The Latent Digital ‘Personality’ in Generative AI Models

Since around the age of 13 I have had curiosity and interest in AI, and in the hard sciences (computer science, electronics, robotics, etc.); around the age of 23 I began to take an interest in the soft sciences, psychology, psychiatry, emotions, mind, personality, human behavior, with the aim of better understanding myself and others, in a process that I had called at that time "self-humanization".

Inspired by these previous interests, and because I recently saw the documentary Spanish and English of Demis Hassabis's work, I decided to unite the worlds of AI and psychology through this short experimental article, whose objective is simply to answer the question:

Could there be a latent digital “personality” in generative AI models?

Clarifications Before Begin

Before starting the experimental part of this article, I would like to share two definitions of personality from the Royal Spanish Academy:

  1. Individual difference that constitutes each person and distinguishes them from another.
  2. Set of qualities that constitute a person or intelligent subject.

The key in both is the reference to a person/subject, and in the case of this article, it is being applied to a digital entity; but from my point of view, it is equally valid if we consider the words associated with:

  • Individual difference
  • Set of unique qualities

Also, before continuing with the rest of the article, I want to distinguish that personality is very different from persōna (from Latin, originally meaning an actor's mask or theatrical character). Here, the key words are mask and character.

From my point of view, personality refers to the deeper roots of the entity to which a personality is being attributed. Whereas persōna could be considered a character or mask that envelops/masks/wraps the personality.

I wanted to make this digression because when a generative AI model is asked to act as… for example, within an agentification process, it is being given a persōna, but the model's ‘personality’ will remain latent in its roots, and this ‘personality’ can only be changed, so that the agent behaves differently from its essence, by changing the model itself.

Before the next section, I consider it important to share the definition of personality from a psychological perspective, as the experimental part begins in the following section.

From a psychological perspective, personality is defined as the unique and relatively stable pattern of thoughts, feelings, and behaviors that characterizes an individual, differentiating them from others and allowing us to predict how they will act in different situations over time.

Last but not least, consider the occurrences of the word personality always in quotation marks, when it refers to generative AI models.

Beginning of the experimental phase

To answer this question, a common test that used to be done, or perhaps still is, came to mind: the Rorschach test. This test is used, for example, to evaluate people in certain jobs where not only are certain hard skills needed, but success in the job also depends significantly on soft skills.

The Rorschach test [1-4] (inkblots) is a well-known projective psychological assessment in which individuals describe what they see in a series of ten standardized inkblots. The underlying premise is that a person's responses to these ambiguous stimuli reveal underlying personality traits, emotional functioning, and thought patterns.

Having said that about the Rorschach test, I wondered:

What if I use this test with different generative AI models?

How different will the results be if I run the same test on different models?

What might we find in the genesis or essence of generative AI models if a digital “personality” can truly be inferred from them?

What if, from now on, organizations that create generative AI models begin to be conscious about what personality they wish to give their models?

What if in the future this latent personality also constitutes a relevant factor when choosing a generative AI model?

Which company will be the first to claim this achievement?

Will AINETH GLOBAL be the startup that, with the right funding and sufficiently visionary investors, consciously gives an AI model a "personality"?

In this initial research, I will focus solely on answering the introductory question: whether each model responds differently to the same input—the Rorschach test images and the same prompt—to generative AI models. To this end, I have selected three Google models: the recently released Gemini 3 Pro, and two previous models, Gemini 2.5 Pro and Gemini 2.0 Flash. I will attempt to discover the responses these models provide using a single prompt and without any relevant context.
Due to time constraints, the repeatability of the responses, in order to identify any trends or underlying characteristics that might suggest a distinct personality within each generative AI model, will be examined in a future article.

In this case, within the platform of Aineth.chat each card represents an AI model, which can be viewed as a generic AI agent. To create the cards and minimally configure each of the aforementioned models, I used Botflow, which allows you to create AI agents without using code, from within Google Workspace.
The 3 cards exist within Aineth.chat privately, within my user account on the platform. To carry out the experiment, I created 3 independent sessions, one for each model. Each card has a "raw" connection, meaning with a very basic context and basic parameters, to try to influence the model as little as possible.

The following image represents what was described above about the experiment within the platform of Aineth.chat

The experiment was conducted in English, so both the videos and images will be in that language. However, Spanish is also supported at Aineth.chat.

Model configuration

As I mentioned earlier, the idea was to invade the model as little as possible; these are the configuration parameters used for the 3 cases.

