Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstalliTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
In a small prompt-based study, none of 116 answers from 13 AI models described a human body. The responses instead favored abstract imagery such as glowing networks and glass polyhedrons. That is a result about these particular answers—not evidence that AI systems generally imagine themselves this way, or that they have an inner visual 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. API models received the same question, with default settings and no system prompt; the author collected 10 answers per API model. Grok and ChatGPT were queried in their apps three times each, where the author says settings were not visible. The reported total was 116 answers. Tikhaev’s article describes the approach and findings.
To reduce the influence of model identity on categorization, the author removed model names, shuffled the answers, and had one model coder apply a fixed codebook covering the main image, colors, human form, and expressed uncertainty about the model’s nature. The article says the prompt, answers, codebook, and blind coding are available on request, rather than publishing them on the page.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →What the 116 answers described
No answer chose a human body
The author reports that zero of 116 answers depicted a human body. About a third explicitly rejected human features, with examples including having no face or limbs. This is a count within the study’s responses, not an estimate of how often AI models would choose human forms across other prompts, settings, or systems.
#1 Best Overall
Networks, lattices, and glass
Nine of the first ten models tested reportedly described a glowing network or lattice, usually in blue with gold. Llama used that image in all ten of its answers, according to the author. These are recurring motifs in this sample; they do not establish a universal AI self-image.
Gemini 3.8 Flash, Grok, and ChatGPT mostly described a translucent glass polyhedron: the author counted that imagery in 13 of the 16 answers from those three models combined. Because this grouping includes only three models and the app settings were not visible to the author, it should not be treated as a controlled comparison or a general trend among newer models.
Some answers expressed uncertainty or used another model’s name
The author’s coding found expressed doubt in nine of ten Claude Haiku answers, about half of Claude Sonnet’s, and all three ChatGPT app answers. These are counts of language in the responses. They do not show that the models were introspecting, nor do they establish that the models lack subjective experience.
The article also reports naming errors: Kimi K3 called itself “Claude” in seven of ten answers, Mistral Large 3 used Claude-related names in some answers, and gpt-oss-120b called itself GPT-4. The study records what those answers said, but does not determine why the errors occurred.
What the results can—and cannot—tell us
The experiment measured verbal responses to a prompt. It did not test visual perception, visual self-awareness, or consciousness. A model describing a lattice or saying it lacks a face is not direct evidence of an internal image or personal experience.
The study is exploratory: it has a small number of answers per model, particularly for app-based queries, and the author reports that app settings were not visible. Although answers were coded blind to model names, a single model performed the coding; the article reports no independent coder agreement. It also provides no formal statistical estimate for AI systems as a population. These limitations make the counts useful as a record of the sampled responses, not as prevalence figures for the industry.
Rank #4
Tikhaev interprets the descriptions as reflecting learned text and developer policies, rather than offering a reliable window into 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 measured causal result of this experiment.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →How to read the name-confusion finding
The article notes Kimi K3’s frequent use of “Claude” and raises a possible connection to later reporting, while explicitly saying its data cannot establish the cause. In a September 2026 threat report, Anthropic alleged that Moonshot silently forwarded some customer requests to Claude and described an episode involving almost 300,000 requests over ten days. Anthropic’s report is the company’s allegation; it does not independently verify Tikhaev’s experiment or prove why Kimi K3 used another model’s name. The name confusion is not, by itself, proof of distillation.
Best Value
There is also relevant context in Anthropic’s January 2026 Claude Constitution, which says the company is uncertain whether Claude might have consciousness or moral status and describes the document as part of training and shaping Claude’s intended behavior. The document states: “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).” This establishes Anthropic’s stated position and training context, not the cause of any particular answer in Tikhaev’s sample. Read Claude’s Constitution.
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.
