How to Screen Intern Applicants Fairly (When They All Have No Experience)

By The Internhuddle Team · Hiring · Published 22 June 2026 · 8 min read

Screening interns is a strange discipline: nearly every applicant has little to no professional experience, so the usual CV signals are missing. Programs that screen well replace "gut feel from a thin CV" with structure — and structure is also the best anti-bias tool you have at this scale.

Quick answer: screen intern applicants in two separate lanes — role fit (evidence of relevant foundations: projects, coursework, self-teaching) and communication (a short structured video introduction) — score both against fixed rubrics, weight role fit first, and let AI triage the volume while a human owns every decision.

Why intern screening goes wrong

• The polish trap. With no experience to compare, reviewers over-index on confidence, fluent writing and presentation. Those correlate with background and coaching more than with potential.
• The volume problem. Internship postings attract hundreds of applications. Under time pressure, unstructured review collapses into skimming for familiar university names.
• The single-score mistake. Blending "seems bright and likeable" with "has actually done related work" into one number lets charisma outvote evidence.

Lane one: role fit (weight this first)

Role fit asks one question: is there evidence this person has foundations relevant to this specific role? For an engineering intern that might be coursework, a personal project, or a hackathon; for a marketing intern, a student society campaign or a self-run channel. Define 3–5 criteria before opening applications, for example:

• Relevant foundation — demonstrated exposure to the domain, however junior.
• Initiative — anything built, organised or shipped without being told to.
• Learning velocity — evidence of picking new things up quickly.
• Practical fit — availability, location and timing actually match the internship.

Score each criterion on the same scale for every candidate, citing evidence from the application — "built a weather app in React (GitHub linked)" — rather than impressions. A candidate with a strong foundation and shaky polish should always outrank a polished candidate with no relevant foundation.

Lane two: communication (short video introductions)

A 60–90 second recorded introduction against fixed prompts adds a dimension CVs cannot: how the person communicates, what motivates them, and whether they can structure a thought under mild pressure. To keep it fair:

• Same prompts for everyone, shown at record time — this measures thinking, not scriptwriting.
• Score against a rubric (clarity, structure, motivation, authenticity), not general likeability.
• Keep the score separate. The video score should inform the overall picture but never override role fit. A brilliant presenter applying for a role they have zero foundation in is still a weak fit.
• Accommodate. Offer re-records and an alternative route for applicants whose devices or circumstances make video hard.

Where AI helps — and where it must not decide

AI screening earns its keep on consistency and volume: it applies the same rubric to application #7 and #400 with equal attention, extracts the evidence for each criterion, and drafts a scored summary a human can verify in seconds. Used well:

• AI triages, humans decide. Every shortlist, rejection of a borderline case, and final call is a human judgement over AI-organised evidence.
• Scores are explainable. Each score cites the evidence behind it, so any decision can be defended to the candidate — or to yourself six months later.
• You calibrate it. When reviewers disagree with a score, that feedback should tighten future screening rather than vanish.
• Silence is a zero, not a guess. Edge cases matter: a video with no discernible speech should score zero for communication, not whatever the model hallucinates.

A screening pipeline that scales

1. Publish criteria in the posting — applicants deserve to know how they will be judged.
2. Collect structured applications: role-fit questions plus a short video introduction, in one flow.
3. First-pass triage (AI or rubric-armed humans) sorts candidates into strong / possible / weak fit with cited evidence.
4. Human review of the strong and possible piles, comparing evidence, not vibes.
5. Decide and communicate quickly. Students juggle multiple applications; a two-week-faster shortlist wins candidates that a prestigious-but-slow process loses.

← All articles