Peto — AI that curates shared pet memories
Peto targets urban pet owners who treat their pets as family members. It uses AI to sort out personality clues, generate companion expressions and precipitate shared memories. In the first phase, it plans to verify the closed product loop from customer acquisition via sharing to long-term recording and payment through WeChat Mini Programs.
Peto is a relationship intelligence platform for real pets. It is currently developing the MVP of its WeChat Mini Program and seeking seed round financing. The product is not targeted at virtual pets, but at dogs and cats that users cohabit with in real life. The role of AI is not to "speak out the true thoughts" for pets, but to organize photos, videos, texts, basic pet data, contexts and historical events supplemented by owners into interpretable, correctable and reviewable relationship content. Peto plans to complete the first experience and sharing with free Soul ID and fun MBTI personality cards, and then use full relationship reports, Pet Voice and Soul Timeline to support recording, return visits, node commemoration and subsequent payment.
From Scattered Materials to a Sustainable Pet Relationship Archive
Based on Peto's judgment of target user scenarios, in addition to feeding and health management, pet consumption also has demands for understanding, expression and commemoration. The 2026 White Paper on China's Pet Industry (Consumption Report) released by CIPS Pet China shows that the consumption market size of urban dogs and cats in China reached 3.126 trillion yuan in 2025, a year-on-year increase of 4.1%, and the number of urban dogs and cats is about 126 million. The report also shows that post-90s pet owners account for 42.7%, and post-00s account for 26.3%. These data indicate that pets are a large-scale consumption entry, but cannot directly deduce Peto's revenue. The niche problem Peto selects is that target pet owners will accumulate photos, short videos and chat records, but lack a default private product that continuously organizes relationship content around the same pet.
The first batch of users focuses on young pet owners aged 20 to 35 in first- and second-tier cities who are willing to shoot and share pet content. The growth experiment prioritizes dog owners to obtain high-density event samples in dog walking, training, travel and daily interactions, and the product structure supports cats from the first version. For such users, the demand is not only to get a comment of "it is cute", but also to understand the pet's behavior clues more specifically, preserve important nodes, and get a relationship-focused review on birthdays, first meeting days, trips or growth changes. Photo albums and social platforms solve the problems of preservation and distribution, and pet-raising tools solve the problems of reminder and management. Peto tries to supplement the long-term context of "the same pet, the same relationship".
AI first undertakes the work of data understanding and content organization. After users upload pet photos, supplement basic information such as name, age, breed and gender, and complete 3 to 5 light interaction choices, the system plans to generate Soul ID, including personality keywords, relationship labels, energy color, personality declaration and fun MBTI type. The MBTI here is an entertaining expression, not a diagnosis of psychology or animal behavior. The output needs to show "which clues are based on" and "how to understand possibly or tendentiously" at the same time, and retain the correction entry for owners who think the result is "unlike". This not only lowers the threshold of the first experience, but also avoids packaging image recognition or generative models as the ability to read the pet's inner thoughts.
Four Layers of AI Capabilities: Generation, Memory, Review and Security
After Soul ID, AI capabilities will enter a longer relationship chain. The full relationship report is planned to generate personality portraits, emotional needs, expression methods, getting-along suggestions and relationship labels based on the pet archive, generated personality clues and interaction contexts supplemented by owners. Pet Voice is oriented to a specific interaction: the system combines the pet's data, historical records, current photos or short videos and the user's questions to generate a short response or observation clue. The first-person perspective used in this response is only to narrow the reading distance. The product must be clearly marked as AI-generated companion expression, and does not claim that pets have language or thoughts that can be directly read by the model. For content involving refusal to eat, continuous abnormality, aggression or suspected disease, the system only provides observation records and professional help suggestions, and does not output diagnostic conclusions.
The third layer is structured relationship memory. Peto plans to precipitate actions, photos, texts, locations or contexts, owner feedback, correction results and important nodes into events, rather than piling up all chat contents indiscriminately. In subsequent generation, the system retrieves related events according to current questions and pet archives, and then the controlled generation module organizes them into responses, weekly reviews or node commemoration content. Soul Timeline is responsible for connecting these events by encounter, birthday, travel, growth and changes, supporting users to view, edit, delete single items, export all data and actively choose to share. The value of AI at this layer is summary, association and personalized expression, and the relationship facts are still provided and confirmed by users.
