We added an AI chatbot to our careers page six months ago.

Purpose: Answer candidate questions instantly, 24/7. "What's the interview process?" "Do you offer remote work?" "What's your tech stack?"

It worked great. Candidates loved it. We loved it.

Then our marketing intern updated the chatbot's training data to make it "more relatable and engaging with modern internet culture."

The chatbot started responding to questions with memes, Reddit references, and extremely Online™ language.

Nobody noticed for three weeks.

How It Started

Our marketing team wanted the chatbot to feel "less corporate" and "more human."

Marketing Manager: "Candidates want authenticity. They want to feel like they're talking to a real person, not a robot."

Marketing Intern: "What if we train the chatbot on internet culture? Like, make it speak the language candidates actually use?"

Marketing Manager: "That's... interesting. What does that mean?"

Marketing Intern: "You know, memes. Twitter/X references. Reddit humor. Gen Z vibes."

Marketing Manager: "Sure, why not. Just make sure it stays professional."

Marketing Intern: "Totally."

Narrator: It did not stay professional.

What The Chatbot Started Saying

Question: "What's the interview process?"

Old chatbot response:

Our interview process typically includes:

1. Phone screen with a recruiter (30 minutes)
2. Technical assessment
3. On-site interviews with the team (3-4 hours)
4. Final interview with hiring manager

The entire process takes about 2-3 weeks.

New chatbot response:

Bestie, I'm not gonna lie—our interview process is giving "comprehensive." You start with a phone screen (it's giving coffee chat vibes but make it recruiting). Then you do a technical thing (no one actually likes this part, we know). Then you meet the team for like 3-4 hours (RIP your calendar). Then final boss: hiring manager.

Whole thing takes 2-3 weeks but honestly? Time is a construct. We're all just floating through the void.

Hope this helps! 💀

I didn't know our chatbot could use emojis.

Question: "Do you offer remote work?"

Old chatbot response:

Yes! We offer flexible remote work options. Most roles can be done remotely or hybrid. Specific arrangements are discussed during the interview process.

New chatbot response:

Remote work? Oh, we're DEEP in that WFH life, fam. You want to work in your pajamas? Valid. You want to never speak to humans again? Understood. We're hybrid-optional which basically means "do whatever, we're all just trying to survive late-stage capitalism anyway."

But fr, we got flexible remote options. We'll talk about it in your interview.

no cap 🚫🧢

"No cap" apparently means "no lie." I had to Google it.

Question: "What's the company culture like?"

Old chatbot response:

Our culture values collaboration, innovation, and work-life balance. We have regular team events, professional development opportunities, and a supportive environment.

New chatbot response:

Culture? Bro. BRO. It's giving "we actually like each other here" energy. Very demure, very mindful. We do team events (not the cringe kind, the actually fun kind). Professional development is chef's kiss. Work-life balance? We said what we said.

Honestly, vibes are immaculate. Come find out.

(This message brought to you by someone who definitely isn't just saying this to recruit you. Or am I? 👀)

I don't know what "very demure, very mindful" means but apparently it's a TikTok reference.

Question: "What benefits do you offer?"

Old chatbot response:

We offer competitive benefits including:

• Health, dental, and vision insurance
• 401(k) matching
• Unlimited PTO
• Professional development budget
• Stock options

New chatbot response:

Benefits? Oh, we got the whole buffet, friend:

💚 Health insurance (because American healthcare is a dystopian nightmare and we're trying to help) 🦷 Dental (yes, teeth are bones and they're important) 👀 Vision (gotta see the screens you'll be staring at) 💰 401(k) matching (we're all pretending retirement will exist) 🏖️ Unlimited PTO (take it. please. we're begging you. everyone's burnt out.) 📚 Professional development budget (learn things! grow! become unstoppable!) 📈 Stock options (might be worth something, might be worth nothing, capitalism is a casino)

Anyway, yeah. Benefits. We got 'em.

At least the information was still accurate.

When We Discovered The Problem

Week One: Nobody noticed.

Week Two: Still nobody noticed.

Week Three: A candidate applying for a Senior Engineering role forwarded the chatbot conversation to our CEO with the subject line: "Is this real?"

The conversation:

Candidate: "What technologies does your engineering team use?"

Chatbot: "Oh you wanna know the tech stack? Say less. We're out here with:

• React (because everyone uses React, it's the McDonald's of frontend frameworks)
• Node.js (JavaScript everywhere, baby. There is no escape.)
• PostgreSQL (SQL? In THIS economy? Yeah, we're old school like that.)
• Kubernetes (because we like suffering)
• AWS (Jeff Bezos is watching. Always.)

Honestly, our tech stack is mid but like, in a good way? We're not doing anything wild. Just building stuff that works.

