Buddhist AI "AI Buddha Zen" Reveals First Anonymous Consultation Data of 2,566 Cases in 17 Days Since Service Launch, Achieves 100% Scripture Source Citation Rate. Most Common Concern Not "Human Relationships" but "Questions About Buddhism Itself."

Image: [Image of a person meditating with a digital interface]

VeritasChain Inc. (Shibuya-ku, Tokyo; Representative: Tokatsu Uemura) conducted a survey analyzing 2,566 consultation data points received by its Buddhist AI dialogue service "AI Buddha Zen" to reveal what modern Japanese people are saying and asking AI.

General mental health surveys are conducted in questionnaire format.

However, it is known that people tend to give socially desirable answers when asked "What are your concerns?" in front of an interviewer (social desirability bias).

This survey, in contrast, statistically analyzes "true feelings" spontaneously expressed during dialogue with AI.

Its characteristics are that it uses naturally occurring data without survey bias, and that the context of Buddhism functions as a space that encourages deeper introspection.

[Data Privacy Box]
AI Buddha Zen is designed so that service providers do not view conversation content during normal operation. In this survey, in addition to category-specific statistical information, anonymized dialogue logs retained for the purpose of improving service quality were analyzed. All analyses were conducted in an anonymized state, and no personally identifiable information has been disclosed. Dialogue data used for analysis was subjected to irreversible anonymization processing after analysis completion.

◼️ Survey Overview
* Target: All consultation messages received by AI Buddha Zen
* Number of analysis targets: 2,566 (2,409 valid consultations)
* Analysis period: March 13, 2026 (service launch date) - March 30 (17 days)
* Unique sessions: 1,138
* Platform: LINE 1,169 (57%) / iOS 882 (43%)
* Average message length: 44 characters (median: 20 characters)
* Long consultations over 200 characters: 63 cases
* Analysis method: Statistical tabulation by categorizing consultation content
* Ethical considerations: Category-specific statistical information and anonymized dialogue logs were used. Quantitative analysis was conducted as statistical information, and qualitative analysis (case descriptions) was performed after processing to prevent individual identification. After analysis completion, dialogue logs were subjected to irreversible anonymization processing.

◼️ Data Processing Pipeline
This survey was conducted through the following three stages of data processing:

* Layer 1: Real-time statistics (normal operation)
AI Buddha Zen categorizes consultation messages upon receipt and increments only the statistical counter for the corresponding category. Quantitative data such as category-specific counts, time-slot specific counts, session depth, and platform-specific ratios are derived from these statistical counters.
* Layer 2: Anonymized analysis (conducted for this survey)
For keyword appearance frequency comparison by time slot ("scared," "can't sleep," etc.) and qualitative case descriptions (use by monks, trends in long consultations, etc.), dialogue logs retained for service quality were analyzed after anonymization. During analysis, user IDs were irreversibly transformed, preventing analysts from identifying specific individuals.
* Layer 3: Post-analysis processing
Dialogue data used for qualitative analysis was subjected to irreversible anonymization processing after analysis completion, making it impossible to restore the original conversation content.

All data disclosed in this survey has been processed to prevent individual identification.
Consultation content described as examples integrates and reconstructs multiple similar cases and does not directly quote specific individuals' consultations.

---

◼️ Discovery 1: The most common consultation category is "Questions about Buddhist teachings themselves"
As a result of classifying consultation content, the most frequent category was "Questions about Buddhist doctrines and practices" (12.0%).

"What is enlightenment?"
"Please teach me how to meditate."
"I want to know the meaning of the Heart Sutra."
"What does it mean to know contentment?"
"What is form is emptiness, emptiness is form?"

Such intellectual curiosity about Buddhism itself was the most common result.

The second most common was "Work and career concerns" (11.1%). A wide range of work-related worries were received, including anxiety about job change, relationships with superiors, loneliness from remote work, fear of layoffs, and stagnation in job hunting.

The third was "Anxiety and fear of the future" (6.4%). Notably, this category included many vague anxieties without a clear object, such as "I'm anxious but don't know what I'm anxious about." In Buddhism, this type of anxiety is expressed as "avidyā" (ignorance) — a state of being unable to see the true nature of things. The data clearly shows that modern Japanese people widely experience this very state.

The fourth was "Family concerns" (6.3%). The suffering unique to close family relationships was expressed, such as parental care, children's school refusal, severed parent-child relationships, and childcare stress.

[Top Consultation Categories]
1st: Questions about Buddhist doctrines/practices 289 cases (12.0%)
2nd: Work/Career 268 cases (11.1%)
3rd: Anxiety/Fear of the future 155 cases (6.4%)
4th: Family concerns 152 cases (6.3%)
5th: Love/Partners 73 cases (3.0%)
6th: Sadness/Hardship 64 cases (2.7%)
7th: Human relationships (excluding family) 48 cases (2.0%)
8th: Motivation/Procrastination 45 cases (1.9%)
9th: Anger/Emotional control 28 cases (1.2%)
10th: Meaning of life/Existential questions 25 cases (1.0%)

While the Ministry of Health, Labour and Welfare's survey on workers' health conditions identifies "human relationships" as the biggest factor in workplace stress, "human relationships (excluding family)" accounted for only 2.0% of consultations to AI Buddha Zen.

This difference suggests that while face-to-face questionnaires tend to elicit responses about "specific current problems (= interpersonal troubles)," AI tends to bring out "deeper questions underlying everyday problems."

