Two minutes into *Love Erosion: Death of a Mermaid*, the camera sits on the glass wall of an aquarium. A mother holds up her phone; the child beside her looks up; the subtitle reads, "The mermaid is out." A few dozen seconds later the scene cuts to a corridor, three people standing and talking, and the mood has already turned. The four-episode omnibus has 860,000 likes on Douyin. But the part worth reading is not the drama. It is the comment section.

 Six of the first eleven comments are closing reports

Six of the top eleven comments are complete reconstructions of the case: who struck first, who finished the job, who dismembered the body, where the legs went, why the shark died.

One user, "头给你打歪" (roughly: "I'll knock your head crooked"), wrote about 200 characters linking four suspects' motives, timelines and physical evidence into a single chain — one man's obsession, the photos he took, a boyfriend who had already discovered the affair and was dosing her food with mercury, a shark that died because it ate the lower half of the body. Another user, "简单要早睡" ("simple, go to bed early"), went further and ranked the killers' sequence of actions, then pointed out a hole: the culprit staged the photos as an alibi, "but the timing clearly doesn't work."

This is not cheerleading. Cheerleading sounds like "so good, please update." This is homework. They treat the drama as a puzzle, check it line by line against the timeline, and argue with each other.

And they are genuinely waiting. The third-most-liked comment, "Squatting for the next episode, kick me when it drops," has 9,646 likes. The creator replied, "I'll kick you" — 5,003 likes.

WATCH NOW

The sarcastic comment got 2,489 likes

Read only the above and you would draw a cheerful conclusion: audiences love deduction. But the single most-liked comment under this video is not a reconstruction. It is sarcasm.

A user named "默言." wrote: "I've watched so many dismemberment cases that I can solve them now. I have reached seven conclusions: 1. The victim was either male or female. 2. It was either someone they knew or a stranger. 3. The killer was either local or from out of town. 4. The victim was alive before they died…" Three more follow, running the "universal conclusions" all the way to height and cause of death.

That is not praise for tight writing. It is a recital of everything a formula can explain, and it earned 2,489 likes — which tells you the audience knows exactly when it is being handled carelessly.

So two things coexist in the same comment section: people solving the case in earnest, and people mocking the template with the template. They are not in conflict; they are two uses of one skill. Viewers already own a full suspense grammar — crime of passion, disguise, twins, amnesia, a lying witness. If the drama supplies detail, they pour that grammar into it. If it doesn't, they turn the grammar into a joke. The difference is not in the audience. It is in whether the script hands over clues or hot air.

You can cut the budget. You cannot cut the answer

According to a November 10 report in *Financial News* (金融时报), DataEye expects China's micro-drama market to exceed 121 billion yuan in 2026, of which AI short dramas and animated ("manju") dramas account for roughly 40 billion. In the first half of 2026 alone, Douyin added about 221,900 AI short dramas and animated dramas; only 1,055 of them passed 100 million views — under 0.5 percent. The same report quotes a Guosen Securities analyst saying some AI short dramas can be produced for 30,000 to 80,000 yuan per title.

At that price, everything that can be cut will be cut: extras, sets, lighting, post-production, voice work. All of it costs money. The one thing that cannot be cut is whether the clues add up. It costs almost nothing, but it requires the writer to know, while writing episode one, who dies in episode four — and requires the storyboard to remember, while shooting episode 19, where the mirror was in episode two.

That is also why viewers split the case into four stages and assign responsibility: they are not savouring the twist, they are auditing whether the twist has debts.

"No Heartburn" — the phrase that means "don't con me"

In November, a line escaped from an AI short drama: "贵是贵点，但吃了不烧心" — "it's a bit pricey, but eating it doesn't burn your stomach." According to *Wenhui Daily*, the #不烧心 (bù shāo xīn) hashtag has passed 5.5 billion views on Douyin, migrating from bun shops to rental listings, takeout reviews and workplace complaints.

What viewers borrow the phrase to say is: don't fool me. For suspense, "not burning the stomach" is more specific. Be convoluted if you like — just don't let me deduce for an hour and find out I was deducing air. The twist can be wild, but someone has to be accountable for it: every detail given in episode one needs someone to answer for it in episode four.

 The people still keeping score haven't left yet

I don't buy the claim that "audiences just want dumbed-down content." Those 200-character reconstructions under *Death of a Mermaid* are viewers handing in homework and grading the writers at the same time. The 2,489-like joke is also a grade — a failing one.

Both are still forms of caring. The real danger is not being mocked; it is when nobody bothers to write a second version — when people swipe away without even complaining. The bar for suspense short drama was never visual effects. It is that timeline chart: only if you dare to let the audience deduce will the audience stay to deduce.