Telegram's own search made a keyword query 30x cheaper than crawling
Reading a channel's history to find 20 posts about one term took 25 requests and 712 KB. Asking Telegram to filter took 1 request and 24 KB. Same 20 results.
Finding the 20 posts in a Telegram job channel that mention one technology cost 25 requests and 712 KB by walking the channel's history, and 1 request and 24 KB by using Telegram's own server-side search — the same 20 results either way. Search mode is therefore not a convenience feature but the difference between paying for every post you walk past and paying only for the posts that matched. Both modes cost the same $0.001 per delivered row, so the saving is entirely in requests, bandwidth and run time.
Key points
The same 20 matching posts cost 25 requests and 712 KB through the channel history and 1 request and 24 KB through Telegram's search endpoint.
A delivered row costs $0.001 in either mode, so search does not buy cheaper rows — it removes the rows you never wanted from the run entirely.
Search runs on Telegram's side, which is why the saving scales with how selective the keyword is rather than with how big the channel is.
Telegram's search stems rather than matching substrings, so the saving comes with a false-positive problem that has to be handled after delivery.
Reading a whole history is still the right mode when you want the channel rather than a topic — search cannot tell you what a channel talks about.
Search mode looks like a convenience. It is a cost structure. The measurement below is why it is the default recommendation whenever the keyword is known in advance.
The measurement
Approach
Requests
Transferred
Results
Read the channel history
Requests25
Transferred712 KB
Results20 matching posts
Telegram server-side search
Requests1
Transferred24 KB
Resultsthe same 20
One job channel, one technology term, twenty matching posts. Twenty-five times the requests and roughly thirty times the bytes for an identical result set.
Why the gap is that wide
Reading a history is pagination: each request returns a page of posts regardless of whether any of them mention your term, and you keep going until you have walked far enough back. The work is proportional to the length of the history.
Search asks Telegram to do the matching and return only the hits. The work is proportional to the number of matches. In a channel that posts fifty times a day about many things, those two quantities are not close.
It does not make rows cheaper
A post and a search result both cost $0.001. Search does not discount the row; it removes the rows you never wanted from the run. That distinction matters when estimating: the saving shows up as a smaller row count, not as a smaller unit price.
The cost of server-side search: stemming
Telegram matches word stems rather than substrings, so a query for one word quietly includes its morphological relatives. That is a real cost of pushing the filter to the server, and it is handled after delivery rather than pretended away — what stemming does to a query.
When to read the history anyway
When the channel is the subject rather than the topic. Reach series, posting cadence, what a channel actually writes about — none of those can be phrased as a keyword, and search cannot answer a question you have not yet formed.
Frequently asked questions
▸Is search mode cheaper per row?
No. A search result and a post both cost $0.001. What search removes is the posts you never wanted: in the measured case, walking the history touched 25 requests and 712 KB to surface 20 relevant posts, where search touched 1 request and 24 KB.
▸Does the saving depend on channel size?
It depends on how selective the keyword is more than on how large the channel is. Search returns matches directly, so the work is proportional to the number of matches rather than to the length of the history walked past.
▸When should I read the whole history instead?
When the channel is the subject rather than the topic — building a reach series, measuring posting cadence, or finding out what a channel writes about in the first place. Search can only answer a question you already know how to phrase.
Sources
Every URL below was requested and returned a page on the date shown.
Documents, voice notes, audio, stickers, locations and round videos never appeared once in the anonymous preview. Fields that would always be false were cut rather than shipped.
Telegram's search stems words rather than matching substrings, so it returns morphological relatives of your term. Filtering them out afterwards is free, and it has to happen before billing.
A walkthrough of reading public Telegram channels anonymously: the four modes, when search beats reading a history, and what the anonymous preview does and does not render.