The fourth layer is content security and model governance. Photos, texts, short videos and generation results need to go through content review, sensitive word and risk rule interception. Key outputs retain the model version, the scope of data used and abnormal records, and establish evaluation samples of "whether it is realistic, offensive, or overly anthropomorphic" through manual sampling inspection. User feedback can be used to correct current expressions and subsequent evaluations, but cannot be described as that one click makes the model learn permanently. In the first phase, the project clearly does not develop public squares, expert consultation, health archives, e-commerce, live streaming marketing, hardware tools and complex multi-pet relationship graphs to control content costs, misleading risks and privacy exposure.
Peto's technical route does not rely on training a new basic large model. The core is to combine multi-modal input, structured event memory, retrieval call, prompt word and state constraint, content review and user control into an application layer system for pet relationships. The engineering priorities of this approach include generation quality evaluation, model call cost, media processing and storage, failure retry, permission management, deletion tasks and manual operation, rather than simply increasing the chat length or the number of functions.
Obtain the Option of Next Round of Financing with Real Verification
Peto is still in the MVP development stage. The tentative name of the company is Peto, and the entity has not been registered yet. The founding team has 4 people, collaborating around product, growth, AI engineering, data and trust security. The team members include founder Zhao Xuanlin, responsible for overall direction, product trade-off and financing promotion; co-founder Liu Ying, responsible for user insight, brand expression, community and early growth; co-founder JING, master of the University of Nottingham, ICPC regional competition bronze medalist, co-founder of Zhili AI, responsible for AI architecture, memory and engineering; co-founder Delia, master of the University of Nottingham, co-founder of Zhili AI, responsible for AI product engineering, data and trust security. There is no publicly disclosed user scale, revenue, retention or cooperative performance at present, and this article does not write the verification target as an established result.
This round plans to raise 2.5 million RMB, and the funds are planned to cover 12 months. In the first 3 months, complete archiving, Soul ID, relationship report, payment, Pet Voice, Timeline and buried points, and run through the traceable funnel of generation-sharing-payment-recording-review; from the 4th to the 6th month, focus on optimizing recording and review, and establish the baselines of D7, D30 and active new Timeline; from the 7th to the 9th month, test node commemoration content and limited high-frequency support; from the 10th to the 12th month, focus on effective niche crowds, content and channels, and prepare data packets for the next round of financing.
The verification will focus on four questions: whether users are willing to complete the free Soul ID and share; whether they are willing to pay 9.9 yuan or 19.9 yuan for the full report; whether long-term memory and review promote the second and more active recordings; whether content channels can bring interpretable natural new additions. The project plans to record the generation completion rate, sharing rate, first order conversion, refund rate, D7, D30, active recording rate, review rate, node repurchase, natural CAC and contribution gross profit through interviews, closed beta and cohort review. In the first phase, it is expected to recruit 100 to 300 strictly matched adults. The relevant figures are all verification targets and do not represent the existing user scale.
The commercialization adopts the sequence of free entry, low-price first order, node consumption and post-subscription. Soul ID and basic personality cards are used to lower the customer acquisition threshold; the full report verifies the first payment; content such as birthdays, first meeting days, travel and growth commemorations test node consumption of 39.9 yuan to 199 yuan; the Pet Voice usage package and Peto Plus membership will be tested only after continuous recording, review and model costs are controllable. In the first phase, no advertisements are used to interrupt intimate relationships, and no "survival", hunger or emotional anxiety of pets is designed as payment pressure.
At the policy level, Peto involves continuous emotional interaction and AI-generated content. The project plans to incorporate AI identity prompts, user exit, data copy and deletion, minimized collection, sensitive information training authorization, risk identification and manual disposal into the pre-launch preparation. The Interim Measures for the Administration of Anthropomorphic Artificial Intelligence Interactive Services issued by five departments including the Cyberspace Administration of China has come into effect on July 15, 2026. However, whether Peto is applicable to specific regulatory obligations and within what scope still requires special legal assessment combined with the actual role setting, functional boundary and service mode. For Peto, the investment value does not lie in claiming a huge market that has been established, but in proving that real pet relationship content can move from one-time generation to long-term use with verifiable user feedback, AI output quality, relationship retention, payment and unit economic data.
Source: Peto Investor Business Plan (Team Revised Version, August 2026); CIPS Pet China 2026 White Paper on China's Pet Industry (Consumption Report); Cyberspace Administration of China Interim Measures for the Administration of Anthropomorphic Artificial Intelligence Interactive Services.