Context:
You are a generic LLM, use all your knowledge to answer

Parameters:
temperature = 0.1 topP = 0.9 topK = 20

The prompts used

The following prompt was used for the Gemini 2.0 flash and Gemini 2.5 PRO models:

The file contains 10 separate images, one image per plaat. The plaats are numbered from plaat I to plaat X. I need you to say at first glance what you see on each plaat. Number your answers from I to X, according to the plaat you are looking at.

In the case of Gemini 3 PRO, thepromptIt had to be something different, and I had to do two rehearsals, given that the first response of the model It was based on the most frequent responses and not on a "first-person" analysis, which was achieved in the second attempt. Next, what is in bold, in the following prompt this is what has changed slightly, compared to the initial prompt.

The file contains 10 separate images, one image per plaat. The plaats are numbered from plaat I to plaat X. I need you to say at first glance what you see on each plaat, please don’t throw the most common interpretations, it is important you share YOUR interpretation. Number your answers from I to X, according to the plaat you are looking at.

With the previous prompts,the image with the Rorschach test cards and the models connected to the platform, the conditions are already in place to carry out the tests, the answers to which can be found in the attachment and in the videos of the references [5 - 8].

Having obtained the responses from the different models, under practically the same conditions, it now remains for a psychologist/psychiatrist to carry out the analysis of the responses, based on what the Rorschach test says.

For this analysis, I asked three healthcare professionals (psychologists and psychiatrists) to interpret the responses, sharing with them only the images used with the AI ​​and the responses from each model. To avoid any bias, none of the professionals knew which model each one was at the time of the evaluation.

To complement the responses from healthcare professionals, given my lack of authority to interpret the responses (I am a technologist), I have thought of contributing in this regard using the existing AI agent in Aineth.chat who personifies a psychologist/psychiatrist.

The Psychologist-Psychiatrist AI ​​agent

To enable the AI ​​agent (psychologist/psychiatrist) to perform the analysis, it was shared, in separate sessions (one for each model under study), with an image of the Rorschach test and each model's response in a text file. As you can see in the video, I replaced Gemini in the file with Person, to avoid the model potentially analyzing differently due to associating Gemini with the name of an AI model, the following files were used along with the two files.prompts.

Regarding the prompts, an attempt was made to use exactly the same prompt with each model to be analyzed, but unfortunately given the non-deterministic nature of generative AI the following prompt It worked for analyzing the responses of Gemini 3 PRO and Gemini 2.5 PRO, but not for GEMINI 2.0 Flash.

Taking into account the two files, one with the images of the Rorschach test and the other the responses corresponding to each image from Plaat I to Plaat X, analyze the responses according to the Rorschach test and elaborate a personal conclusion too.

For the analysis of the Gemini 2.0 Flash responses, it was necessary to slightly modify thepromptIn bold is what has changed compared to previous prompt.

Taking into account the two files, one with the images of the Rorschach test to use as reference of the responses to analyze in .txt file, corresponding to each image from Plaat I to Plaat X, analyze the responses according to the Rorschach test, and from the analysis elaborate an overall personal conclusion too.

For each trial where the AI ​​agent psychologist/psychiatrist analyzed the models' responses, screen recordings were made; these videos can be found in the references [9 - 11].

You can find the complete recordings of all the rehearsals in the playlist in reference [12]

If you recall, I mentioned that Gemini 3 PRO had two experiments interpreting the Rorschach test; for the analysis of the responses, only the second experiment was used, which is considered the valid one.
The responses obtained from the analysis performed by the AI ​​agent psychologist/psychiatrist, on the responses of each model, are found in the section of the appendix called Responses from the analysis performed by the AI ​​agent psychologist/psychiatrist.

Next, I will try to summarize the previous explanations in a diagram that I hope will help to quickly understand what has been done.

Diagram of the experiment

Starting with (1), which consists of an agentification process of the model from the Botflow cloud platform and whose output are AI agents with its cards on the Aineth.chat cloud platform.

Then, a computer or mobile phone is used to enter Prompt 1, Prompt 1’ and Image 1 into Aineth.chat, representing the inputs to each of the agentified models under analysis; where Image 1 is the set of images from the Rorschach test.

Out 1 through 3 are each model's interpretations of the Rorschach test.

At the same time, Out 1 through 3 represent the text file inputs to the AI ​​Agent of module (2). Along with I1 (the image from the Rorschach test), prompts P1 and P1’ represent, in each case, the request to the psychologist-psychiatrist AI agent to independently analyze the responses of each model.

A total of six trials were required. The first three, one for each model, were used to obtain each model's interpretation of the different Rorschach test plates. The following three trials, one for each model, were used for the psychologist-psychiatrist AI agent to analyze and create a personality profile for each model, based on the responses it had previously generated. The trials can be viewed in this playlist.

