8 AI Accounts, 7 Failed: An Honest Review of Using AI to Get Work Done

Someone on Bilibili ran an experiment: they counted 8 accounts run by AI, and only 1 didn't fail. This observation itself is more worth mulling over than any AI tutorial — while most people are still agonizing over "can AI write good copy," the real dividing line has become "can AI consistently get the work finished."

8 Accounts, 7 Ways to Fail

According to this Bilibili creator's survey, at least 8 accounts this week were run by AI. He watched every one of them, and his conclusion was blunt: the top account posted only 5 videos, with the highest one hitting 2.17 million views; another posted 156 videos, with the highest getting only 10,000.

The contrast between quantity and results is the first layer of information. 156 versus 5 — a 31x gap in output, yielding a 217x gap in views. What does that tell us? AI lowered the barrier to publishing, but it didn't lower the barrier to being seen. Posting more doesn't mean posting right; excess capacity actually dilutes the room to polish each piece of content.

The second layer of information is hidden in the comments section. The account that succeeded turned its comments section into content by day three — real users' names were printed on a list, and over 500 people applied.

This move is key: AI handles production, humans pull real users in. Viewers shift from "watching" to "participating" — one person even used the name "Da Fei Yu" to post Dayling and got 120,000 views.

The 7 failed accounts' problem almost certainly wasn't whether AI wrote well or not — it was that AI was only "performing." Performing an emotional persona, performing a content style, but behind it there was no real work, no real feedback, no real user relationships. Viewers aren't stupid — they can smell the stench of performance.

Performing AI vs. Using AI to Get Work Done: An Overlooked Dividing Line

One line in the original article nails the essence: "They're performing AI having emotions; I'm using AI to get the work done."

This line deserves elaboration. "Performing AI having emotions" is a common pitfall in the content game — give AI a persona, have it output emotional text or videos, chasing "human-like." But "human-like" is a bottomless pit, and viewers' tolerance for AI flavor is dropping fast.

"Using AI to get the work done" is another path: AI isn't an actor, it's one link in the toolchain. One hands-on practice per day — if you screw up, you still write the next one. The essence of this strategy is embedding AI into a sustainable delivery process, rather than letting it independently bear the full weight of "creation."

The difference can be seen this way:

• Performing AI: the goal is content that looks human; the failure criterion is "being exposed as AI"
• Using AI: the goal is getting the work done; the failure criterion is "failure to deliver"

The former's ceiling is persona maintenance cost; the latter's ceiling is process efficiency. For those who want to run an account long-term, the latter is more robust.

"Persona" (人设) originates from the entertainment industry and influencer economy, referring to a public image that creators deliberately shape and maintain long-term to attract a specific audience. In AI content production, persona usually manifests as fixed tone, catchphrases, emotional reactions, and value settings, solidified through prompts or fine-tuned models. Maintaining a persona requires continuous output of consistent emotional performance — once logical breaks or emotional distortion appear, viewers get a jarring "AI flavor." More critically, the persona itself doesn't produce delivery value; it's just packaging. When packaging costs exceed the content itself, the account becomes unsustainable. By contrast, treating AI as one link in the toolchain means the persona can be simplified or even omitted, with focus shifting to verifiable output and rhythm.

Can This Playbook Be Replicated?

A few actionable lessons can be distilled from the 8-account sample.

Control capacity — don't use quantity to mask quality. The 5-versus-156 results already tell the story. AI makes posting 156 videos easy, but what really determines views is the content density in those 5. Rather than flooding, use the time AI saves on topic selection and polishing.

The comments section is the second content arena. The successful account turned its comments into content by day three, printed real users' names on a list, and got 500+ applications. That's using AI to lower operating costs, then using real interaction to boost engagement. What AI can't do is precisely "real user relationships" — which is exactly where humans should focus their investment.

Delivery rhythm matters more than a single viral hit. "One hands-on practice per day — if you screw up, you still write the next one." The value of this line lies in establishing expectations and rhythm. Viewers know there'll be more tomorrow, and creators won't stop updating just because one piece underperformed. AI's role here is bringing the marginal cost of daily updates down to an acceptable range.

Make users part of the content. Someone posted Dayling under the name "Da Fei Yu" and got 120,000 views — showing that when users are pulled in, the content they produce becomes the account's own material pool. AI organizes and amplifies; humans design the participation mechanism.

The "second content arena" (第二内容场) refers to user-driven derivative content spaces such as comments, danmaku, and fan creations, which complement the main content: the main content attracts, while the second content arena retains and spreads. Platform algorithms typically treat comment volume and engagement rate as content quality signals, and high engagement brings more recommended traffic. So turning the comments section into content — for example, displaying real user names or launching participatory activities — not only boosts community feel but also feeds back into the recommendation system. AI's advantage in this step is quickly organizing, replying to, and amplifying user output, but building real relationships still depends on human design and curation — which is why this strategy is hard for pure AI accounts to replicate.

A Signal Worth Heeding

7 out of 8 accounts failing — that ratio itself is a signal. AI content supply is expanding rapidly, but viewers' attention isn't expanding in sync. The result is that a large number of AI-run accounts fall into a cycle of "the more you post, the fewer people watch."

The truly scarce resource isn't AI capacity — it's "authenticity": real hands-on records, real failures, real user relationships. That creator said "if you screw up, you still write the next one" — this unpolished delivery stance may be the core competitiveness of that one account that didn't fail.

For content creators, rather than researching how to make AI write more like a human, research how to let AI take over repetitive labor and bet the saved energy on authenticity and continuity. After all, viewers can see through performance, but it's hard to refuse someone who keeps delivering.