Guide · Telegram

How do Telegram bots and Mini Apps get recommended by AI?

Updated 2026-11-10 · kagentx

Short answer

A bot's own page has almost no content, so put a web landing page on your own site. On 2026-11-10 we fetched t.me/Stickers and t.me/BotFather: status 200, and the raw HTML has only a title, og:title and og:description (one sentence), with no meta description or canonical; t.me's robots.txt is 404. Telegram's docs state: description up to 512 characters, about text up to 120 characters, commands up to 32 characters. The landing page should state what it is, the commands, the limits and the update date, and link to t.me. Whether AI cites t.me pages: no evidence.

What does a t.me bot preview page contain in its raw HTML?

Very little: a name and a sentence or two. On 2026-11-10 we fetched two public official bots with curl (ordinary Chrome identifier, not logged in, no scripts run):

Itemt.me/Stickerst.me/BotFather
Status code and sizeHTTP 200, 11,074 bytesHTTP 200, 11,169 bytes
<title>Telegram: Launch @StickersTelegram: Launch @BotFather
og:titleStickers BotBotFather
og:descriptionCreate Telegram Stickers and get usage stats for your stickers with this bot.BotFather is the one bot to rule them all. Use it to create new bot accounts and manage your existing bots.
meta description, canonical, robots metaNoneNone; the response headers have no X-Robots-Tag either

t.me/robots.txt returns 404, which means no rules at all. The sentence shown on the page is the same as og:description. The title of both pages is “Launch @name”, which does not say what the bot does; the one sentence is all there is. We found no statement from Telegram about t.me pages and AI crawlers.

This means that all AI sees of your bot is a name and a sentence or two, and that sentence is the text you set in BotFather, so write it carefully.

What are the character limits for the name, about text, description and commands in BotFather?

Telegram's official docs (core.telegram.org/bots/features and the Bot API page, viewed 2026-11-10; the pages show no update date) state:

FieldBotFather commandLimit the docs stateWhere users see it
Name/setnameBot API setMyName: 0–64 characters; the features page gives no numberAt the top of the chat and on the t.me page
About text/setabouttextUp to 120 characters (Bot API: 0–120)On the bot's profile page; sent along with the link when the bot is shared
Description/setdescriptionUp to 512 characters (Bot API: 0–512)Under “What can this bot do?” at the start of the chat
Commands/setcommandsA command is up to 32 characters, lowercase English letters, digits and underscores only; a command description is 1–256 characters; at most 100 commands (Bot API)The command menu next to the input box

Name, about text and description can each be set per language. As for which field shows on the t.me page, all we can confirm is that the sentence on the t.me/Stickers page matches the bot's about-style text; which exact field it is, we did not verify.

The about text is only 120 characters, so the first sentence should say plainly “who this bot is for and what it does”, not a slogan.

Why does a bot need a web landing page?

Because AI search mostly reads web pages, and a bot's own page has almost no content. The t.me page has a sentence or two; the bot's features, commands and usage are inside Telegram, where crawlers cannot go. Mini Apps are the same: Telegram's docs (core.telegram.org/bots/webapps, viewed 2026-11-10) describe them as JavaScript web interfaces that run inside Telegram and open from a URL the bot specifies; they are normally used inside the chat, crawlers cannot see their content, so a public page is needed to say what they do.

Put the landing page on your own site as plain HTML with the body rendered on the server (see Do AI crawlers render JavaScript?). An example structure (invented):

Title:        YourBot: <what it does> on Telegram
First line:   YourBot is a Telegram bot that <does X> for <who>.
Open it:      https://t.me/YourBot  (plain link in the text)
Commands:     /start  - <what happens>
              /help   - <what happens>
Limits:       <what it cannot do>, <supported languages>
Updated:      2026-11-10
  1. 1. Say what it is in the first sentence

    “YourBot is a Telegram bot that does Y for X.” Use the same sentence as the about text in BotFather.

  2. 2. Put the t.me link in the body

    Use a plain link whose text gives the bot's name and purpose, not just “click here”.

  3. 3. Write the commands and limits

    One sentence per command; state the supported languages and what it cannot do. AI needs these concrete details to quote you.

  4. 4. Give the update date

    Bots change; a date tells AI and users that it still works.

  5. 5. Add use-case pages

    Use cases such as “what to do with a Telegram bot”, one page each; for how to write them, see How to write "X for Y" use-case pages that AI will cite.

Can AI read bot directory pages?

