31]. Artificial intelligence can expand choice while also undermining autonomy if consumers believe that an algorithm invisibly determines the options, recommendations, or outcomes available to them [32].

Creative self-efficacy is the belief that one can produce creative outcomes [33]. In consumer co-creation, this belief should matter because GenAI can either function as capability scaffolding or as capability substitution. Scaffolding helps consumers articulate ideas they could not otherwise execute, preserving the perception that the output originates from their intent. Substitution provides a completed solution that may be impressive but offers little evidence of the consumer’s own competence. The distinction parallels the difference between being a co-creator and being an editor of machine output [11].

Perceived authorship is defined here as the extent to which consumers attribute the product’s defining ideas and choices to themselves. It is not equivalent to legal authorship, nor does it require exclusive creation. Rather, it reflects a subjective attribution: “this design expresses my intentions because my decisions shaped what it became.” Perceived authorship should arise when the consumer supplies identity-relevant inputs, makes consequential selections, and can trace the final output to those inputs. GenAI complicates this traceability because fluent outputs can obscure how much of the concept came from the consumer and how much came from the system.

Psychological ownership is the state in which individuals feel that a target is “mine,” even without formal ownership rights [34], [35]. It develops through control over the target, intimate knowledge, and investment of the self. Consumer research demonstrates that even temporary tactile interaction can increase perceived ownership and valuation [36]. Co-creation provides several ownership routes simultaneously: Consumers control attributes, learn the product in detail, invest effort, and project personal preferences into the outcome. The extended-self perspective explains why ownership is especially consequential for identity-relevant products [37]. A customized product can represent the consumer’s taste, competence, group membership, or aspirations. However, when a generative system supplies the identity-defining elements, consumers may question whether the product genuinely reflects them. This concern introduces authenticity as a second mechanism. Authenticity judgments concern whether an offering is experienced as sincere, credible, and true to an asserted identity or origin [38], [39].

Authenticity is likely to become particularly important when consumers know that GenAI was involved. Algorithmic judgments can be appreciated in objective tasks but resisted when the task is perceived as subjective, identity-sensitive, or dependent on uniquely human understanding [40–43]. Product design occupies an ambiguous position: functional optimization may benefit from algorithmic competence, whereas expressive design may be evaluated through human meaning and authorship. Accordingly, a system that produces technically attractive options can still reduce authenticity if consumers believe that the result no longer represents their own voice.

Table 1 synthesizes the central insights and unresolved questions across the five literature streams. The common gap is that existing research does not explain how the allocation of creative agency affects the transition from participation to psychological ownership and market behavior.

prompts, clicks, or iterations.

GenAI Co-Creation Framework Contribution Interactive activity may no Distinguishes participation longer indicate substantive from directive control and consumer contribution perceived authorship.

when a system generates the defining content.

Research emphasizes Specifies assistant, productivity and creativity collaborator, and more than ownership, autonomous-creator roles.

authenticity, and market response.

It is unclear when GenAI Models perceived control and scaffolds consumer creative self-efficacy as capability versus substitutes enabling mechanisms.

for it.

GenAI can reduce effort Links authorship attribution while increasing capability, to psychological ownership creating competing effects on and downstream behavior.

ownership.

The authenticity of a co- Introduces authenticity created product may depend erosion as a displacement on whether human mechanism and identity contribution remains visible relevance as a boundary and meaningful. condition.

The proposed framework treats GenAI-enabled co-creation as an allocation of creative agency between consumer and system. Creative agency is the capacity to make consequential decisions that determine what the product becomes. Two design dimensions define this allocation. The first is consumer directive control: the extent to which the consumer sets goals, establishes constraints, chooses among alternatives, initiates revisions, and controls final acceptance. The second is GenAI generative agency: The extent to which the system independently determines the concept, attributes, aesthetic form, or solution logic.

These dimensions are related but not perfect opposites. A well-designed collaborative interface can provide high system capability while retaining high consumer directive control. For example, a consumer may define the target user, functional requirements, visual identity, prohibited features, and evaluation criteria, while GenAI quickly produces and revises alternatives. Conversely, a platform can appear interactive while offering low consumer control if it asks for a short prompt, generates a nearly complete design, and limits the user to selection. The framework therefore focuses on experienced role structure rather than the mere presence of Fig. 1 presents the dual-pathway framework. The enabling pathway connects interface control to perceived control, creative self-efficacy, perceived authorship, and psychological ownership. The displacement pathway connects high GenAI generative agency to authorship dilution and authenticity concerns. The resulting levels of ownership and authenticity influence purchase intention, Willingness to Pay (WTP), advocacy, and continued participation. Product identity relevance, process transparency, consumer artificial intelligence literacy, and customization depth condition these effects.

