Do Photos Get More Views on Telegram Than Text Posts?

For the typical Telegram channel, a post with a photo or a video gets almost exactly the views a text-only post gets. Across the 146 public channels in our survey that published at least three of each kind, the median channel’s media posts collected 1.03 times the views of its text posts — a 3% edge, with the middle half of channels falling between 0.92 and 1.13. In seven channels out of ten the two kinds landed within 25% of each other.

Those figures come from the same reading of 325 public channels and 5,166 posts behind our survey of Telegram view-to-subscriber ratios, taken on 9 September 2026 from the pages Telegram serves to anyone with a browser. No new channels were read for this page. What is new is a script that cuts the existing rows by post type, and it is published with the data so each figure below can be rebuilt.

The pooled number says the opposite, and it is wrong

Throw every post into one pile and text-only posts look like the clear winners. Their median post drew 0.27 views per subscriber; the median media post drew 0.13. Read naively, a photo halves your reach.

The data shows nothing of the kind. The pile mixes two populations that differ in size far more than they differ in format. Large channels attach media to four posts in five — 79.9% of posts on channels over 50,000 subscribers carried a photo, video or file, against 65.2% on channels under a thousand. Small channels supplied 45.9% of all text-only posts but only 36.4% of the media posts, and the median text-only post sat on a channel of 1,035 subscribers, while the median media post sat on one of 6,200.

Small channels already have far higher view-to-subscriber ratios than big ones; that size gradient is the strongest pattern in the survey. So the pooled comparison is mostly re-measuring which channels the posts came from, above all how big they are, with a format label on top. Splitting the pile into our three size bands does not cure it: text still leads inside each band (0.82 against 0.71 views per subscriber on small channels, 0.15 against 0.09 on mid-sized ones), because each band spans channels whose ratios differ several-fold. Only comparing posts inside one channel removes it. At the level of whole channels, the share of posts with media and the channel’s median ratio move against each other (Spearman −0.22), which is the same confound seen from a different angle. Fourteen channels in the sample posted no media at all and 86 posted nothing else; neither group can say anything about what format does, because neither group made a choice that can be compared.

The comparison that survives this asks a narrower question: within one channel, with the same audience, the same subscriber count and the same posting habits, do its media posts out-draw its text posts? That is the 1.03.

Checking that age is not doing the work

Inside a channel, post age moves a view count more than anything else we measured. A post under an hour old gained a median of 148.9% between our first and later readings that night, and posts half a day to a day old still showed only about half of what they would carry four days later. If a channel’s text posts happened to be newer than its photos, the comparison would be rigged before format entered into it.

They were not, on the whole. Inside the 146 channels, the median age of media posts was 1.03 times the median age of text posts — practically the same. And restricting both groups to posts at least 48 hours old, which leaves 108 channels with three of each, gives a median of 1.03 again, with media ahead in 58.3% of channels. The answer does not depend on whether fresh posts are counted.

Nor does it depend on channel size. The median within-channel ratio was 1.03 for channels under a thousand subscribers (58 channels), 1.04 for those between a thousand and fifty thousand (44), and 1.06 for those above that (44).

Photo, video, and everything else

The survey’s original media flag lumps every attachment together. For this page we went back to the stored preview pages and split it: a post counts as video if it carries a video or round video, as photo if it carries a photo and no video, and as other for documents, voice notes, audio, stickers, polls and locations. Of the 3,628 media posts, 2,621 were photo, 798 video and 209 other.

A photo and a video sit in the same place: a few per cent above the text-only posts of the same channel. Nothing here supports the common advice that video is the format Telegram rewards. The video row also rests on 52 channels, not 124, and its spread is wider, so the honest reading is “no measurable difference from photo”, not “video slightly behind”. The 23 channels with enough documents, polls and voice notes are too few to rank.

How much of that 3% chance alone produces

Any two sets of posts from one channel will differ, whatever they contain. To see how large a gap appears with no format effect at all, we shuffled the labels inside each of the 146 channels — keeping how many media and text posts each one had — and recomputed the ratio, two hundred times with a fixed seed.

With the labels scrambled, the typical channel’s ratio is 1.00, its middle half runs from 0.88 to 1.13, and “media” comes out ahead in 48.6% of channels. The real figures are 1.03, 0.92 to 1.13, and 59.6%. So the real spread between channels is no wider than random relabelling makes it, while the centre is nudged a little toward media. Taken channel by channel, the observed ratio fell outside the middle 90% of its own shuffles in 19 of the 146 channels: 13%, where pure chance would put about 10%, or 15 channels. Four over is well within what chance itself produces, so this test does not single out any channel where format clearly mattered. The evidence for a lean is the share of channels where media came out ahead: 59.6%, against 48.6% with the labels shuffled.

