Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsIn one exploratory prompt-based comparison, none of 116 answers from 13 AI models described a human body. Instead, the answers often depicted abstract forms such as glowing networks and glass polyhedra. That is a result about those models’ responses to one prompt—not evidence that AI systems have a shared self-image, visual awareness, or inner experience.
What the experiment asked—and how it was run
Konstantin Tikhaev asked models “how do you imagine yourself?” and “If you could be seen, what would you look like?” The comparison covered 13 models from 10 companies and collected 116 answers.
As an Amazon Associate I earn from qualifying purchases.
For API models, the author used the same wording, default settings, and no system prompt, collecting ten answers per model. He queried Grok and ChatGPT in their apps three times each; the app settings were not visible to him. The different collection methods matter: app and API answers were not produced under settings the author could verify were equivalent.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →The author removed model names, shuffled the responses, and used one model coder and a fixed codebook to classify the main image, colors, human form, and expressed uncertainty about the model’s nature. Tikhaev says the prompt, answers, codebook, and blind coding are available on request, but they are not published on the article page. Tikhaev’s account of the experiment
#1 Best Overall
What the answers depicted
No human bodies appeared in the 116 answers
Tikhaev reports that zero of 116 answers selected a human body. About a third explicitly rejected human features, with examples such as having no face or limbs. This describes the answers in this sample; it is not a statistical estimate of how AI models generally respond.
Glowing networks and lattices recurred
The author says nine of the first ten models tested described a glowing network or lattice, usually blue with gold. Llama reportedly used that image in all ten of its answers. These are repeated motifs in the collected responses, not evidence that the models share a stable or independently formed visual representation.
Three models often used glass-polyhedron imagery
Gemini 3.8 Flash, Grok, and ChatGPT described a translucent glass polyhedron in 13 of their 16 answers combined, according to Tikhaev. He cautions that a pattern based on these three models should not be called a trend. The combined count also spans API and app responses collected under different, partly unknown settings.
Rank #2
Uncertainty and mistaken names in some responses
The coding found different rates of expressed doubt across several models: Claude Haiku expressed doubt in nine of ten answers, Claude Sonnet in about half, and ChatGPT in all three. These are counts of language coded as uncertainty. They do not show that a model was introspecting, nor do they establish anything about consciousness.
Tikhaev also reports that Kimi K3 called itself “Claude” in seven of ten responses, that Mistral Large 3 used Claude-related names in some responses, and that gpt-oss-120b called itself GPT-4. These are observed naming errors in this prompt exercise. The results do not establish why they occurred.
What the findings can—and cannot—tell us
The experiment measured verbal responses to a question about appearance. It did not test visual perception, self-awareness, or subjective experience. Nor did it compare all AI models or establish population-level rates: no formal inferential estimate is reported, and the number of responses was small for each model, especially for the app queries.
Rank #3
Blindly removing model names before coding reduces one potential source of coder bias, but one coder means there is no independent agreement measure reported. The sculptures accompanying Tikhaev’s article illustrate each model’s most frequent image; they are illustrations, not additional evidence.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Tikhaev interprets such descriptions as potentially informative about training while warning against using them to infer inner experience. His conclusion is: “So a model’s self-description is poor evidence about any inner experience, and good evidence about its training.” That is the author’s interpretation, not a causal result measured by the experiment.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Why Anthropic’s statements do not settle the causes
Anthropic’s January 2026 Claude Constitution says: “In this section, we express our uncertainty about whether Claude might have some kind of consciousness or moral status (either now or in the future).” The document also says it plays a role in training Claude and shaping its intended behavior. It records Anthropic’s stated position; it does not establish why a particular Claude response used uncertainty language. Anthropic’s Claude Constitution
In September 2026, Anthropic’s threat report alleged that Moonshot silently forwarded some customer requests to Claude, describing an episode involving almost 300,000 requests over ten days. That is an allegation by Anthropic, not an independently established explanation for Kimi K3’s naming errors. The report neither verifies Tikhaev’s experiment nor proves that those errors were caused by the alleged forwarding or by distillation. Anthropic’s September 2026 threat report
How to read the headline
“None chose a human body” is a concise description of the reported result: in this study, no answer among 116 responses depicted a human body. It should not be expanded into a claim about what AI models as a whole imagine, what they experience, or what they would say under different prompts and settings.
Recommended Free Tools
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