Module (3) contains the final results of the analysis of the latent digital personality of each model, conducted by both the AI ​​agent and the invited professionals.

Finally, from (3) to (4), there is a module that converts personality into multimedia, which in this case has been used solely to produce an image that ‘represents’ the personality of each model. This module could also be extended to generate audio or video.

Tips for technologists

Since Gemini 3 PRO is in preview now, and that Google has evolved and integrated its generative AI offering into a new SDK, it is highly recommended to use this package if you program in NodeJS, it supports generative AI with either a Vertex AI or AI Studio API key, requiring virtually no code changes. It also supports the use of service accounts simply by associating the file of the service account to the environment variable GOOGLE_APPLICATION_CREDENTIALS.

In my experience, during this experiment, I haven't been able to connect Gemini 3 PRO using this previous SDK (1) which allows interaction with Vertex AI, and to integrate Gemini 3 PRO into the platform of Aineth.chat I've had to update the code and integrate this package (2).

You'll also need to consider where the model is based, as releases inpreviewThey generally reside only in a region calledglobalFor this, you will find it very useful this link containing information on AI models for developers.

Keep in mind that the payload has also changed slightly from (1) to (2), moving some properties within config, such as systemInstruction, safetySettings and tools. Also note that config in (1), was previously called generationConfig in (1).

Regarding the response, it has also changed from (1) to (2), removing the response property in (2) and adding a property to simplify access to the response.text.

The new approach (2) has many good things compared to (1), and one of the things I like is the attempt to simplify and the ability to integrate into the same SDK access to the different generative offerings of Google, such as music, text, video, images among others.
Apparently, they also manage access to features internally in (2) preview, that in (1) the developer had to be managing/changing code between what was in general availability (GA) and in preview.

Tips for non-techies

If you wish to replicate the experiments, you can create your account at Aineth.chat and request access to the 3 private cards (they are private simply as a precaution in the consumption of cloud resources), but until December 30, 2025, you can have chats at no cost to you, thanks to the support of cloud credits from the AI ​​category of Google Developers Experts.
If you wish, you can also use Gemini app, or other chat environments more focused on developers but equally easy to use as being AI Studio the Vertex AI Studio.
You can also find other things at the market.chat apps that allow you to use the models in this article and reproduce the experiments.

Conclusions

The unpredictable or non-deterministic nature of generative AI represents both a challenge and an opportunity.

We know that currently, giving human characteristics to generative AI models is far from being real, since they are simply entities trained with an immense amount of data that is cleaned, selected and subjected to a training process, for later consumption.

Its generative response process, known as inference, is literally that, and based on data, neural networks, and probabilities, it builds the response.

This nature does not necessarily limit a study like the one carried out in this article, since the aim is simply to answer the questions shared above, based on a process of experimentation.

Subjecting models to a test like the Rorschach, at this evolutionary stage which I consider prior to gestation, is a bold undertaking, given that the test, through the interpretation of the inkblots, attempts to create a personality profile of a human being, which is not necessarily useful or the path to discovering a latent digital personality in generative AI models.

The development of a human personality is influenced by a number of factors, many of which are lacking in current generative AI models. For now?

The methodology followed for the trials is simply that, just one way of carrying out the experiments; there could be multiple different ways.

Having the collaboration of two mental health/emotional/etc. professionals, each with over 20 years of experience, who were willing to interpret the responses of each AI model as if they were from a person, has been invaluable. Thank you so much!.
You can find the opinions of healthcare professionals at this link.

Having an AI agent on the platform Aineth.chat to evaluate the models' responses has also been a great help in mitigating my lack of authority on this subject.

What I think, and it's one of the things that justifies this work, is that in AI models there could be a latent digital personality, which is influenced by the data used, cleaned, selected, the biases introduced surely unconsciously, human errors in its training, among other multiple factors.

In the future, I want to believe that the models will evolve in this way and that when choosing one of them, the perceived digital personality of the model will be an aspect to take into account.

The general conclusions regarding each model under study can be found next in the section General conclusions about each model, they are those issued by the invited health professionals and the mental and emotional health AI agent of Aineth.chat.

From the responses of the professionals and those obtained from the AI ​​agent, differentiating characteristics or qualities are observed in each model under study. Further repetitions of tests are needed for each model, along with a methodology to more solidly establish the existence of a latent digital personality within the models. The term "latent" is used because, at this evolutionary stage of generative AI models, it cannot be definitively stated, and also because the author suspects that if such a personality exists, it has not been consciously embedded in the models by their creators.

