Telegram’s AI bot market appears to be moving away from general-purpose model wrappers and toward tools built around a specific task.
New market data for August 2026 shows that general AI wrappers lost 11% of their combined monthly active users in a single month. AI photo and video bots declined 18%, while AI character and roleplay bots fell 19%.
The broader trend is even more pronounced over time. General AI wrappers are now 61% below their December 2025 peak, according to an independent analysis of Telegram’s public bot statistics.
That does not mean people have stopped using AI inside Telegram. It suggests that simply placing several popular models behind one bot is becoming less compelling on its own.
What the numbers show
The August dataset tracks Telegram bots with at least 10,000 visible monthly active users. Across the AI categories measured:
- General AI wrappers fell by 1.51 million users, or 11%, to 11.8 million combined MAU.
- AI photo and video bots fell by 458,000 users, or 18%, to 2.02 million combined MAU.
- AI character and roleplay bots fell by 226,000 users, or 19%, to 960,000 combined MAU.
Only 12 of the 73 general AI wrappers in the dataset grew during the month. Seven of 38 AI photo and video bots increased their audience, while nine of 17 character and roleplay bots grew.
The figures describe category-level movement rather than the performance of every individual product. There are still successful bots in each group, but the overall direction is negative.
This is a notable change from earlier in the year. A previous survey identified 183 AI bots with a combined 44.3 million monthly active users, illustrating how quickly Telegram had become a distribution channel for AI products.
The latest numbers suggest that the initial expansion phase is giving way to a more selective market.
What counts as an AI wrapper?
A general AI wrapper usually gives users access to one or more external models through a Telegram chat. Instead of opening ChatGPT, Claude, Gemini or a separate image generator, the user sends a prompt to the bot and receives the result inside Telegram.
That remains convenient, particularly in regions where direct access to some AI services is limited. Many of the AI chatbots listed on Telegram AI Tools also combine several models, media generators and payment options in one interface.
The problem is that access alone is becoming easier to copy.
Two bots may offer similar model menus, similar subscription plans and nearly identical commands. If neither has better memory, workflow integration, privacy controls or a clearly defined use case, users have little reason to remain loyal to one of them.
The underlying models are also increasingly available through official apps, websites and other aggregators. A Telegram wrapper therefore needs to provide more than a different chat window around the same model.
Specialized does not simply mean smaller
The August data does not show that every niche AI category is growing. Photo, video and roleplay bots also declined as groups.
The more useful distinction is between bots that expose a collection of models and bots that solve a recognizable problem.
One counter-example in the research was an AI-positioned fortune-telling bot that grew from 46,000 to 164,000 monthly users. The analysis also identified AI song generators as a smaller category worth tracking.
These products do not ask the user to choose between model names or understand technical differences between providers. They begin with an outcome: generate a song, interpret something, edit a particular type of media or complete a defined workflow.
That clarity can make onboarding easier. It also gives users a reason to return that is not dependent on whichever bot currently offers the lowest price for the same underlying model.
For developers, the lesson is not necessarily to build an astrology or music bot. It is to make the product’s purpose understandable before asking the user to send their first prompt.
Telegram still has advantages for AI products
The decline should not be read as evidence that Telegram is becoming irrelevant to AI developers.
Telegram continues to provide native bot discovery, file handling, payments and interfaces that work across mobile and desktop. Its Bot API now supports features designed for AI interactions, including streaming replies, threaded conversations and structured rich messages. Telegram’s official bot documentation explicitly presents AI agents as a core use case.
Launching through Telegram also removes several steps associated with a standalone application. Users do not need to create another account, install another app or learn an unfamiliar interface.
But easier distribution also creates more competition. When dozens of bots can connect to the same models and reproduce the same basic experience, the advantage moves away from model access and toward product design.
Persistent memory, integrations, reliable task completion, transparent pricing and a focused workflow are harder to reproduce than a list of model names.
The data has limitations
Telegram’s visible MAU number is useful for comparing direction, but it is not a complete measure of the market.
The figure represents accounts that interacted with a bot during a rolling monthly window. It does not identify unique users across multiple bots, so one person can appear in several category totals.
Activity taking place entirely inside a Telegram Mini App may not be reflected in the bot’s public number either. Bots used in groups can also accumulate activity differently from bots designed for private conversations.
MAU does not reveal revenue, retention or how frequently each person uses a product. A giveaway, advertising campaign or regional restriction can substantially change the visible number without showing whether the underlying business is healthy.
The figures are best treated as a directional market signal rather than a precise count of Telegram’s AI audience.
What this means for Telegram AI tools
The first generation of Telegram AI bots largely competed on access: more models, fewer restrictions and lower prices inside a familiar messenger.
That proposition is becoming less distinctive.
The next group of durable products is more likely to compete on what happens after the model responds. Can the bot remember previous work? Can it connect to other services? Can it complete a multi-step task? Does it produce a specific result more reliably than a general chatbot?
General AI wrappers are not disappearing, and 11.8 million combined monthly users remains a substantial audience. But the August decline indicates that convenience alone may no longer be enough to sustain growth.
Telegram remains a strong interface for AI. The bots most likely to benefit are those that feel less like a menu of models and more like a tool built for a particular job.