Objective interface controls matter only to the extent that consumers experience them as consequential. A platform may offer numerous options but still feel controlling if the system repeatedly overrides preferences, narrows the solution space, or makes opaque changes. By contrast, explicit goal setting, reversible actions, selective acceptance, version comparison, and final-decision authority should communicate that the consumer directs the process. Perceived control is therefore the first psychological translation of interface design. Proposition 1 (greater consumer directive control in a GenAI-enabled co-creation interface increases perceived control over the design process and product outcome). Perceived control should strengthen creative self-efficacy because consumers can observe a reliable link between their decisions and changes in the design. GenAI can make previously inaccessible actions achievable, for example, turning verbal descriptions into visual forms, without necessarily displacing the consumer. When the system behaves as responsive scaffolding, consumers receive evidence that they can generate and improve creative outcomes. In contrast, when outputs appear independently of meaningful user direction, the system demonstrates its capability rather than the consumer’s capability.

Proposition 2 (perceived control positively influences creative self-efficacy during GenAI-enabled product co-creation). Creative self-efficacy should increase perceived authorship. Consumers who believe they were capable of shaping the design are more likely to attribute the final outcome to their own creative contribution. This mechanism is especially important because GenAI makes technical execution less diagnostic of authorship. The relevant question shifts from “did I physically create every element?” to “did my ideas and decisions cause the product to become what it is?” A co-creator role provides this causal experience, whereas an editor role can signal that the central creative act occurred before the consumer entered the process.

Proposition 3 (creative self-efficacy positively influences perceived authorship of a GenAI-assisted product). Perceived control and authorship provide complementary routes to psychological ownership. Control creates a sense of possession over decisions, whereas authorship embeds the self into the product’s form and meaning. A consumer can control a purchase without feeling that the product expresses the self; similarly, a consumer can inspire an idea without retaining control over its final execution. Psychological ownership should be strongest when both conditions are present.

Proposition 4 (perceived control and perceived authorship each positively influence psychological ownership, and their joint presence produces stronger psychological ownership than either mechanism alone).

GenAI generative agency becomes problematic when it substitutes for consumer decisions that would otherwise carry identity or meaning. The system may determine the central concept, aesthetic direction, feature combination, or narrative rationale while presenting the consumer with polished options. Such output can reduce effort and improve apparent quality, but it also weakens the causal connection between consumer intention and final design. The consumer becomes a selector of machine proposals rather than an originator or director.

Proposition 5 (as GenAI generative agency substitutes for consequential consumer decisions, perceived authorship decreases). Authorship dilution is not only a question of contribution volume. A small human contribution can remain highly meaningful if it establishes the design’s identity, whereas extensive low-level editing may not restore authorship after the system has determined the core concept. Accordingly, the location of consumer contribution in the decision hierarchy matters. Goal definition, identity cues, trade-off decisions, and final integration should contribute more to authorship than cosmetic adjustments.

Proposition 6 (the negative effect of GenAI generative agency on perceived authorship is stronger when the system controls identity-defining and concept-level decisions than when it performs execution-level tasks). High GenAI agency can also create authenticity concerns. A co-created product is implicitly presented as an expression of the consumer’s needs or identity. If the system supplies the defining ideas, that claim may feel less credible. This issue is analogous to task-dependent algorithm aversion: consumers are more receptive to algorithms in objective tasks than in contexts requiring subjective judgment or personal understanding [42], [43]. Functional optimization may therefore tolerate high system agency, whereas expressive design may require visible human authorship.

Proposition 7 (high GenAI generative agency reduces perceived authenticity when consumers expect the product to express personal identity, taste, or lived experience).

Psychological ownership should influence several market outcomes. First, ownership increases purchase intention because rejecting the product can feel like abandoning something already incorporated into the self. Second, consumers may accept a price premium because the offering contains personal meaning and is difficult to replace with a standard alternative. Third, ownership can encourage advocacy because displaying or recommending the product communicates both taste and creative competence. Fourth, it can increase continued participation by making future co-creation personally rewarding rather than merely efficient. Proposition 8 (psychological ownership positively influences purchase intention, willingness to pay, advocacy, and continued participation in GenAI-enabled co-creation).