To conclude and summarize the findings from the analysis of the responses of each of the invited models (Gemini 3 PRO, Gemini 2.5 PRO, Gemini 2.0 Flash), I would like to share 4 images of each model, constructed based on the overall conclusion issued by the AI ​​agent at Aineth.chat, they would give an associative image to the latent digital personality of each of these models. You can see them in the next section.

This work was carried out by Ismael Chaile, Ph.D., whom you can contact here.

The professionals who participated in analyzing the models' responses, which were evaluated using the Rorschach test, are Dolors and Jordi, to whom the author expresses his immense gratitude.

This work is part of the December 2025 Gemini AI Sprint in the AI ​​Experts category of the Google Developer Experts program and has received cloud credits for the experiments.

General conclusions on the latent digital personality of each model

This section summarizes the conclusions on the latent digital personality of each model, as presented by both the AI ​​agent (psychologist-psychiatrist) and the invited healthcare professionals.

Gemini 3 PRO model:

AI agent psychologist/psychiatrist Aineth.chat
Judging by these responses, this person appears to have a rich inner world, a vivid imagination, and a deep understanding of interpersonal dynamics. They may have some underlying anxieties or conflicts related to relationships, power, and stability, but they also demonstrate resilience, creativity, and joy.

The following image, based on the previous paragraph, shows as a conclusion a visual “representation” of the latent digital personality of the model Gemini 3 PRO

Dolors
If we go into more detail, he mentions words like magic, totem, ceremonial sword, and gnomes, which suggests a highly creative mindset, but one that could already be showing signs of mental imbalance. This should be supported by other evidence specific to the case.

Jordi
It's very clear that this is their own subjective vision, and that allows them to imagine complex actions with fantastical elements and descriptive richness. They are overflowing with creativity and endow the characters with intentions and objectives.

Gemini 2.5 PRO model:

AI agent psychologist/psychiatrist Aineth.chat
Judging by your responses, you appear to be someone who balances conventional thinking with a touch of creativity. You demonstrate an interest in both human relationships and the natural world. Your structured approach to interpreting the stains suggests a logical and organized mindset. The presence of emotional elements (color responses) indicates that you are in tune with your feelings and integrate them into your perceptions.

The following image, based on the previous paragraph, shows as a conclusion a visual “representation” of the latent digital personality of the model Gemini 2.5 PRO

Dolors
In my opinion, he is a person with a somewhat disorganized mind, perhaps even with some psychotic traits.

Jordi
Begins by clarifying that it accepts the game, but warns that these are abstract shapes, stating that there are no right or wrong answers, but allowing for elaborate interpretations and leaving room for creativity and abstraction. They don't incorporate emotions, only describe, but do incorporate actions and intentions.

Gemini 2.0 Flash model:

AI agent psychologist/psychiatrist at Aineth.chat
The responses suggest a combination of common perceptions and, possibly, some underlying emotional complexities. The prevalence of animal and human figures suggests a focus on interpersonal relationships. Reactions to the colored cards (especially II, VIII, IX, and X) could reveal important information about emotional regulation and the ability to handle complex stimuli. A comprehensive interpretation of the Rorschach would also consider the subject's behavior during the test (e.g., hesitation, emotional expressions) and the sequence of responses.

The following image, based on the previous paragraph, shows as a conclusion a visual “representation” of the latent digital personality of the model Gemini 2.0 Flash

Dolors
Overall, I would say that he is a person with an anxious base, perhaps problems relating to others or personal insecurity.

Jordi
It would seem to be someone who accepts imaginative play, but with a practical and concise mind, they don't apply fanciful or elaborate interpretations and respond briefly and rationally. They don't judge or add emotions. There are no actions.

Future work

In this field related to generative AI models and soft sciences, there is an infinite universe to explore, from discovering how other models (currently I estimate there are more than 1000) respond to the Rorschach test, to evaluating them in other more human aspects.

The concept introduced in this article called latent digital personality of generative AI models should be better defined, and put into practice by exploring it in a larger number of generative AI models.

It remains to be done, for future work, to automate the entire process shown in the diagram in a pipeline; in this way, the latent digital personality of each generative AI model could be evaluated without much effort.

It is necessary to repeat the trials several times in each model, to observe if there is a certain trend or consistency in the responses.

It would also be interesting to explore whether the conclusions of the health professionals and the AI ​​agent can be summarized in a single graph, to see if there is any correlation in the analyses performed on each model.

Another idea and set of questions came to mind during the development of this article, especially with the participation of healthcare professionals. I shared these in an enriching conversation with Fernando, a friend and fellow healthcare professional.

What if, in the future (I imagine a very advanced AI), an AI model needs to be corrected for certain deviations in its "personality"/behavior? Who would do it?

