MegaTGAI · Practical guide
Collect Telegram audiences and filter people by their messages with AI
Collect people in one list, then keep those relevant to your offer. AI assesses what they wrote rather than just looking for a keyword.
Click a screenshot to enlarge it.
Add your sources and start collecting
Open Parser → Audience collector. Paste groups and channels together, one link per line.

- Paste your full source list.
- Choose how recently a person must have been online.
- Choose approximately how many people to collect per source.
- Click Start parsing after enabling message collection below.
Collect the messages AI will need
For interest-based filtering, enable Writers only in advanced settings. The collector saves authors’ message text so AI can distinguish genuine discussion from advertising.

- Enable Writers only before starting collection.
AI receives up to 1,000 characters of saved text per person, not their entire conversation history. People without text are kept in a separate list.
Describe whom to keep
Open AI filtering on the completed collection. Describe the topics and intentions you need, and whom to exclude, in plain language.

- Open the AI filter beside your collected audience.
For example: “Keep people discussing home renovation or looking for a contractor. Exclude renovation-company ads and building-material sales.”
Check the criteria, then process the audience
Review a small sample first. If AI keeps the wrong people, refine your criteria before processing the entire audience.

- Enter your criteria.
- Click Test (10 users). A result of 1 means keep; 0 means exclude.
- When the decisions look right, click Run on the entire database.
Download the relevant audience
Download AI-approved users and use the list for your campaign. The no-message list is available separately: AI does not infer interests when there is no collected text.

- Download the selected audience. This example keeps 7,820 of 18,640 people with message text.
Screenshots show real public sources; processing totals and results are illustrative.