Authenticity should exert both direct and interactive effects. Directly, consumers should respond more favorably to products they regard as genuine expressions of their preferences. Interactively, authenticity should validate the meaning of ownership. A consumer may feel possession toward a highly automated product but hesitate to pay a premium or display it if the product seems inauthentic. Conversely, visible continuity between consumer input and final output can make ownership commercially consequential. Proposition 9 (perceived authenticity positively influences market responses and strengthens the positive relationship between psychological ownership and those responses).

The combined framework predicts a non-linear relationship between GenAI generative agency and psychological ownership. At very low system agency, consumers retain authorship but may receive insufficient capability enhancement; the process can remain difficult, frustrating, or limited by technical skill. At moderate agency, the system expands the feasible design space while the consumer retains directive control, creating an ownership-preserving augmentation zone. At very high agency, the system generates most defining content, and the displacement pathway dominates. Fig. 2 illustrates this predicted inverted-U relationship.

Proposition 10 (the relationship between GenAI generative agency and psychological ownership is inverted-U shaped, with the highest ownership occurring when the system operates as a capable collaborator under meaningful consumer direction).

Process transparency is the first boundary condition. Transparency is often treated as universally beneficial, but its effect should depend on what it reveals. A contribution map that shows how consumer goals, constraints, and revisions shaped the final design can reinforce authorship. A disclosure that emphasizes autonomous machine generation may instead make displacement salient. Thus, transparency should be designed to make role allocation intelligible rather than merely expose technical complexity.

Proposition 11 (process transparency strengthens psychological ownership when it makes the causal influence of consumer inputs visible, but weakens ownership when it primarily highlights system dominance over defining decisions). Product identity relevance is the second boundary condition. Consumers diverge more strongly in domains used for identity signaling [44]. Clothing, home décor, gifts, personal accessories, and creative media therefore require greater protection of human authorship than utilitarian configurations such as storage capacity or delivery scheduling. Third, artificial intelligence literacy should reduce uncertainty about system behavior and help consumers use GenAI as a controllable instrument rather than treat it as an opaque agent. Fourth, customization depth should matter: superficial choices cannot compensate for system control of the core concept.

Proposition 12 (the negative effects of high GenAI generative agency on authorship and authenticity are stronger for identity-relevant products and superficial customization, but weaker among consumers with greater artificial intelligence literacy and interfaces that preserve deep, consequential customization). Table 2 summarizes the framework’s principal constructs and offers operational guidance for future studies. The measures should assess consumers’ experienced role allocation rather than infer it from system settings alone.

Many consumer-facing GenAI tools are optimized for speed: a short prompt produces a polished output with minimal effort. This design can improve conversion at the beginning of the journey but weaken attachment to the outcome. Firms should therefore optimize for ownership-preserving ease rather than friction elimination alone. The interface should reduce execution barriers while preserving consequential decisions. Useful features include explicit goal definition, editable constraints, selective acceptance of suggestions, side-by-side alternatives, reversible revisions, and final integration by the consumer.

A practical design principle is to keep the consumer “above” the model in the decision hierarchy. The consumer should determine why the product exists, whom it serves, which trade-offs matter, and what identity it expresses. GenAI can operate “below” that layer by translating intent into alternatives, checking consistency, or implementing revisions. This arrangement converts technical automation into experienced augmentation.

Firms should make the consumer’s causal contribution visible. A contribution history can show the initial concept, accepted and rejected suggestions, key preference decisions, and how the final version evolved. Such a contribution history is not simply an explainability feature; it is an ownership cue. Naming user-created versions, allowing consumers to annotate design rationales, and generating a concise “design story” based on verified interaction history can strengthen perceived authorship without falsely claiming that the system made no contribution.

Attribution language must remain accurate. Marketing claims such as “designed by you” are risky when the consumer only selects among autonomously generated options. More credible labels include “directed by you and developed with artificial intelligence” when the consumer truly controlled goals and revisions. The platform should avoid deceptive anthropomorphism and should disclose system involvement in a way that clarifies complementary roles.

A single co-creation architecture is unlikely to work across categories. High system agency may be acceptable for utilitarian optimization but harmful for identity-expressive products. Firms should therefore vary role allocation by product meaning. In functional categories, the system can recommend technically efficient configurations while preserving consumer veto and priority setting. In identity-relevant categories, the consumer should retain concept-level decisions and receive richer tools for expressing personal narratives, symbolic preferences, and aesthetic boundaries.

