Telegram bot marketing: send ad traffic to a bot, not a website

Telegram's self-serve ads can't link to websites, only to channels and bots. Why a bot is the better landing spot anyway, what it must do to pass review, and how to build one.

By Andrej Ruckij · · 8 min read

Why send Telegram ad traffic to a bot instead of a website?

By Andrej Ruckij · November 14, 2026 · Telegram’s ad rules and bot docs checked the same day

TL;DR: On Telegram’s self-serve ad platform you don’t get to pick a website: “Links to external sites are not allowed”, so the ad opens a channel or a bot. A bot is the better of the two for selling, for three reasons. It opens inside Telegram in one tap. Once someone presses Start, the bot can message them again, which a website visitor never allows. And a ?start= code in each ad’s link tells you which ad brought each user. The catch is that the bot has to be a real bot: Telegram declines “mostly noninteractive bots designed to redirect elsewhere”. Build a welcome that gives something useful first, then the offer, then follow-ups that don’t get you blocked. It’s the funnel we build for clients, and with CPMs around half a dollar, the views are cheap enough to test it properly.

This is the funnel we build and run for clients: ad, bot, offer, follow-ups. Telegram ads at Primores.

On Telegram’s own ads, a website isn’t on the menu

Telegram’s getting-started page is plain about it: “All links included in the Text and URL field must link to a channel or bot on Telegram, using the format t.me/link or @link”, and “Links to external sites are not allowed” (Telegram).

The ad guidelines also carry rules for websites, such as “Inactive, expired and inaccessible websites are not allowed” (guidelines), but Telegram’s pages don’t say which accounts can use a website. On the self-serve account, the question isn’t bot or website. It’s bot or channel.

For selling, I’d pick the bot almost every time. A channel is good for building an audience. A bot can take an order, open an app, send an offer and tell you exactly which ad worked.

Why a bot sells better than a website would

It stays inside Telegram. The user presses Start and they’re in the bot, with no browser and no new login. Telegram describes bots as having “seamless authorization and notifications through Telegram out of the box”, unlike websites (Telegram). Vendors who sell this traffic, such as Mondiad, claim mini apps “remove the biggest conversion killers: external redirects, app installs, and slow-loading landing pages” (Mondiad). That’s a vendor’s claim with no numbers, but the mechanics are real: the user never leaves the app.

The user becomes reachable. “Bots can’t start conversations with users. A user must either add them to a group or send them a message first” (Telegram). Read that the other way round: once someone presses Start, your bot can write to them. A website visitor who leaves is gone unless you pay to retarget them. A bot user is a contact.

You know which ad worked. Telegram’s guidelines say: “If the destination is a Telegram bot, links may include start parameters for the bot” (guidelines). Put a different code in each ad, like t.me/yourbot?start=ad12, and the bot receives it when the user presses Start. Codes can use letters, digits, _ and -, up to 64 characters (Telegram). Telegram documents this for logins and commands; we use it to tell ads apart, and one developer’s write-up does the same (DEV).

It can sell inside Telegram. A bot can open a mini app, which Telegram says “can completely replace any website” (Telegram), and take payment for digital goods in Stars (Telegram). For physical goods or a sign-up on your site, the bot can still hand over a link, after it’s done its job.

The best public number I found isn’t on Telegram’s own ad platform, so take it as direction, not proof. In a PropellerAds case on their own Telegram ad format, sending users into a Telegram group before the website cost $5.80 per conversion, against $23.64 for sending them straight to the site (PropellerAds). The budgets differed and it was a group, not a bot. I couldn’t find a clean bot-versus-website test with equal budgets.

What the bot must do to pass review

Telegram reviews the destination as well as the ad. The rules that catch bots:

  • 4.1: the bot needs a profile image and a complete description (guidelines).
  • 4.2: “Bots must respond to commands properly”, and “Destination bots must respond to commands properly on both mobile and desktop devices.” Bots also “must be beneficial to the user”.
  • 4.3: “Destinations must match the ad text and may not be used only for redirecting to other landing pages.” Banned outright: “Mostly noninteractive bots designed to redirect elsewhere.”
  • 4.7: the bot answers in the same language as the ad and the channels you target.

Rule 4.3 is the one that matters. A bot whose only message is “click here to go to our site” breaks it. The bot has to give something in Telegram first.

How the funnel is built

This is the shape we use:

  1. Ad with a button to t.me/yourbot?start=<code for this ad>.
  2. Start. The bot saves the code, so every later step is tied to the ad.
  3. Welcome that pays off at once. Something the ad promised: a price list, a quiz result, a free first lesson, a calculator. This is what passes rule 4.3, and it’s also why people stay.
  4. The offer. A button that opens a mini app, a Stars invoice, or a link to your sign-up page, now that the user has a reason to click.
  5. Follow-ups to people who didn’t buy: a reminder, then an offer. How to do that without getting blocked is its own article: how to message your bot users.
  6. Measure per ad. Users, offers opened, sales and blocks, all by start code. Telegram tells the bot when a user blocks it (Telegram), so you can see if one ad brings people who leave fast.

The traffic side is cheap enough to test this properly. On our own campaigns the median CPM was $0.48 per 1,000 views, so $1,000 buys about 2 million views at that price; see CPM by region for where it’s cheaper and where it isn’t.

The honest costs

People block bots. On Telegram’s own bug tracker, users ask for a universal stop command because “The only reliable solution is often to block the bot” (Telegram bug tracker). A user can unblock, but until they do, you can’t reach them at all. Every follow-up spends a little of that goodwill.

No pixel. There’s no Meta-style pixel inside a bot. Tracking means start codes plus your own events. If you also run Meta ads, one developer’s write-up shows passing Meta’s click id through the start code and sending conversions from your server (DEV).

A bot is software. It needs building, hosting and someone to fix it when it breaks. It has to work on phone and desktop or review declines it.

No proof it always beats a website. Every “bots convert better” claim I found was a vendor’s assertion or a single case. We build it this way because a bot keeps the user reachable and tells us which ad worked, but I’d still test it on your offer.

How to do it

What I’d do this week to get a first bot funnel live:

  1. Write the promise. One sentence: what the user gets the moment they press Start. That line is both your ad’s hook and your review pass.
  2. Build the welcome. One message with that thing, plus one button for the next step. Keep it in the ad’s language.
  3. Set up start codes. One code per ad, saved with the user. A spreadsheet export is enough at first.
  4. Fill the profile. Profile image, full description, and commands that answer, /start first. Test both on a phone and a laptop.
  5. Add the offer step. A mini app, a Stars invoice or your sign-up page behind a button. Never as the only thing the bot does.
  6. Pick channels in the same language; we start with 20 to 40, as a rule of thumb: how to find Telegram channels to advertise in.
  7. Run a small test, then read it per start code: users, offers opened, sales, blocks. Our calculator helps size the budget.

For the full picture of buying on Telegram, see the guide to advertising on Telegram. For where Telegram fits next to other channels, see alternative ad channels.

Sources

https://t.me/aidaily_business