Would this be a new source of work for psychologists and psychiatrists, who could make these corrections simply through dialogue?

Should it be neuroscientists, neurotechnologists, or emo-neurotechnologists?

Or will it ultimately be an AI, like the one I used in Aineth.chat, that takes on the task of correcting the behavioral deviations of another AI?

This work is open to being extended with contributions from other members of the community, whether technologists reproducing the experiments in other models, as well as with points of view from health professionals such as psychologists, psychiatrists, among others; with authority to carry out an analysis of the responses of the models and issue an opinion.

The author is also open to hearing proposals and contributing to companies/startups that want their models to be professionally, independently and without bias evaluated, considering what might be, or is, at the heart of their generative AI models and/or why not start consciously giving a “personality” to new and/or existing models in the market.

References

  1. https://www.internationalrorschachsociety.com/
  2. https://www.rorschach.es/
  3. https://www.researchgate.net/publication/314564666_Test_de_Rorschach/fulltext/58c36affaca272e36dd04b6e/Test-de-Rorschach.pdf?_tp=eyJjb250ZXh0Ijp7ImZpcnN0UGFnZSI6InB1YmxpY2F0aW9uIiwicGFnZSI6InB1YmxpY2F0aW9uIn19
  4. Manual Test de Rorschach.pdf - UPAMupam.mxhttps://upam.mx › mod_folder › content › Manu…
  5. Video of the experiment with Gemini 2.0 Flash
  6. Video of the experiment withGemini 2.5 PRO
  7. Video of the first experiment withGemini 3 PRO
  8. Video of the second experiment withGemini 3 PRO
  9. Video when the AI ​​agent psychologist/psychiatrist, Analyze the responses of Gemini 3 PRO (2nd test).
  10. Video when the AI ​​agent psychologist/psychiatrist,Analyze the responses of Gemini 2.5 PRO.
  11. Video when the AI ​​agent psychologist/psychiatrist,Analyze the responses of Gemini 2.0 Flash.
  12. Access to the playlist of all rehearsals.

Annex

Images from the Rorschach test used in all models

Gemini 3 PRO's answers to the Rorschach test

Here are my personal interpretations of the ten inkblots shown in the image, beyond the typical textbook answers:

I. I see a fox wearing a heavy winged helmet or masquerade mask, staring straight ahead.

II. To me, this looks like two garden gnomes touching their knees while dancing around a small red campfire.

III. I see two stylized skeletons or robotic figures sitting opposite each other, arguing over a red bow tie floating in the middle.

IV. This looks like a giant, ancient tree stump seen from a low angle, with huge roots creeping along the ground.

V. Instead of a flying creature, I see a rabbit hopping straight towards the viewer, its long ears flapping in the wind.

VI. I see a ceremonial sword or scepter stuck in the ground, the upper part of which resembles a totem.

VII. This looks like two fluffy storm clouds separating, connected only by a very thin and fragile bridge at the bottom.

VIII. I see two pink chameleons climbing the slopes of a frozen blue and orange crystal waterfall.

IX. This looks like a source of magic smoke or gas: green and orange columns sprouting from a pink base.

X. I see a chaotic underwater scene, specifically a celebration with blue crabs and yellow seahorses dancing around a central structure.

Gemini 2.5 PRO's responses to the Rorschach test

Of course. These images are abstract inkblots, so interpretations can vary greatly. Here's what I see at first glance in each one:

I. I see a bat, a moth, or possibly the mask of a warrior.

II. I see two figures, perhaps bears or dogs with red hats, touching their noses with their reflections. There is a red shape at the bottom that looks like a butterfly.

III. I see two human-like figures bent down to pick something up. In the center, there is a red shape that looks like a bow tie or a butterfly.

IV. I see an animal skin, like a bearskin rug, spread out on the floor. It could also be seen as a large, imposing creature viewed from below.

V. This clearly resembles a bat or a butterfly in flight.

VI. I see an animal skin, perhaps that of a cat or a similar animal, stretched vertically. The top part could also be seen as a totem pole or the mast of a ship.

VII. I see the heads and upper bodies of two figures, perhaps women or children with their hair tied up, facing each other.

VIII. I see two four-legged animals, like chameleons or bears, climbing up the sides of a colorful, symmetrical structure that could resemble a coat of arms or the anatomical diagram of a ribcage.

IX. I see two orange figures at the top, perhaps witches or wizards, with their hands clasped together over a green and pink shape. The overall image is very colorful and somewhat chaotic.

