I Ran 5 Bilibili Automation Accounts — 3 Survived, 2 Got Shadowbanned. The Difference Wasn’t Volume.

Six views. Day 12. That’s what account #3 pulled — down from 2,400 the day before. No warning, no notification, just a silent algorithmic kill switch: a shadowban. Same automation tool, same niche, same content batch. Yet three of my five Bilibili accounts kept growing. The two that got torched? They weren’t killed by volume. They were killed by timing.

This isn’t another “automation is risky” sermon. It’s the exact parameter set that separated a hidden account from a growing one on a platform that shadow‑restricts far more than it suspends.

The mistake every Bilibili automation beginner makes (I made it three times)

Like most operators, I assumed the danger zone was raw numbers. “Keep actions under 100 a day and you’re fine.” Bilibili disagreed — harshly.

My initial config on a popular desktop tool looked like this:

• Follow 30–40 accounts per day per profile
• Like every video I commented on
• Comment lead‑gen phrase on 1 in 3 videos visited
• Run the script from 08:00 to 23:00, no off‑hours

That’s a bot‑burst pattern, and Bilibili’s anti‑manipulation layer spotted it instantly. Accounts #1 and #3 were ghosted within the first week — reach cratered, comments invisible to anyone but me. Meanwhile, operators on Kuaishou and TikTok were reporting identical patterns with one common fix: rethink pacing, not just capping.

The 14‑hour inactivity window is your cheapest shield

The single change that revived account #2 and kept #4 and #5 alive: enforcing a 14‑hour no‑action window daily — 23:00 to 13:00 the next day. Not a rest day, a daily dead zone longer than most people sleep.

Why it matters: Bilibili’s mid‑morning moderation sweeps are aggressive. A burst of likes or follows at 09:00 every day looks like a cron job. By pulling automation back to a narrow afternoon‑evening slot, I mimicked a student who only opens the app after class.

The exact alteration that worked:

• Auto‑actions permitted only 13:00–22:00
• No more than 2 consecutive days of full activity before a mandatory off day
• Every interaction interval randomised between 45 seconds and 4 minutes (no fixed 3‑minute timers)

This lines up with what I later learned about Xiaohongshu’s 14‑hour “sleep” rule — the principle is nearly identical. Platforms aren’t hunting raw numbers; they’re hunting clockwork regularity.

Bilibili doesn’t punish volume — it punishes predictability. 30 actions in 5 minutes will get you flagged faster than 200 scattered unpredictably across a week.

Fingerprint isolation: the hidden kill switch for matrix operators

Accounts #1 and #3 shared more than a niche — they ran inside the same browser profile. Same cookies, WebGL hash, Canvas fingerprint. Even with separate logins, Bilibili’s client‑side tracking tied them together instantly. Two “different” accounts pinging from identical device fingerprints within minutes? That’s a flag, not two humans.

When I rebuilt the surviving three, I switched to a matrix tool — NoobClaw in this case — that spins per‑account fingerprint‑isolated browser profiles. Each Bilibili profile opened in its own container with independent GPU, audio, font, and WebGL fingerprints. The tool also enforces per‑account daily caps and randomised pauses out of the box, so I couldn’t override the safety ceiling even if I wanted to (and I absolutely would have, given my earlier over‑enthusiasm).

For any operator running more than two Bilibili accounts, profile isolation isn’t a luxury — it’s the difference between a matrix and a ban cascade. If your current automation tool opens accounts in the same browser instance or uses a single user‑data directory, assume Bilibili can connect them.

Niche targeting: the engagement multiplier that also keeps you safer

Most people automate Bilibili by carpet‑bombing popular videos in hot categories. I tried that for three days on account #2 pre‑reset. Result: a few followers, and an engagement footprint that looked off because the account posted niche anime reviews while the comments landed under trending gaming clips.

After the reset, I fed hyper‑specific keywords: not just “动漫推荐” (anime recommendation), but the actual seasonal titles my account was already covering. The tool scraped recent videos in that narrow keyword cluster, liked only those, and left comments that genuinely related to the video content. Because the AI comments were seeded with a lead‑gen phrase at a 15% probability — not every comment — the activity profile resembled a real fan, not a growth bot.

The numbers after 30 days (3 surviving accounts):

Note the follower‑gain gap. The lower caps didn’t slow growth — they enabled it by keeping accounts visible. When I later temporarily raised daily follows to 6 on account #5 for a week, reach dipped for two days before I pulled back. The safe ceiling for Bilibili appears to be ≤5 total interactions per day (follows + likes + comments combined), with comments in particular needing to stay below 3 per day.

Safe vs. risky Bilibili automation settings — at a glance