How AI is being used to evaluate the authenticity of paintings, amid conservators' concerns of whether the tech can account for wear, damage, and other factors
Machine learning can be the difference between a charming picture and a masterpiece worth millions
Context & Ripple Effects
This fits a broader expansion of AI into art practice: artists were already weighing creative tools such as DALL-E and Midjourney against risks to their livelihoods in the earlier debate over AI tools in artists’ work. Authentication extends that debate from making images to deciding the status and value of existing ones.
The stakes are unusually high because an authenticity assessment can separate an appealing work from one valued in the millions. Later work on AI-assisted painting restoration suggests that machine learning is moving into multiple parts of the art-care workflow, not just image generation.
First-order effects
- Authentication workflows gain a machine-learning input that can influence how paintings are classified and, consequently, how their market value is assessed.
- Conservators must test AI findings against physical evidence such as wear and damage, since those conditions may not be adequately represented in a model’s assessment.
Second-order effects
- Dealers, collectors, and other market participants may demand clearer expert corroboration when AI is used in a high-value attribution decision, rather than treating an output as conclusive.
- Providers of art-analysis tools are pressured to show that their methods handle condition-related variation, not merely visual similarity in cleaner reference images.
Third-order effects
- If adoption persists, art authentication is likely to become a hybrid process in which computational screening is paired with specialist judgment and condition assessment.
- The broader shift is toward formalizing how AI evidence is documented and weighed in cultural-asset decisions, especially where an erroneous result can materially alter value.
The trend: AI is moving from creative production into expert cultural workflows, where its usefulness depends on fitting alongside—not replacing—domain-specific human judgment.