The rapid integration of generative artificial intelligence into scientific visualization is precipitating a credibility crisis within the academic publishing sector and broader scientific communication landscape. As AI models enable the creation of highly realistic scientific imagery from simple text prompts, traditional markers of visual authenticity are losing their authority. This shift threatens public trust in research findings and complicates the dissemination of evidence-based knowledge. Recent incidents underscore the scale of the challenge. In 2024, multiple peer-reviewed papers faced retraction after publication of AI-generated figures containing biologically impossible structures. Most recently, in April 2026, the New England Journal of Medicine withdrew a clinical study following the discovery of AI-manipulated imagery. These cases reflect only a fraction of a growing trend, with experts warning that fields reliant on visual evidence, particularly materials science and medicine, face mounting risks. The proliferation of synthetic visuals is further complicated by detection systems that consistently trail behind generation capabilities. AI image detectors rely on known pattern recognition and require continuous retraining as new generative models emerge, leaving a persistent window where fabricated content passes initial editorial and peer review. The erosion of trust stems from the disruption of established cognitive heuristics. Historically, scientific imagery derived credibility from technical complexity, institutional affiliation, and direct alignment with pre-existing knowledge. Generative AI democratizes the production of polished, sophisticated visuals, decoupling them from verifiable instrumentation or observational contexts. When technical polish and institutional provenance become unreliable indicators, audiences default to motivated reasoning. Authentic research that challenges established views is more readily dismissed as synthetic, while fabricated images that confirm pre-existing beliefs gain unearned credibility. Academic publishers are exploring automated verification tools, yet experts emphasize that technological detection alone is insufficient. The prevailing recommendation centers on institutionalizing image provenance standards akin to existing data and methodology disclosures. Researchers and institutions are urged to explicitly document the origin, generation parameters, and modification history of all visual materials. Clear labeling of AI-assisted creation, simulation, or enhancement should become a baseline requirement. Studies indicate that audiences familiar with AI technology often view explicit disclosure as a marker of transparency, sometimes increasing the perceived credibility of labeled synthetic content over unattributed material. Preserving the authority of scientific visualization requires a structural shift toward documented authenticity. The enduring impact of landmark imagery, such as the 1968 Apollo 8 Earthrise photograph and the 2026 Artemis II mission visuals, relies not on aesthetic appeal but on traceable connections to physical instrumentation, documented missions, and verifiable observational protocols. In an era where synthetic and authentic imagery are visually indistinguishable, scientific institutions can no longer rely on implicit trust. Establishing rigorous, field-specific standards for image verification and transparent attribution will be essential to maintaining the integrity of scientific communication and preventing a broader crisis of empirical credibility.