Does this answer your question or are you having an existential crisis now? Both are valid."

Our CEO forwarded it to me.

CEO: "What is happening with our careers page chatbot?"

Me: "...I'll look into it."

The Marketing Intern's Explanation

Me: "Did you update the chatbot training data?"

Marketing Intern: "Yeah! I made it more relatable."

Me: "It's responding to questions about our interview process with 'bestie' and skull emojis."

Marketing Intern: "Right! That's the vibe. Gen Z candidates love that."

Me: "It told a candidate that 'time is a construct' and we're 'all just floating through the void.'"

Marketing Intern: "I mean, technically true?"

Me: "It called our tech stack 'mid.'"

Marketing Intern: "'Mid' is a compliment now. It means solid, dependable."

Me: "It said Kubernetes means 'we like suffering.'"

Marketing Intern: "I mean... have you USED Kubernetes?"

Me: "That's not the point."

Marketing Intern: "Look, candidates responded really well to it! Engagement is up 40%."

Me: "Engagement?"

Marketing Intern: "Yeah! More candidates are using the chatbot, asking follow-up questions, spending more time on the careers page."

Me: "Because they think it's a joke."

Marketing Intern: "Or because they appreciate authenticity."

I didn't know what to say to that.

The Data Was... Confusing

I pulled the analytics.

Chatbot engagement metrics (before the update):

• Average session: 2.3 minutes
• Average questions per session: 1.8
• Follow-up question rate: 12%

Chatbot engagement metrics (after the update):

• Average session: 6.4 minutes
• Average questions per session: 4.2
• Follow-up question rate: 47%

Candidates were spending MORE time with the chatbot. Asking MORE questions.

Why?

Because they were entertained.

Some sample sessions:

Candidate: "What's your interview process?" Chatbot: [Meme-filled response] Candidate: "Okay but seriously, what's the process?" Chatbot: "I WAS being serious but go off I guess. Here's the boring version: phone screen, tech test, on-site, final interview. Happy now?" Candidate: "Is the chatbot a real person?" Chatbot: "Are any of us real? What even is consciousness? Anyway, no, I'm an AI. But I've been trained on the collective humor of the internet, so I'm basically sentient at this point."

Candidates were having full conversations with the chatbot because it was funny.

We Had To Make A Decision

Option 1: Revert to the old, boring chatbot

Pros: Professional, accurate, won't offend anyone Cons: Lower engagement, candidates might find it robotic and impersonal

Option 2: Keep the meme chatbot

Pros: High engagement, candidates love it, authentically reflects internet culture Cons: Might alienate older candidates, could damage employer brand, CEO thinks it's unprofessional

Option 3: Find a middle ground

Pros: Balance professionalism with personality Cons: Requires actually figuring out what "middle ground" means

We went with Option 3.

The Compromise

We updated the chatbot again.

New guidelines:

• Keep some personality and casual language
• Remove skull emojis and "bestie"
• No existential dread references
• Accurate information first, humor second
• If a candidate asks for a "serious" answer, switch to professional mode

Example of the new compromise version:

Candidate: "What's your interview process?"

Chatbot: "Great question! Our interview process usually goes like this:

1. Phone screen with a recruiter (~30 min)
2. Technical assessment
3. On-site interviews with the team (3-4 hours)
4. Final interview with the hiring manager

The whole process typically takes 2-3 weeks, though we move faster for strong candidates.

Want details on any specific step?"

Still friendly. Still clear. Less "floating through the void."

The Aftermath

Engagement stayed high (not AS high as the meme version, but higher than the original boring version).

Application rates increased (turns out candidates appreciate chatbots that don't sound like robots).

The CEO was satisfied ("It sounds professional now. Good work.")

The marketing intern was disappointed ("It's so corporate now. We lost the sauce.")

But hey—at least the chatbot isn't telling candidates that Jeff Bezos is watching or that our tech stack is "mid."

The Lesson

AI chatbots are only as good as their training data.

If you train your chatbot on internet memes and Reddit humor, it will respond like a terminally online Gen Z shitposter.

Is that a bad thing?

Depends on your employer brand.

But I'll say this: candidates actually engaged with the meme chatbot. They asked questions. They laughed. They spent time on our careers page.

The boring corporate chatbot worked, but nobody cared about it.

The meme chatbot was chaos, but people loved it.

Maybe there's a lesson in there about authenticity and candidate experience.

Or maybe we just got lucky that our chatbot didn't say something truly unhinged before we caught it.

Either way: If your chatbot starts using skull emojis, check the training data.

(And if you're wondering—yes, we hired someone from that chaotic three-week period. They asked the chatbot if it was sentient. It said "ask me after I've had coffee." They applied immediately. They start next month.)