Why Vocal Went Down: AI Bots, Infrastructure Stress, and the Fix

It’s just past 1:30 AM here in NY, and I wanted to write this resource for the Vocal community—because we know the last few days haven’t been great.

Between the downtime, slowdowns, publishing issues, and moments where things like a ❤️ just didn’t work, the platform wasn’t what it should be. That’s on us—and I want to give you a clear, transparent update.

We’ve been heads down, working around the clock to fix it. Tonight, we shipped an update that we believe will mitigate these issues for the foreseeable future. So here’s what happened, what we’ve done to stabilize things, and how we’re making sure your experience on Vocal stays fast, safe, and reliable going forward.

WTF Happened?

Over the past couple of weeks, Vocal was targeted by a surge of AI bots.

These aren't your typical bots. They're fast, aggressive, and designed to mimic real users—cycling through IPs, hiding behind Google Cloud and Amazon infrastructures, and hammering the platform at scale.

Their goal? To access as much publicly visible content as possible, as quickly as possible, to train large language models.

Vocal already takes an aggressive stance against allowing AI LLMs to scrape your content for their training purposes. We block AI companies—like OpenAI and others—at multiple levels.

• We disallow them in our robots.txt file (which tells legitimate crawlers to stay out).
• We actively block known scraping hosts and data centers via Cloudflare.
• And we monitor for abnormal traffic patterns that suggest unauthorized scraping.

To give you an idea of scale, here’s the volume of AI-related bot traffic we’ve actively blocked over the past few weeks:

What We Did to Fix It

We didn’t just patch things—we rebuilt critical parts of the platform to make sure Vocal can handle this kind of pressure going forward. Here’s what we shipped:

• Scaled up infrastructure: The platform now has significantly more headroom to absorb sudden spikes in traffic without slowing down or timing out.
• New edge-layer bot protection: We deployed a new set of Cloudflare Workers built specifically to catch and filter this kind of evasive bot traffic before it ever reaches Vocal’s core systems.
• Smarter WAF and fingerprinting: We’re now detecting suspicious behavior based on patterns and signals—not just IPs—so we can identify bots pretending to be human users and block them faster.

If you’ve seen more CAPTCHA prompts than usual lately, that’s part of this. It’s not ideal, but it’s helping prevent bad actors from abusing the platform and dragging down performance for everyone else.

A Systemic Problem

What we experienced this week wasn’t isolated—it’s part of a much larger pattern unfolding across the internet.

AI scraping bots are putting enormous strain on platforms everywhere. These aren’t just passive crawlers—they’re aggressively harvesting content at scale, often without permission, to feed commercial language models. And they’re not subtle about it. They cycle IPs. They run through cloud providers. They mimic real traffic. And they hit platforms like Vocal—platforms that are open, human-first, and content-rich—because they see us as a high-value target.

We’re not alone in responding. In recent weeks, developers across the web have begun pushing back—some going as far as blocking access from entire countries or de-indexing from search engines altogether just to slow these bots down.

If you want to dig into this more, here are a few good reads:

• AI crawlers haven't learned to play nice with websites - 03/18/25 theregister.com
• Open source devs are fighting AI crawlers with cleverness and vengeance - 03/27/25 on techcrunch.com
• Trapping misbehaving bots in an AI Labyrinth - 03/19/25 on cloudflare.com

Looking Forward

Vocal is performing well again, and the team is actively monitoring everything to make sure it stays that way.

We’ll continue refining things, but the updates we’ve shipped give us more control, more stability, and better visibility going forward.

We know the last few days were frustrating. Some of you couldn’t publish when inspiration struck. Some of you hit errors at the worst possible time. You deserve better than that.

Thanks for your patience—and for continuing to trust us to get it right. If anything still feels off, or if you have questions, reach out to support. We’re here. We’re listening.

You can also subscribe to status updates at status.vocal.media.