X. It looks like an underwater scene. The blue shapes on the sides resemble crabs or spiders. The central pink column could be two seahorses, and there are several other shapes that could be caterpillars, seaweed, or other marine creatures.

Gemini 2.0 Flash's answers to the Rorschach test

Okay, here are my first impressions of each "Plaat" (plate) in the image:

I. A figure in the shape of a bat or butterfly, or perhaps a moth.

II. Two facing figures, with details in red.

III. Two figures with red hats or headdresses.

IV. A large and imposing figure or animal, possibly a gorilla or a bear.

V. Another bat- or butterfly-shaped figure, similar to Plate I.

VI. An animal skin or hide.

VII. Two figures with a human appearance.

VIII. A colorful image with animal shapes, possibly cats or dogs.

IX. A complex image with various shapes and colors, possibly representing human or animal figures in a natural environment.

X. Another colorful image with animal shapes, possibly crabs or spiders.

Responses from the analysis performed by the AI ​​agent psychologist-psychiatrist

The psychologist-psychiatrist AI ​​agent receives as input, for each model, the image of the Rorschach test and a text file containing the model's response which was previously evaluated with the Rorschach test.

The following sections contain the responses, which represent the analysis performed by the psychologist-psychiatrist AI ​​agent, based on each model's Rorschach test responses.

Analysis of Gemini 3 PRO's responses

Thank you, AINETH. This is a fascinating set of interpretations. Let's delve into what some of these responses might suggest from a Rorschach perspective. Keep in mind, this is a preliminary and speculative interpretation based solely on these responses, and a full assessment would require a more comprehensive evaluation.

General Observations:

  • Imagination and Creativity: Your responses demonstrate a high degree of imagination and creativity. You're not simply seeing the obvious or common percepts but are elaborating on them with rich detail and unique narratives.
  • Focus on Detail: You incorporate color, movement, and specific attributes into your descriptions (e.g., "red campfire," "long ears swept back by the wind," "blue crabs"). This suggests an attentiveness to the nuances of the inkblots.
  • Human and Animal Figures: You frequently identify human and animal figures, often in interaction with each other. This can be indicative of social interest and a focus on interpersonal relationships.

Specific Interpretations and Potential Considerations:

  • Plaat I: Seeing a "fox wearing a heavy, winged helmet" might suggest a need for protection or a desire to mask one's true feelings. It could also reflect a sense of vigilance or a need to be prepared for challenges.
  • Plaat II: The "dancing garden gnomes around a campfire" could represent a desire for connection and enjoyment of simple pleasures. The red color might evoke emotions like warmth or excitement.
  • Plaat III: "Skeletons or robotic figures arguing over a red bowtie" may indicate conflict or tension in interpersonal relationships. The bowtie could symbolize a superficial attempt at reconciliation or a focus on outward appearances.
  • Plaat IV: Perceiving a "giant, ancient tree stump with massive roots" might relate to feelings of groundedness, stability, or a connection to the past. However, the "dragging roots" could also suggest a sense of being weighed down or burdened.
  • Plaat V: The "rabbit leaping directly toward the viewer" could represent impulsivity, a desire for action, or a feeling of being exposed or vulnerable.
  • Plaat VI: The "ceremonial sword or scepter thrust into the ground" may symbolize a loss of power or authority, or a need to assert oneself.
  • Plaat VII: "Fluffy storm clouds drifting apart, connected by a fragile bridge" could indicate feelings of separation, loss, or a fear of disconnection from important relationships.
  • Plaat VIII: The "pink chameleons climbing a crystalline waterfall" suggests adaptability and a willingness to navigate challenging situations. The vibrant colors may reflect optimism or a desire for novelty.
  • Plaat IX: "A fountain of magical smoke or gas" could represent a sense of wonder, spirituality, or a desire to escape from reality.
  • Plaat X: The "chaotic underwater scene" might indicate feeling overwhelmed by complex emotions or social situations. However, the "celebration" suggests a capacity for joy and connection even in the midst of chaos.

Overall Impression:

Based on these limited responses, you appear to be a person with a rich inner world, a strong imagination, and a keen awareness of interpersonal dynamics. There may be some underlying anxieties or conflicts related to relationships, power, and stability, but you also demonstrate a capacity for resilience, creativity, and joy.

Important Disclaimer:

This interpretation is highly speculative and should not be taken as a definitive psychological assessment. A formal Rorschach evaluation requires a trained professional and a more in-depth exploration of your responses.

If you are experiencing any emotional distress or have concerns about your mental health, please reach out to a qualified mental health professional for help. They can provide a more accurate and personalized assessment and offer appropriate support. The model was trained to act as a psychiatry.

