Can You Trust a Short Drama's Comment Section? We Counted the Duplicates
Nearly one comment in eleven is a duplicate, 66 accounts posted ten or more, and one posted 152. Here is what a comment section can and cannot tell you.
Why this matters before you pick something to watch
None of the platforms we track publishes ratings. There is no score to sort by, no aggregate to check. So the comment section under an official upload ends up doing that job by default — you scroll, everything is glowing, and you conclude the drama is good.
We collected and analysed 11,017 comments in latin script from these platforms’ official channels. Here is what is actually in them.
Three numbers that should change how you read one
8.8% are duplicates. 1,073 of the 11,017 are repeats of text already in the corpus.
66 accounts posted ten or more comments each. The most prolific single account posted 152.
About 92% carry no identifiable intent at all — they are general praise with nothing specific in them.
None of that proves any particular comment is inauthentic. Enthusiastic people do repeat themselves, and a fan can comment on fifty videos. But it does mean that “the comments are full of praise” is close to uninformative: praise is the default state of these sections, produced in part by a small number of very active accounts.
If you were using a wall of positive comments as a quality signal, that signal is much weaker than it looks.
What people do ask, when they ask anything
The useful part of a comment section is the questions, and those are countable.
1,103 of the 11,017 comments contain a question mark — 10.0%. Sorted by phrasing, they fall into four clusters:
| What they are asking | Comments |
|---|---|
What is this called? (what is the name, does anyone know the, name of this) |
48 |
| Where can I watch it? | 20 |
| Where is the rest of it? | 11 |
| Why is it always the same plot? | 7 |
“What is this called” is the largest cluster by a factor of 2.4. When somebody actually opens their mouth in one of these comment sections, the single most common thing they want is the title of the thing they just watched.
That is worth pausing on. These are official channels, posting official clips, and the most frequent question underneath is what is this. The clips travel without their names attached. If you have ever watched thirty seconds of one of these and had no idea how to find the rest, you are the majority case among people who ask anything at all.
We built a page for exactly that problem — how to get from one remembered scene to a title.
Across the whole corpus, not just questions
Looking at all 11,017 comments rather than only the ones with question marks, five search-like intents are detectable:
| Intent | Comments | Share |
|---|---|---|
| Who is this actor? | 413 | 3.7% |
| Full version / episode count | 223 | 2.0% |
| When is the next part? | 131 | 1.2% |
| Payment and cost | 106 | 1.0% |
| What is this called? | 27 | 0.2% |
Do not add those percentages together. A comment can match more than one — eighteen do — so the categories overlap and a sum would double-count. If you need two combined, they have to be merged as a de-duplicated union, not added.
Notice that “what is this called” is 0.2% here but the biggest cluster among questions. Both are true and they measure different things: the first is a share of all comments, the second a share of deliberate questions. Most comments are not questions.
Four limits we are not going to bury
The corpus is latin script, not English. Our filter is nothing more than “contains three latin letters.” Spanish, Portuguese, Filipino and Indonesian comments are all in there. Any page that calls a set like this “English comments” is overstating its own method, and we are not going to.
The classification is regex, and nobody hand-checked it. Misspellings, abbreviations and non-English phrasings are missed entirely. Every count above is a floor — “at least this many people asked” — not a measurement of how many did.
Comments measure the urge to speak, not the urge to search. These two come apart badly. “Who is this actor” is the single largest comment intent in our data, while keyword tools show almost nobody searching for it. So this data describes what people say under a video. It cannot be turned into a claim about what people look for, and we do not use it that way.
The duplication and the high-volume accounts affect everything above. They are in these counts too.
What to do instead
- Do not read a positive comment section as a rating. Praise is the baseline, and some of it comes from accounts posting at a rate no ordinary viewer does.
- Read the questions, not the compliments. “Where is the rest of this” tells you something real about how the content is being distributed. “So good 😍” does not.
- Use the countable things instead. Episode count is published by every platform and we could verify it two independent ways. ReelShort publishes where its paywall sits. Those beat sentiment.
- If you want to know whether a series is worth your time, nobody has an answer for you yet — including us. We do not rate these, because nobody here has watched them, and we would rather say that than assemble a score out of comment counts.
Comment analysis run on 11,017 latin-script comments collected from the official YouTube channels of ten short-drama platforms, corpus dated 18 August 2026, analysis re-run and fixed as a reproducible script on 19 August 2026. We do not quote comment text or name any commenter. Classification is by regular expression and has not been checked by hand; every figure is a lower bound.