TG Inspector methodology

Every published conclusion should state the following six fields:

• Source. Where the underlying public data came from.
• Snapshot. When it was collected or which saved snapshot was used.
• Sample. How many posts or channels were included and how they were selected.
• Method. How the published metric or score was derived.
• Result. The factual conclusion supported by that sample.
• Limitations. What the data cannot establish and where error is possible.

Checks use public Telegram data only: subscriber counts, recent public posts, views, reactions, comments, and channel metadata.

Each published conclusion is tied to a saved snapshot. Newer monitoring data is not silently mixed into an older report or FAQ.

The cleanliness index is calculated by deterministic code with fixed rules. It combines independent signals: reach relative to audience size, view variation and maturation, reactions, comments, and metadata.

An LLM helps analyze comment language and write summaries or FAQ answers. It cannot change the numerical index, grade, or traffic-light verdict.

Signal weights and the grade scale are published on the About page.

• Repeated checks are deduplicated; only the latest available snapshot of each channel is counted.
• ERR and reaction aggregates use medians so a single outlier cannot dominate the result.
• An unavailable value remains unavailable and is never replaced with zero.
• Niche breakdowns hide groups containing fewer than five channels.
• The sample size and latest snapshot date are shown alongside the conclusion.

The index estimates statistical risk; it does not prove intentional manipulation. Content format, post age, external traffic, and missing public signals can all affect channel metrics.

Database aggregates describe channels that TG Inspector users checked or added to the registry. They are not a representative sample of every Telegram channel.