We obtained only some samples; the rest is unverified (2026-11-10):

Directoryrobots.txtPage sampleUnverified
storebot.meUser-agent * Disallow: (empty, nothing blocked) plus a Sitemap indexOne bot page: HTTP 200, title “Zenmoney Bot - StoreBot”, canonical points to itself, no meta description, one JSON-LD blockWhether other bot pages match
botostore.comUser-agent * Allow: /; blocks /install/ and parameters such as ?do=open_linkOnly a collection page was obtained: 200, with a title, meta description and canonicalThe content of an individual bot page
tgstat.comcurl could not connectNot obtainedEverything is unverified

We have no evidence on whether these directories lead to AI citations. All we can confirm is that your bot's directory entry should keep the same name and the same sentence as your landing page. For the more general question of how to check directory pages before submitting, see Can the GPT Store and AI agent directories get you recommended by AI?.

An example of the approach: catching Telegram searches with a web guide

If you build something for Telegram, you do not have to build a bot. MochiStamp is a site for making stickers, and it has a page, MochiStamp's Telegram sticker guide (English page; MochiStamp is a site for making LINE, WhatsApp and Telegram stickers), that answers “how to make Telegram stickers and add them to your own pack” on the web instead of hoping a t.me page gets found. On 2026-11-10 we fetched that page: HTTP 200, title “Telegram sticker maker: make and add your own sticker pack | MochiStamp”; the page has en, ar, fa and ru versions and no Chinese version (the Chinese addresses return 404), so all three language versions of this guide link to the English page.

This page's approach is also what a bot should learn: turn what users want to do in Telegram into a web page that can be found and quoted. It is an example of an approach, not performance data; whether this page has been cited by AI, we have no evidence.

Checklist before you launch

  1. 1. First look at your own bot's t.me page

    Run these three lines to confirm that the name and the one sentence are what you want people to see:

    # What does a bot's public preview page say, without logging in?
    curl -s https://t.me/YourBot | grep -io '<meta property="og:title"[^>]*>'
    curl -s https://t.me/YourBot | grep -io '<meta property="og:description"[^>]*>'
    curl -s https://t.me/YourBot | grep -io '<title>[^<]*</title>'
  2. 2. About text within 120 characters, description within 512 characters

    State in the first sentence who it is for and what it does.

  3. 3. Put a landing page on your own site

    See the structure above. First use AI crawler check to confirm your own site does not block AI crawlers.

  4. 4. Keep the name and one-line description the same everywhere

    Use the same name and the same sentence in BotFather, the landing page, bot directory entries and social accounts.

  5. 5. Ask another round after two weeks

    Ask AI the questions users would ask and record the answers and sources; for the method see How do you measure AI visibility? Four metrics and a table you can follow.

Slack apps and Discord bots are also products built inside someone else's platform; for tests of them see How do Slack apps and Discord bots get recommended by AI?, and for how mobile apps fare in app stores see How do I get my app recommended by ChatGPT?.

What sticker products actually get impressions in search, by page type, is in How do sticker and emoji packs get found in search and AI answers?.

FAQ

Can AI read my bot's t.me page?

On 2026-11-10 we fetched t.me/Stickers and t.me/BotFather: status 200, and the raw HTML has a title, og:title and og:description (which is the bot's about-style text), with no meta description or canonical; t.me's robots.txt returns 404 (no rules). The pages contain very little, just a name and a sentence or two. Whether AI has cited these pages, we have no evidence.

Do the name, about text and description in BotFather have character limits?

Telegram's official docs (viewed 2026-11-10) state: description up to 512 characters, about text up to 120 characters, commands up to 32 characters; in the Bot API the name is up to 64 characters, a command description is 1–256 characters, and there are at most 100 commands.

Does a Mini App need a web landing page?

Yes, same as a bot. A Mini App is a web interface opened from a URL, but it is usually used after login, so crawlers cannot see what is inside; for search and AI to read what you built, you need a public page that explains it.

Are bot directories (such as StoreBot and BotoStore) useful?

All we confirmed is that the robots.txt of storebot.me and botostore.com does not block the home page or content pages, and that one bot page on storebot.me has a title and canonical but no meta description. Whether they lead to AI recommendations: no evidence; we could not connect to tgstat.com, unverified.

How do you know whether AI recommends you?

Write down 5 questions users would ask (such as “is there a Telegram bot that makes stickers”) without your brand name, ask each 3 times with web search on, and record the answers and sources. One result is not a conclusion; for the method see “How to measure AI visibility”.

Sources

kagentx

Do AI answers
mention you?

The evaluation checks whether crawlers can get in and whether the text is served directly, then asks ChatGPT, Perplexity and Claude (with web search on) your buyers' questions and sees whether you are mentioned. The Basic plan is $1, and results usually take 2–3 minutes.