The context of Buddhism may be functioning as a safe space to discuss existential questions.

◼️ Discovery 2: 40% of consultations concentrated between 8 PM and 11 PM. "Scared" appears 5 times more frequently at night than during the day.

Image: [Chart showing consultation frequency by time of day]
Time-slot analysis revealed a prominent peak from 8 PM to 10 PM. These three hours alone accounted for approximately 26% of all consultations, and extending to the four hours from 8 PM to 11 PM, it reached about 40% of the total.

The data shows that people confront their worries during this "golden time for introspection at the end of the day" – after finishing work, having dinner and a bath, and picking up their smartphone in bed.

[Consultation Counts by Time Slot]
9 AM: 148 cases (morning peak)
6 PM: 102 cases
7 PM: 182 cases
8 PM: 228 cases (most frequent)
9 PM: 221 cases
10 PM: 162 cases
11 PM: 125 cases
12 AM: 110 cases
1 AM: 69 cases

Even more interesting is how the "quality" of consultations changes with time. Comparing keyword appearance frequency between late night (12 AM - 4 AM) and daytime (9 AM - 5 PM) revealed the following trends:

[Keyword Frequency Box]
"Scared": 5.3 times more frequent at night than during the day
"Can't sleep": 12.2 times more frequent at night than during the day
"Loneliness": 3.1 times more frequent at night than during the day
"Painful": 2.0 times more frequent at night than during the day

The data indicates that users who remain active during late-night hours are in a more urgent psychological state.

While corporate mental health measures typically involve annual stress checks and consultation services during weekday business hours, people's minds are most shaken "at the end of the day" and "on sleepless nights."

This data clearly supports the need for 24/7 digital self-care tools.

Furthermore, differences in message length were also observed by time slot. The average for 9 AM was 64 characters, the longest, suggesting an intention to "organize and convey thoughts" in the morning.

Conversely, the average for 11 PM was 29 characters, the shortest, with many brief appeals like "always feel anxious," "can't sleep," or "I can't take it anymore."

◼️ Discovery 3: Over half continued conversations for 2 or more exchanges. The deepest conversation was 71 exchanges.
Analysis of session depth showed that only 43.7% of users ended the conversation with a single question, while over half (56.3%) continued for two or more exchanges.

* 1 exchange only: 497 cases (43.7%)
* 2-3 exchanges: 550 cases (48.3%)
* 4-5 exchanges: 52 cases (4.6%)
* 6 or more exchanges: 39 cases (3.4%)

The longest dialogue session reached 71 exchanges, followed by 47 and 43 exchanges.
The user with 71 exchanges has been using the service continuously since its early days, suggesting they utilize AI as a partner for daily introspection.

For users engaging in deep conversations, themes like the following were concentrated:

* Marital relationships (16 exchanges)
* Children's school refusal
* Economic anxiety due to rising prices (19 exchanges)
* The grand question of "world peace" (43 exchanges)

These users actively constructed dialogues by refuting, delving deeper, and self-analyzing, rather than passively receiving AI responses.

Meanwhile, the 63 long consultations exceeding 200 characters included many details that could only be openly discussed with AI, such as detailed family situations, histories of workplace relationships, and dilemmas at life's turning points.

◼️ Discovery 4: Monks and Buddhist researchers also consult AI for "Dharma talks."

Image: [Image of a Buddhist monk]
Among the 2,566 cases, users identifying themselves as monks were confirmed. Additionally, there was a case where an individual operating multiple Zen meditation sessions at a temple described in detail (over 1,500 characters) the differences in character between Friday and Sunday Zen sessions, the policy for sesshin (intensive training), and how to create Dharma talks, then sought evaluation from the AI.

Users of Buddhist AI are not just general citizens with worries.
Buddhist specialists themselves are utilizing AI as a "sounding board" to objectively re-evaluate their own propagation activities.

◼️ Discovery 5: The "weight" of messages differs between morning and night.

Image: [Image of a person looking at a smartphone in bed]
When the average message length of consultations was tabulated by time slot, two peaks appeared: 9 AM (average 64 characters) and 12 AM (average 53 characters).
Conversely, the average was shortest at 3-4 PM, at 29 characters.

This trend suggests that in the morning, there is a need to "verbalize and organize thoughts," while at night, there is a need to "simply vent," with different motivations driving usage.
Even for the same "work concerns," in the morning, it might be "anxiety," while at night, it tends to be expressed briefly and urgently as "I hate work" or "I can't do it anymore."

This data indicates the possibility that adjusting the response tone according to the time of day (more specific advice in the morning, more listening and empathy at night) could be effective in the design of AI dialogue services.

◼️ Discovery 7: All 2,392 AI responses clearly cited scriptural sources. The most cited scripture was the "Kālama Sutta."

Image: [Image of ancient Buddhist texts]
AI Buddha Zen is designed to clearly indicate the source of all its responses from Buddhist scriptures by scripture name, chapter, and verse number.

Aggregating RAG (Retrieval Augmented Generation) logs during the analysis period showed that for 2,392 total searches, there were 2,392 source citations and 0 cases of no match, achieving a 100% source citation rate.

While general generative AI is said to include factually incorrect information in up to 27% of its responses, AI Buddha Zen structurally eliminates fictional citations (hallucinations) by searching and citing verses from a database of over 10,000 scriptural verses according to the content of the consultation.

[Top 10 Most Cited Scriptures]
Image: [Chart showing top 10 most cited scriptures]