"Find me 5 internships for a 17-year-old interested in graphic design, remote, this summer." Ten seconds later I had five listings. Company names, application links, even a line on why each one "fit my profile." It felt like cheating in the best way.

I applied to the first one before I'd even read it properly. The link 404'd. So did the second. The third company was real, but had never posted that internship — I checked their actual careers page. Two out of five were, politely put, invented.

Not lies exactly. ChatGPT doesn't "know" it's making things up. It's predicting what a plausible internship listing looks like, and a plausible listing is convincingly specific. That's the part nobody warns you about before you paste your first search prompt in.

---

Why it does this, and it's not a bug

A search engine looks something up. A chatbot like this one writes something that sounds correct — those are different jobs wearing the same interface. Unless it's actually searching the live web in that response, it's generating text based on patterns, not looking up today's actual open roles.

Confident, specific, and completely made up. This is the failure mode to watch for.

This is called a hallucination, and it gets worse the more specific you ask for. "What is an internship" — safe, general knowledge. "Find me 5 open design internships in Bangalore this July" — now it's inventing specifics to match a pattern, because it wants to give you a satisfying answer more than it wants to give you a correct one.

---

What it's actually good at

None of this means don't use it. It means use it for the right half of the job.

Research assistant, not search engine. That's the actual job description.

I went back and used it differently: "explain what a UX design internship actually involves day to day" — genuinely useful, accurate, saved me from applying somewhere I'd have hated. "Rewrite this cover letter paragraph to sound less like I copy-pasted it" — excellent. It's a strong editor and a weak reporter.

---

How I actually found real internships after that

The boring way, which is also the way that worked:

1. Company careers pages directly. If a company interests you, their own site's "Careers" or "Internships" page is the source of truth — not a chatbot's guess about it.
2. LinkedIn and Internshala, searched myself. Filtered by "internship," "remote," and my actual interest. Slower than asking AI. Also actually real.
3. College and community boards. A surprising number of real, low-competition internships never make it to big platforms at all.

Slower than a chatbot. Also actually true.

Then I used AI on top of that real list — not to find it, but to help me apply to it well. Once a listing is confirmed real, that's exactly when the editor version of AI becomes genuinely valuable.

---

The one part it can't do for you

Here's the part that isn't really about AI at all. Even a perfect, fully-accurate list of internships doesn't get you the internship. Someone still has to read your application and believe you specifically. A generic AI-written cover letter, sent to five companies unchanged, reads exactly like what it is.

The verification and the specificity are yours to do. AI can draft, AI can explain, AI can edit. It can't confirm a listing is real, and it can't make a hiring manager believe you actually care about their company instead of every company. Both of those are still, entirely, your job.