Bots on Telegram channels can attract tens of thousands of subscribers, participate in giveaways, and even mimic live conversation. We decided to test the scale of this phenomenon on our own channel, KLJOMPUS: we conducted a five-day experiment using Telegram Gifts and discovered who is actually behind those huge subscriber numbers.

We decided to test this on our own Telegram channel KLJOMPUS; at the time of launch, we had about 1,500 subscribers.

We weren't interested in promoting the channel. We wanted to see what kind of audience would show up on its own for a regular Telegram Gift giveaway and what would remain of that audience.

Day 1 — 59,597 participants

The very first draw yielded a result we hadn't expected.

It quickly became clear that a significant portion of the new audience looked unusual. Accounts with names of a certain type were appearing by the thousands in the sign-up log:

H0t4e, C0ld2, N379e, 4lph44, 4ng3l, Dr1f73r, F4d3i, D1g1t4l, M0l3cul3.

The words, numbers, and individual characters changed, but the principle behind the formation of the names repeated itself over and over again.

These weren't just a few random accounts scattered among the regular audience.

There were thousands of profiles like that.

And the most interesting thing is that this massive influx had virtually no impact on the channel's operations.

Even with tens of thousands of followers, posts continued to receive only a few hundred views. At one point, the statistics showed 306 views, 14 reactions, and one share.

Tens of thousands of accounts were listed on the counter, but we didn't see any audience corresponding to them.

The bot farm is starting to run out of steam

The next day, we held our second Telegram Gift giveaway.

The conditions remained the same. But we were unable to repeat the result of the first draw.

The next major wave resulted in:

33,452 new subscriptions in the past 24 hours.

Then: 15,321.

After that, the number continued to decline, and by the fifth day, it had dropped to about 8, 000 accounts.

The trend turned out to be very telling:

59,597 → 33,452 → 15,321 → … → ~8,000

An account that had already subscribed to KLJOMPUS for the previous giveaway remained subscribed to the channel. The next day, it could not become a new subscriber again until it unsubscribed from the channel. For this reason, we set the start time of the new giveaway to one hour before the end of the current one.

To maintain the initial pace, more and more accounts were needed.

There were fewer and fewer of them.

Some of the accounts that had already participated began leaving the channel at the same time, so some might have returned in subsequent waves. For this reason, the daily influx cannot simply be added up to determine the total audience size: while some accounts were joining, others were already leaving in a second wave.

But the trend was clear. Each subsequent draw resulted in a significantly lower volume.

The bot farms ran out of steam right in the middle of the experiment.

They knew how to talk

Subscriptions weren't the only automated activity we observed.

After a new post appeared on the channel KLJOMPUS, 3–10 new accounts would appear in the chat almost immediately.

The accounts could have posted a perfectly coherent message that was relevant to the topic of the post or what was happening in the chat.

But there was one detail.

These profiles regularly featured the same or very similar 18+ content in their profile pictures.

The accounts appeared in groups, almost simultaneously; they had a similar design and exhibited the same pattern of behavior.

Taken together, this pointed to a network of linked accounts.

And here, the experiment revealed yet another problem with Telegram.

A bot no longer has to look like a bot. It can have a well-designed profile, monitor new posts, and leave messages that seem perfectly human.

Day Five

By the end of the five-day experiment, the Telegram channel KLJOMPUS had reached:

80,000 subscribers

The net difference between the initial audience and the peak was approximately 79,945 accounts.

But if you looked at the result rather than the score, the picture was completely different.

We haven't gained a significant number of new regular readers.

We did not see the expected increase in the number of active chat participants.

We didn't manage to engage the audience that came for the giveaways, took an interest in KLJOMPUS , and stayed.

The conversion of nearly 79,000 additional subscribers into a loyal, active audience turned out to be zero.

We also reviewed the Telegram Stars ad separately

At the same time, we tested Telegram Stars’ built-in channel promotion feature. We purchased Stars, paid for the ads, and launched the campaign directly within the Telegram ecosystem.

The result turned out to be practically the same— we didn't see any influx of active users.

For us, this was yet another telling moment in the experiment. Even when we’re not talking about giveaways or third-party services, but rather paid promotion within Telegram, simply buying reach doesn’t necessarily mean you’ll attract real readers.

In our case, the money was spent, the ad campaign was effective, but it didn't yield any results for the channel's live audience.

What Happened After the Experiment

Once the giveaways were over, almost the entire audience began to leave. And this is evident not only from our internal Telegram statistics.

According to TGStat, the channel gained 79,945 new subscribers in May 2026. By comparison, the increase was only 24 people in March and 9 in April.

After the experiment ended, the influx of new subscribers practically stopped: 0 new subscribers in June and only 2 in July.

By August, the channel had about 1,400 subscribers left—practically the same number as before the experiment began.

Thus, the external service recorded the same pattern we observed within Telegram: a sharp influx of tens of thousands of accounts during giveaways and a near-complete return to the original audience size once they ended.

Conclusion KLJO

We wanted to test a simple question: To what extent is Telegram actually a “live” messaging app?

How many real people are actually behind channel subscribers, and to what extent do the numbers that channels and projects use to measure and demonstrate their success reflect their actual audience?

The experiment revealed a huge gap between these two things.

Tens of thousands of accounts can create the appearance of a large community in a matter of days: following, participating in giveaways, and even leaving thoughtful comments. But when almost all of that crowd disappears and the number of real readers remains virtually unchanged, the subscriber count itself loses its meaning.

There’s another paradox with Telegram: multi-accounting in TMA projects. In pursuit of rewards, airdrops, and in-game advantages, a single person can use multiple accounts at once. For the project’s statistics, each of these accounts is counted as a separate user, even though they’re all controlled by the same person.

As a result, another disconnect arises between accounts and users: part of the audience consists of bot farms, and part consists of multiple accounts belonging to real users, yet the public metric combines all of this into a single, attractive figure.

That is precisely why Telegram's large user numbers should be evaluated not based on how impressive they look, but on how many real, unique people are actually behind them.

Because if one person is represented by ten accounts, and a significant portion of the rest belong to bots, the counter's reading no longer reflects the actual audience size.

That's not an audience anymore. It's just a number.