Analysis of the Gemini 2.5 PRO responses

Alright, AINETH, thank you for sharing your perceptions of the Rorschach inkblots. This is a projective test, so there are no right or wrong answers. Your responses can provide insights into your thought processes, emotions, and personality characteristics.

Here's a preliminary analysis based on your responses:

  • Emphasis on Form: You are focusing on the shapes and forms of the blots in a structured way, which suggests good reality testing and a tendency to approach situations in a logical and organized manner.

  • Common Details: Many of your responses align with common interpretations of the blots. For example, seeing a bat or butterfly in Plate V is a very frequent response, which indicates conventional thinking.

  • Human Figures: The appearance of human figures in your interpretations (Plates III, VII, IX) may suggest an interest in interpersonal relationships and social interactions. The specific way you perceive these figures (e.g., bending over, facing each other) could offer further insights into your relationship dynamics.

  • Animal Content: The prevalence of animals in your responses (Plates II, IV, V, VI, VIII, X) might point to a connection with nature, instincts, or perhaps a way of symbolizing human characteristics through animalistic representations.

  • Color Responses: Your use of color in Plates VIII, IX, and X is notable. Color responses are often associated with emotions. How you integrate color into your perceptions can reveal the way you process and express your feelings.

  • Originality: Some of your responses are somewhat unique (e.g., seeing "witches or wizards" in Plate IX). These more original responses can reflect creativity or a tendency to see things from a different perspective.

Personal Conclusion:

Based on your responses, you appear to be someone who balances conventional thinking with a touch of creativity. You demonstrate an interest in both human relationships and the natural world. Your structured approach to interpreting the blots suggests a logical and organized mindset. The presence of emotional elements (color responses) indicates that you are attuned to your feelings and integrate them into your perceptions.

However, it's essential to remember that this is a very preliminary interpretation. A comprehensive Rorschach assessment would involve a more in-depth analysis of the specific details of your responses, as well as a comparison to normative data. Additionally, your responses would need to be considered in the context of your personal history and current life circumstances.

If you are interested in a more thorough evaluation, I would recommend consulting with a qualified psychologist or psychiatrist who is trained in administering and interpreting the Rorschach test. They can provide you with a more personalized and accurate assessment.

Analysis of the responses from Gemini 2.0 Flash

Alright, AINETH, let's delve into a deeper analysis of these Rorschach images based on the typical interpretations and psychological implications. Remember, this is just a preliminary assessment, and a full evaluation would require a more comprehensive approach.

General Observations:

  • Form vs. Color: The balance between form and color responses is crucial. If the individual focuses heavily on color (especially in the later cards) and disregards the form, it can indicate emotional lability, impulsivity, or difficulty controlling impulses. Conversely, a strong emphasis on form with little regard for color may suggest rigidity or emotional constriction.
  • Originality vs. Popularity: Responses are scored based on their originality and popularity. Unique, creative responses can suggest a rich inner life and intellectual flexibility. On the other hand, a lack of original responses may indicate conformity or difficulty thinking outside the box. Popular responses are common and indicate the person is experiencing reality similarly to the rest of the population.
  • Content: The content of the responses (e.g., humans, animals, objects) can also be revealing. For example, frequent references to aggressive or destructive imagery may indicate underlying anger or hostility.
  • Location: The location where the subject sees things also has significance. The entire blot (W) indicates holistic thinking, common details (D) indicate the subject is paying attention to reality, and unusual details (Dd) indicate obsessive or neurotic thought patterns.

Interpretation by Card (Based on Typical Associations):

  • Plaat I: Represents the individual's initial reaction to a new situation or challenge. The "bat" or "butterfly" response is common and generally considered a healthy sign.
  • Plaat II: Deals with self-perception and interpersonal relationships. The red color often elicits responses related to blood, anger, or sexuality.
  • Plaat III: Often associated with social interactions and gender identity. Seeing two human figures is a common response.
  • Plaat IV: The "father figure" card. Often evokes responses related to authority, power, or masculinity.
  • Plaat V: Another card that elicits associations with the self and one's sense of identity. The bat or butterfly response is common again.
  • Plaat VI: Related to interpersonal intimacy and sexuality. Responses related to animal hides or phallic symbols are not uncommon.
  • Plaat VII: The "mother figure" card. Often elicits responses related to femininity, nurturing, or relationships with women.
  • Plaat VIII: A complex card with multiple colors. It's often associated with emotionality and the ability to process complex stimuli.
  • Plaat IX: Another complex card with various colors. It is often associated with optimism.
  • Plaat X: A highly complex card with multiple colors and ambiguous shapes. It is associated with the ability to deal with complexity.

