Telegram rich messages: structure readable AI bot replies

The ability to send longer, more complex Telegram messages does not automatically make a bot more useful. For an AI assistant or support bot, the real design problem is helping someone find the answer, evidence, and next action quickly on a phone screen.

Telegram introduced headings, lists, tables, quotations, collapsible sections, footnotes, and inline media for bots in its June 11, 2026 Rich Messages update. The accompanying Bot API 10.1 added sendRichMessage and sendRichMessageDraft, which can stream a formatted AI response. Versions 10.2 and 10.3 later expanded the available blocks, buttons, files, and expandable quotations.

These capabilities are useful for organizing a complex answer. They are not a reason to turn every response into a document.

Decide whether rich formatting is necessary

Plain text is usually clearer for a short answer, a confirmation, or a single instruction. Rich structure becomes more useful when a response contains:

• A comparison of several options and their conditions.
• A troubleshooting sequence that must be followed in order.
• A knowledge-base answer with a conclusion, evidence, limits, and related links.
• AI research or file analysis that combines a summary, details, and sources.

Telegram says a bot's rich message can contain up to 32,768 characters, with a Show More control after roughly the first 8,000. That capacity is a ceiling, not a target. If a reader must expand the message before seeing the conclusion, the response should be reorganized.

Use a six-layer response structure

A long answer can be arranged from most important to least important:

1. Direct answer: answer the actual question in two or three sentences instead of restating the request.
2. Key steps: use a numbered list when order matters and bullets for parallel information.
3. Conditions and exceptions: keep scope, failure cases, and changeable facts close to the conclusion they qualify.
4. Comparison details: use a table only for genuine side-by-side comparison; keep mobile tables narrow and cells concise.
5. Optional depth: put logs, derivations, and extended background in collapsible sections so they do not obscure the result.
6. Sources and next action: finish with verifiable sources, required confirmations, or a clearly labeled action.

This order gives a scanning reader an actionable answer while preserving evidence and detail for anyone who wants to continue.

Support bots and general AI assistants need different templates

A knowledge-base support bot should resolve business questions consistently and with traceable grounding. In scenarios related to YourCopilot, a useful template is “answer — steps — policy or knowledge source — human handoff if unresolved.” When an answer concerns an account, order, or price, mark anything that requires a live check. Strong formatting must not make an uncertain fact look settled.

A general AI assistant handles a wider range of requests. In scenarios related to BearChat, the structure can be “short answer — optional expansion — sources or assumptions — follow-up question.” For answers based on an image, file, or web result, identify the type of input behind the conclusion and make unverified parts visible.

Neither template should expose a model's private reasoning process. Readers need evidence they can inspect, relevant limits, and a clear next step—not a transcript of internal deliberation.

Keep streamed replies structurally stable

sendRichMessageDraft lets a bot send a rich reply while it is still being generated. Streaming reduces perceived waiting time, but it can also make headings jump, renumber lists, or reveal buttons too soon.

A more stable approach is to:

• Establish the block order before streaming the body.
• Hide buttons that could trigger irreversible actions until the answer is complete.
• Show an explicit generating state and handle stopped or failed generation.
• Recheck links, numbering, sources, and conclusions after completion.
• Keep a plain-text fallback that remains understandable if rich delivery fails.

Pre-release checklist

Before enabling complex replies in a bot, verify that:

• The opening answers the question instead of providing background only.
• Every heading, list, and table has a clear purpose.
• The message remains readable on a narrow phone screen, and links can be tapped and copied accurately.
• Collapsible sections contain optional detail rather than essential warnings.
• Sources, dates, and scope are verifiable.
• An interrupted stream tells the reader what happened and how to continue.
• The plain-text fallback is complete when rich formatting or a newer client is unavailable.
• Sensitive data and unnecessary internal processing details are not exposed.

Product boundary

This article covers Telegram platform capabilities published between June and August 2026 and information-design practices relevant to AI assistants and support bots. It does not claim that BearChat or YourCopilot currently supports Rich Messages, rich streaming, or every block described here. Check the relevant product page, bot interface, and release notes for current availability.

Sources

• Telegram Team, Smartwatch Apps, Rich Text for Bots, AI Guardians for Groups, and Much More, published June 11, 2026; accessed September 23, 2026.
• Telegram, Telegram Bot API, change logs for Bot API 10.1, 10.2, and 10.3; accessed September 23, 2026.