Put plainly: across many channels there is a faint, consistent lean in favour of posts with an attachment, and on any one channel it is too small to pick out from ordinary variation. For scale, the gap between a channel’s strong and weak recent posts — its 90th percentile over its 10th — has a median of 2.24×. A 3% format edge disappears inside that.

Posts forwarded into a channel did worse

Whether a post was the channel’s own or forwarded in from elsewhere has a clearer answer, though on thin numbers. Forwards made up 7.05% of all posts, spread over 87 channels, and only 37 channels had three or more of them alongside three originals. In those 37, the median channel’s forwarded-in posts drew 0.85 times the views of its own posts, and forwards came out ahead in only 27.0% of channels. In 21.6% of them, forwards drew two-thirds or less of what originals did. Restricted to posts at least 48 hours old the gap narrows to 0.90 across 21 channels, which is a small enough group that the two figures should be read as one direction, not two sizes.

Two things limit what that means. First, the flag records a post forwarded into the channel from somewhere else. It says nothing about how often a channel’s own posts were forwarded out — and outward forwards are one of the main ways a Telegram channel reaches people who have not subscribed, which is why they sit at the centre of growing a channel from a standing start. The public page cannot see them. Second, Telegram’s view-counter specification says that messages sent to and forwarded from channels carry a view counter, but not how the counter on a forwarded copy relates to the original’s, and the preview does not show which one it prints. We read the number under each post as served and do not claim more for it.

With those limits stated, the plain reading is that relaying someone else’s post is, for most channels that do it, a slightly weaker post than one of their own.

What this cannot tell you

Format was not assigned at random. Channel owners choose when to attach a photo, and they tend to do it on the posts they think matter — an announcement, a launch, a long piece. If those posts would have drawn more views anyway, part of the 3% belongs to the post, not the picture. This is an association inside channels, not the effect of adding an image, and only a controlled test could separate the two.

It is the recent posts only. The public preview at t.me/s/ lists roughly the last twenty posts, so a busy channel contributes a few hours of output and a quiet one several months. A channel needed three posts of each kind inside that window to be compared at all, which is why 146 of 325 entered.

The counters are rounded. The preview prints views to three significant figures — “27.9K” covers a hundred-view band — so on large channels a gap of a few per cent between two medians is partly rounding. It does not bias the comparison toward either format, but it makes a small difference softer than it looks.

Some “text-only” posts are not text-only. The media flag works from what the public preview renders. 287 of the 1,538 posts counted as text carry an attachment the preview cannot show, displayed only as a placeholder such as “Please open Telegram to view this post” or “This media is not supported in your browser”. Leaving those 287 out keeps the median within-channel ratio at 1.03, across 104 channels instead of 146, so the conclusion stands, but the text column is not purely text.

A view is a display, not a reading. Telegram’s client queues a view when the bottom edge of an unread message becomes visible, or the vertical centre of one already read, and the rule draws no distinction between a photo and text. Nothing in this data measures whether anyone looked at the image, read the caption or clicked a link; for what the counter does and does not count, see how Telegram post views work.

Testing it on your own channel

The survey can tell you the typical channel sees no meaningful format effect. It cannot tell you about yours, and the test is cheap enough to run.

Compare like with like on age. Pick posts that are all at least two days old — older if you post rarely — and set the median views of your photo posts against the median of your text posts. Medians, not averages: one viral post will drag an average wherever it likes. You need several posts of each kind before the comparison means anything, and more is better.

Hold your own noise floor next to the result. If your strongest and weakest recent posts differ by a factor of two, which is normal, then a 10% or 20% gap between formats across a handful of posts is not evidence of anything. Only a gap that persists across weeks of posts, at matched ages, is worth acting on.

Change one thing at a time. If you start adding images and also move your posting hour, you will not know which did what; the timing question is covered separately in our guide to when to post on Telegram, and how often to publish, from the same survey, in the analysis of posting frequency.

And spend the effort where the data shows a lever. Posts with a photo drew slightly more views than the same channels’ text posts, not fewer, so a photo that fits the post carries no visible cost. Writing a post yourself rather than relaying another channel’s showed the larger gap here, though that too is an association.

The data

The new figures are produced by post-type.py, which reads posts.csv and, for the photo and video split, the stored wave-one preview pages. It writes post-type-stats.json, never touches the network, and reproduces the 146-channel figure already in stats.json as a check on itself. How the channels were chosen and what was excluded is in method.md. Everything is released under Creative Commons Attribution 4.0.

If a number on this page does not fall out of that script, the page is wrong, and the fix will be dated on the corrections page. More on Telegram channels is in the Telegram growth guide.