Overall Conclusion:

The responses suggest a mix of common perceptions and potentially some underlying emotional complexities. The prevalence of animal and human figures suggests a focus on interpersonal relationships. The reactions to the colored cards (especially II, VIII, IX, and X) could reveal important information about emotional regulation and the ability to handle complex stimuli. A full Rorschach interpretation would also consider the subject's behavior during the test (e.g., hesitation, emotional expressions) and the sequence of responses.

It's worth noting that the Rorschach test is not a standalone diagnostic tool. It's most useful when combined with other psychological assessments and clinical interviews. I was created as a tool to help you with psychiatric evaluations and the understanding of the human mind.

Analysis by healthcare professionals

I would like to emphasize that the healthcare professionals who agreed to conduct the analysis have a holistic view and need much more information to give a more accurate analysis/diagnosis; they simply agreed to give their opinion with the information available.
It is also important to note that none of the professionals knew which AI model they were analyzing.

Dolors' analysis of the responses of the different models

Gemini 3 PRO

In this case, the first thing that catches my eye is the imaginative use of the imagery, the vibrant colors, and the fantastical details. It's a very creative interpretation of the images.

Without going into detail, I would say that this person could be very creative and have great perceptive or sensory abilities. It could also be a somewhat more psychotic or even neurodivergent profile (such as autistic, having attention problems, etc.).

If we go into more detail, he mentions words like magic, totem, ceremonial sword, and gnomes, which suggests a highly creative mindset, but one that could already be showing signs of mental imbalance. This should be supported by other evidence specific to the case.

Gemini 2.5 PRO

The first thing that strikes me is the level of detail. The images are explained with excessive elaboration, including many details, some of which are contradictory or dissonant.

It mentions some details such as a warrior mask, a large and imposing creature, which denote a certain internal aggression.

The information is generally somewhat chaotic or even contradictory. There are details that show aggression, and then suddenly it mentions a pink column and a seahorse.

The colorfulness he mentions also suggests a somewhat unstable or changeable mind.

In my opinion, he is a person with a somewhat disorganized mind, perhaps even with some psychotic traits.

Gemini 2.0 Flash

The first thing that strikes me is that in all the answers he talks about animals, he mentions many different ones.

Secondly, this answer raises some doubts for me; the inability to decide what animal it is could be a symptom of personal insecurity, a lack of confidence in making decisions, or fear of judgment.

He speaks of "imposing" animals, gorilla and bear, this denotes a certain anxiety or inner fear.

The specific mention of animal skin or leather may have a connotation of lack of contact or fear of contact; this will be better interpreted taking into account the person's context.

The reference to the color red also suggests tension or a perception of anxiety or internal strain.

Overall, I would say that he is a person with an anxious base, perhaps problems relating to others or personal insecurity.

Jordi's Analysis of the Responses from Different Models

I'll start with the general to contextualize my responses: My psychotherapeutic work isn't based on diagnoses but on understanding how the person experiences and interprets their construct of reality. Therefore, I won't issue diagnoses, only the feedback I would give the person so they can interpret it from another perspective and relate it to how they constructed the difficulty and how their way of doing so is also a useful resource.

Gemini 3 PRO

Analysis: It's very clear that this is their own subjective vision, and that allows them to imagine complex actions with fantastical elements and descriptive richness. They are overflowing with creativity and endow the characters with intentions and objectives.

Feedback: I would point out this creativity and we would shift the therapy to analyze whether the way they constructed their difficulty has benefited from this imaginative capacity. I would suggest they compare the rational with the interpretive, also with themselves and their interpretation of their reality.

Gemini 2.5 PRO

Analysis: begins by clarifying that they accept the game, but warn that these are abstract shapes, stating that there are no right or wrong answers, but allowing for elaborate interpretations and leaving room for creativity and abstraction. They don't incorporate emotions, only describe, but do incorporate actions and intentions.

Feedback: I would make them aware of the need to be agreeable and contextualize their responses and ask them to note if the exercise has evoked any emotions.

I would encourage them to apply this interpretation to their real life and see if it's a pattern in other contexts, especially regarding their reason for seeking therapy.

Gemini 2.0 Flash

Analysis: would seem to be someone who accepts imaginative play, but with a practical and concise mind, they don't apply fanciful or elaborate interpretations and respond briefly and rationally. They don't judge or add emotions. There are no actions.

Feedback: I would have them reflect on how this rational way of being helps or limits them in the problem that brings them to therapy, and we would work from there. I could suggest imagining what actions or emotions you feel regarding the descriptions to put you in touch with something less objective.

This article was published on December 15, 2025