How we rank Telegram bots
Nine criteria, public weights, hard filters and a written reason for every editorial adjustment. Sponsorship never affects the score.
Criteria and weights
| # | Criterion | Weight | How we measure |
|---|---|---|---|
| 1 | Functionality | 25 | How completely the bot solves the task of its subtopic. Measured with a fixed test scenario per subtopic (for translators: 5 phrases × 3 languages, a voice message and a photo with text). |
| 2 | Reliability & speed | 15 | Share of successful automatic checks over 30 days and median response time to /start. |
| 3 | Popularity & trust | 15 | Monthly users when the bot publishes them, subscribers of the bot's channel, mentions in independent sources and ratings in other catalogs. |
| 4 | Price & free tier | 10 | Free-tier limits compared to competitors and how transparent the prices are. |
| 5 | Usability | 10 | Onboarding, inline mode or Mini App, buttons versus commands, number of steps to a result. |
| 6 | Safety & privacy | 10 | What data the bot asks for, whether a privacy policy exists, whether it is official, and the absence of suspicious patterns. |
| 7 | Freshness | 5 | Date of the last visible update and whether support answers. |
| 8 | Localization | 5 | Russian and English interface, availability in the target regions. |
| 9 | No intrusive ads | 5 | Frequency and aggressiveness of promotional messages. |
Weights are global by default; a subtopic may use its own profile (for anti-spam bots reliability is 25 and price is 5). An editorial adjustment is limited to ±10 points and is always published. Sponsorship never affects the score: sponsored cards are placed separately and labelled.
Hard filters (disqualification)
- No reply to /start in three consecutive checks.
- Payment or phone number required before any function is shown, without explanation.
- Signs of scam or phishing, copying someone else's brand, fake numbers.
- Violation of Telegram rules or the law: piracy as the main function, gambling, grey schemes.
- No Russian and no English interface.
Process
- Collect 15–40 candidates per subtopic: Telegram search, other catalogs, GitHub, Habr, Reddit, Product Hunt, official company bots, what AI search engines already cite.
- Automatic liveness check and profile snapshot (name, about, avatar, inline mode, public monthly users).
- A hand test against the subtopic's scenario with screenshots and facts.
- Scoring on nine criteria; a bot with an editorial adjustment shows the written reason on its page.
- Publication with the check date. Liveness is re-checked weekly, the full score quarterly or after complaints.