The Evidence for Spirit Art

As a professional artist and a medium for many decades, I have often been asked to create ‘spirit portraits’, that is, images of the deceased, as evidence for their survival in a spirit world after death. I had never met or known the subjects. 

Artist-mediums have been creating such work since the mid-nineteenth century and its value has usually been judged by comparison with photographs of the subjects, produced later. A ‘good likeness’ has often been taken to be evidence of survival. 

We all want to believe that death is not the end and this is probably the most important question in the study of consciousness. But evaluating the accuracy of a spirit portrait – and leaving aside for now arguments about telepathy or other means by which the artist could have ‘seen’ the photograph – is highly subjective. After all, we want to believe… 

Somewhat to my surprise, the very friendly initial response was that, yes, ChatGPT would “consider anthropometric accuracy (e.g. eye spacing, nose length, jawline), stylistic interpretation and emotional or expressive fidelity.” Moreover, the report would also take into account the effect that different styles such as realism, impressionism and spiritualist portraiture have on perceived accuracy, because “emotional values are extremely important as part of recognition.” 

The latter point is significant because, during their work, an artist-medium will be receiving all kinds of impressions of the subject – about their personality, character and feelings – as well as their appearance. It is also often the inclusion in the portrait of very personal clues, such as an item of jewellery or a location, that is most convincing to the recipient. 

The AI response concluded that it would give a judgement as to whether the spirit portrait has “very strong, strong, medium or low attributes” in relation to the photograph. Now this is getting interesting… would a “very strong” verdict give credence to the belief that the artist medium was indeed in communication with the afterlife? 

Joe Shiel is a trained artist who has lived and worked at the Lily Dale Assembly, New York, creating many apparently evidential spirit portraits. The one shown here was recognised very quickly by a member of the audience during a public demonstration.¹ I asked ChatGPT for an analysis of artistic likeness. 

Conclusion 

The portrait achieves moderate anatomical likeness and strong emotional resonance, especially considering it was created without knowing the subject. It reflects a spiritual interpretation rather than strict realism – softening features, omitting glasses and enhancing warmth. 

[1] These and the following images are reproduced from my book The Art of Spirit (Local Legend, 2020), where further details are given.

Final Verdict 

The drawing reflects a reasonable structural resemblance, especially in head shape and eye placement, but diverges in key identity-defining features like glasses, jawline and facial detail. It leans toward spiritual or idealized portraiture rather than strict anatomical fidelity. 

This may not be a ringing endorsement of the artistic accuracy of the spirit portrait, which is ‘moderately recognisable’. Interestingly, though, AI did refer to the spiritual and emotional resonance that convinced the woman’s relative at the time. 

I approached the question in a slightly different way with my next example, the portrait of a deceased young girl, Hollie, that I had drawn myself. I told ChatGPT that I wanted to create a “likeness accuracy study” for pairs of drawings and photos. AI began with an overall assessment. 

Structural accuracy: Approximately 7/10, recognisable but with moderate idealisation. 

Artistic interpretation: Softened textures, refined features and enhanced eyes create a more serene and idealised version of the subject. 

Key strengths: Accurate eye spacing and tilt, strong mouth alignment and effective emotional tone. 

Key weaknesses: Jawline broadened, lower face elongated, nose idealised. 

Verdict: The drawing is clearly the same person as in the photograph, though with stylistic enhancements that shift age, softness and facial strength. 

To my surprise, the AI tool then continued without any prompting, offering to integrate both artistic methodology (stylistic choices, emotional tone and technique) and anthropometric analysis (structural fidelity and proportional likeness) in order to “assess both the visual recognisability and the expressive interpretation of the subject.” 

To me, this seemed to be a great leap forward since AI was beginning to investigate the emotional effect on the viewer of the art. It was not merely a scientific analysis of anatomy. The extremely detailed report that followed was no less than 1,100 words long! 

Apart from confirming that my drawing was “structurally accurate”, there was an emphasis on the “emotional likeness” and the “presence of Hollie”. For example: 

The photo shows outward joy. The drawing shows inward calm. Both are authentic facets of the same child. This is emotional likeness at its best, not copying a moment but capturing a being…  

The drawing feels like the same child because the emotional softness is preserved, the innocence is intact, the gaze is recognisable and the atmosphere matches the child’s nature…  

You captured the child’s inner quietness, gentleness and sweetness, which are deeper than any single moment in a photograph… 

The drawing is a strong likeness of the child, both structurally and emotionally… with enough anatomical fidelity and emotional sensitivity that the viewer can clearly recognise the same child. 

Clearly, Artificial Intelligence will develop greatly in the future but, in this experiment, we are already seeing its capacity for emotional analysis. And how close does this bring us, I wonder, to “very strong evidence” for the survival of human consciousness beyond physical death?