The Internship Screen Stopped Grading Your Answer

Sponsored by

How 2M+ Professionals Stay Ahead on AI

AI is moving fast and most people are falling behind. 

The Rundown AI keeps you ahead of the curve. 

It's a free AI newsletter that keeps you up-to-date on the latest AI news, and teaches you how to apply it in just 5 minutes a day.

Plus, complete the quiz after signing up and they’ll recommend the best AI tools, guides, and courses — tailored to your needs.

Sign up to start learning.

Welcome to today's SCALIS EarlyCareers newsletter! 🚀

You submit the take-home. It is clean. The code runs, the deck is tight, the writing sample is polished. Two weeks later you get a callback and the recruiter says: "We would love to walk through it live with you. Can you share your screen?"

That is not a formality. That is the actual screen.

Something shifted in intern recruiting over the last twelve months, and almost nobody told students about it. Employers stopped treating the submitted artifact as the signal. The Wall Street Journal reported this month on companies rebuilding their interview process around verification: candidates asked to remove virtual backgrounds, pan their cameras around the room, share screens during live written exercises, even wave a hand in front of their face. One remote-first startup, Passion.io, moved away from take-home assignments entirely after finding candidates were submitting heavily AI-generated work. Another company, Nominal, now runs final rounds in person with engineers coding on company laptops.

Two forces collided to make this happen. First, skills-based hiring became the default: 70% of employers in NACE's Job Outlook 2026 survey report using it, up from 65% the prior year, and 71% of those use it at least half the time. Skills-based hiring means they want to watch you use the skill, not read your claim about it. Second, the fraud panic. NACE has flagged a rise in employers reporting fake candidates attempting to access company systems, and a Checkr survey of 3,000 managers found 59% suspected candidates of using AI to misrepresent themselves. Treat that one directionally.

Here is why this lands hardest on you. A mid-career candidate has eight years of employment history to vouch for them. You have a transcript and a project. When the artifact stops being trusted, you are the only remaining evidence. The interview is now the verification layer, and the students who get offers this cycle will be the ones who can think out loud on demand.

Learn AI in 5 minutes a day

You don't have to scroll every AI thread, track every new tool, or watch every demo. 

The Rundown AI breaks it all down for you — the latest AI news, tools, and tutorials in one free 5-minute email every morning. 

Trusted by 2M+ professionals at Apple, Google, and NASA.

Sign up to start learning.

4,304 BRAND NEW internship positions freshly posted today. View the full list below and start applying!

Assume every take-home becomes a live screen share

The take-home is not going away, but its weight is collapsing. It is a filter to earn the real round, not the round itself. So the work you submit needs to be work you can rebuild, defend, and modify in real time with someone watching.

Practical rule: never submit anything you could not recreate from a blank file in twenty minutes. If you used a library you do not understand, swap it for one you do. Something slightly less impressive that you fully own beats something polished you cannot explain.

Then rehearse it. Record yourself narrating the project for eight minutes. You will find out fast which parts you actually understand.

Practice narration, not answers

Almost every new format rewards the same behavior: audible reasoning. Live exercises, adaptive follow-ups, screen shares, in-person whiteboards. They are all built to capture the thing an AI assistant smooths over, which is the messy middle of figuring something out.

Most students practice the finished answer. Practice the process instead. Use this four-beat structure out loud:

Assumption, tradeoff, choice, what would change your mind. Four sentences. That pattern is nearly impossible to fake in real time and it is exactly what a structured rubric is scoring.

Set up a room that survives a camera pan

If you are asked to remove a virtual background or show your surroundings, the ask itself is routine now. Some candidates reportedly end interviews rather than comply, which reads as guilt even when it is just discomfort. Do not be that person.

Set up once and stop worrying. Real background, no blur, no virtual image. No second monitor in frame. Phone face down and visible on the desk, which quietly signals you are not reading from it. Wired earbuds or none at all. If you share a dorm room, tell the recruiter at the start: "I am in a shared space, my roommate may walk through, happy to show you the room if that helps." Naming it first removes the suspicion entirely.

Also: interview from the location on your application. IP checks are common enough now that a mismatch between your stated city and your connection creates a problem you never get to explain.

Learn AI in 5 minutes a day

You don't have to scroll every AI thread, track every new tool, or watch every demo. 

The Rundown AI breaks it all down for you — the latest AI news, tools, and tutorials in one free 5-minute email every morning. 

Trusted by 2M+ professionals at Apple, Google, and NASA.

Sign up to start learning.

Own the part of your project that broke

The verification question is almost always the same shape: "Walk me through a decision you made here." They are not testing the decision. They are testing whether a decision was ever actually made, or whether the work simply appeared.

So lead with the friction. "I tried it this way first and the join kept duplicating rows, so I checked the key and found the source table had dupes." "My first survey draft got twelve responses, so I rewrote the intro and reposted it and got sixty." Failure, diagnosis, fix. Nothing signals authorship faster than remembering what went wrong, because generated work has no scars.

If your project genuinely went smoothly, say what you almost did instead and why you rejected it. Roads not taken are still evidence of a driver.

Ask what the AI policy is, then use it openly

Employers are not banning AI. Recruiters report that many companies actively encourage AI use on take-homes and coding exercises, while still expecting you to demonstrate your own reasoning elsewhere. The problem is that students guess at the rule instead of asking.

Ask, in writing, before you start:

That question removes the risk of an integrity flag, signals that you think about process, and often gets you the honest answer, which is that they care far more about the follow-up conversation than the artifact.

What to say when you blank

You will get a question you cannot answer. The old recovery move was to bluff something plausible. In a format built to detect exactly that, bluffing is the worst play available.

The recovery script:

That answer scores. It shows a diagnostic instinct, awareness of your own limits, and willingness to ask for help, which is close to a complete description of what a good intern does. Interns are hired on trajectory, not a finished skill set. NACE puts the intern-to-full-time conversion rate at 63.1%, the highest in five years. Companies are not paying an average of $23.35 an hour to hire someone who already knows everything. They are paying to find out how you learn.

The algorithm made everyone's output look the same. The only thing left to differentiate on is the part you were probably trying to hide, which is how you actually think when you are not sure yet.