Send live web research summaries to Telegram with GPT-4o

See What's Earning in AI Automation Freelancing.
DigiNo helps new AI automation freelancers earn faster by tracking what clients actually pay for.

Clients who write newsletters or run solo consulting practices burn hours chasing answers across tabs. This AI automation puts a live research assistant inside their Telegram, answering questions with sourced web results the moment they ask.

What This Automation Does

Receives a question from a Telegram message and routes it through a two-stage AI pipeline for intent clarity and web research
Uses a reasoning model to refine and reframe the user query before any search is performed, improving result accuracy
Executes live web searches via SerpAPI and compiles the most relevant results, including links and images, into a clean summary
Sends the finished research response back to the same Telegram conversation within seconds of the original message

Tools Used

Telegram
OpenAI
n8n

Where to Get Hired for This Skill

On Contra, top freelancers across this stack have earned 101 combined verified reviews from real client projects.

Source: Contra freelancer search · refreshed 26 May 2026

Start Earning as a Freelancer on Contra

Contra is a commission-free professional network for independents. Browse live AI automation work and keep what you earn.

Join Contra Free →

How To Build It

Wire up the Telegram bot trigger

Register a bot through BotFather, add the API token to your credentials, and configure the automation to fire every time a user sends a message to that bot.

Configure DeepSeek R1 as the intent layer

Connect the DeepSeek API key and instruct the first AI agent to receive raw user queries, clarify ambiguity, and output a clean, precise research prompt for the next stage.

Build the GPT-4o mini research agent with SerpAPI

Attach the SerpAPI tool to a GPT-4o mini agent so it can execute live web searches against the refined query and return structured results including titles, snippets, and source URLs.

Add a memory buffer to sustain conversation context

Enable a sliding window memory buffer on the research agent so it retains recent exchanges, allowing follow-up questions to stay coherent without repeating context each time.

Format and return the Telegram response

Pass the agent's output into a Telegram send-message action that delivers the formatted summary, including any relevant links or images, back to the user who originally asked the question.

Pitfalls

SerpAPI's free tier caps at 100 searches per month, so a client with even light daily usage will hit the limit inside a week unless you scope usage expectations clearly or upgrade the plan before handoff.
DeepSeek R1 and GPT-4o mini are billed separately, and both accounts need positive balances at the same time. If either runs dry mid-conversation the workflow silently fails, so set up billing alerts on both accounts as part of your delivery checklist.
Telegram bots do not support multiple simultaneous users by default without explicit routing logic. If a client expects multiple team members or community subscribers to use the same bot concurrently, you will need to build per-chat-ID routing or the responses will cross-contaminate between users.

FAQ

Can I build this without coding?

Yes. The entire workflow is built with pre-built integrations and a visual interface. The only configuration that touches anything resembling code is writing the system prompt instructions for each AI agent, which is plain English.

How long does it take?

A competent first build with all credentials ready takes two to four hours. Repeat builds for additional clients drop to under an hour once you have a tested template to clone and reconfigure.

What can I charge?

Pricing is yours to set based on your market and the client's perceived value. A common structure is a one-time setup fee for the delivered bot plus an optional monthly retainer covering API cost monitoring, prompt tuning, and uptime checks.

Which tool is required vs optional?

Telegram and SerpAPI are required since the entire delivery mechanism and web research depend on them. DeepSeek R1 handles query refinement and can be swapped for another reasoning model if a client already has a preferred provider, making it the most substitutable piece.

This is original DigiNo analysis. The underlying automation pattern is a community workflow template – view the original on n8n.