Consumer expertise also matters. Novices may benefit from stronger scaffolding, examples, and structured prompts, whereas experts may experience the same guidance as constraining. Adaptive systems should infer desired assistance from explicit preferences and interaction behavior, but they should not silently remove control. A visible “assistance level” setting, ranging from execution support to exploratory collaboration, can allow consumers to choose the relationship they want with the system.

Managerial dashboards should assess process and psychological outcomes in addition to design quality. Relevant indicators include perceived control, authorship, psychological ownership, authenticity, abandonment after generation, conversion to purchase, WTP, sharing, and return to the co-creation platform. A tool that raises aesthetic ratings but reduces ownership may appear successful in prototype tests while underperforming commercially. Firms should therefore test both creative performance and consumer attachment before scaling an autonomous design feature.

Study 1 should test whether the human role, independent creator, editor of machine output, or active human– GenAI co-creator, changes creative self-efficacy, authorship, psychological ownership, and purchase response. Participants could design a neutral but customizable product such as a desk lamp, travel bottle, or backpack. The independent-creator condition would use standard digital tools without GenAI. The editor condition would receive a completed system-generated concept and limited editing controls. The co-creator condition would iteratively set goals, accept or reject suggestions, and direct revisions.

A sample of approximately 450–600 participants would permit planned contrasts and mediation analysis while allowing exclusions for failed manipulation checks. The primary prediction is not that one condition universally maximizes objective creativity. Rather, the co-creator condition should outperform the editor condition on self-efficacy, authorship, and ownership, with these mechanisms explaining purchase intention and WTP. External raters should separately evaluate output novelty, usefulness, and aesthetic quality so that psychological effects are not confounded with artifact quality.

Study 2 should use a factorial experiment manipulating consumer directive control and GenAI generative agency. High-control participants would specify goals, constraints, and trade-offs and retain final integration authority; low-control participants would select among bounded options. High-agency systems would generate concept-level and identity-defining elements, whereas low-agency systems would execute consumer-defined ideas. Product identity relevance should be manipulated or compared across categories, for example, a personalized sneaker or gift versus a utilitarian storage configuration.

This design can test the displacement pathway and the predicted interaction between system agency and identity relevance. The key outcomes are perceived authorship, authenticity, psychological ownership, purchase choice, and WTP. A moderated serial-mediation analysis should assess whether high system agency reduces market response through authorship dilution and authenticity erosion, particularly in identity-relevant categories. The study should also record process data, prompt revisions, accepted suggestions, time spent, and decision reversals, to compare objective interaction patterns with subjective role perceptions.

Study 3 should test the ownership-preserving augmentation zone in a field or high-fidelity marketplace setting. Participants could receive a real budget to customize and purchase a low-cost product, or make incentive-compatible choices among products created under different assistance levels. The interface would offer a continuous assistance setting, enabling estimation of a non-linear relationship between system agency and ownership. Actual purchase, price premium, sharing behavior, and return visits would provide stronger evidence than intentions alone.

A field design should also test process transparency. One version of the interface could display a contribution map showing how user inputs shaped the final product; another could display only a generic disclosure that artificial intelligence was used. This comparison would distinguish role-clarifying transparency from mere disclosure. Longitudinal follow-up could assess whether ownership persists after delivery and use, rather than disappearing once the novelty of co-creation fades.

Measures should be adapted conservatively from established scales and validated for the GenAI co-creation context. Perceived control can draw on consumer-control research [30], creative self-efficacy on established efficacy measures [33], psychological ownership on the control and self-investment tradition [34], [35], and authenticity on consumer-based authenticity scales [38], [39]. Perceived authorship requires careful scale development because it is distinct from effort, control, and legal ownership. Items should capture causal contribution, self-expression, and continuity between the consumer’s intention and the final design.

The empirical program should separate manipulations from mediators, include pretests of role perception, and use independent evaluations of output quality. Behavioral WTP is preferable to a single self-reported item. Studies should be replicated across cultures and product categories because autonomy, authorship, and authenticity may not operate identically across contexts. Finally, researchers should archive prompts, system versions, and interface logic because GenAI behavior changes over time and can undermine reproducibility if the technological implementation is not documented.