How to stand out as a job applicant when AI raises the bar for everyone

When everyone can produce polished output, judgment and visibility do the sorting

TL;DR: AI has raised the bar for appearing competent, making formal job applications harder to win than ever. Two things still cut through: judgment you can defend, and the surface area you've built before you needed a job.

Acknowledgements: thank you to Kevin Xia for your feedback on this post!

The baseline has moved

A well-structured report, a clear email, a researched summary, a tidy presentation. These used to take skill and time. Now they take a good prompt and five minutes.

The floor for looking competent has risen sharply. Some employers forbid AI use in test tasks, but the realistic assumption is that some applicants use it anyway. Where it’s allowed, everyone uses it. Either way, the floor rises, and the bunching at the top gets worse.

You still can’t get hired without looking competent, but it’s no longer sufficient. The question employers are asking has shifted from “can this person do the work?” to “is there a real mind behind it?”

How employers are adapting

Some employers are redesigning test tasks to surface what AI can’t easily fake.¹ The strongest version I’ve seen asks for thinking, not just output: a Loom walkthrough of your approach, a written reflection on what you’d do differently, a live discussion of the work, and the questions you’d prioritise and why. The test becomes “can you defend your reasoning?” rather than “can you produce a deliverable?”²

I have also heard about tasks that require real-world contact. Interview a stakeholder. Get a quote from a supplier. Call three orgs and synthesise what you heard. AI can’t (yet?) build trust with a human in 48 hours.³

What these have in common is that they reward agency and judgment over polish.

What signals something real

Two things are harder to replicate with AI:

Judgment you can defend. AI output is sensitive to your prompt. A vague prompt gets a vague answer. Framing a question well requires you to already understand the problem, and evaluating the answers requires knowing whether they're right in your specific context. That depth has to come from experience or real understanding.

A visible point of view. Writing publicly about what you’re working on, what you’ve changed your mind about, what you notice that others don’t. Your track record of showing how you think will be more important than the size of your audience.

The practical takeaway

AI makes it harder for genuinely strong applicants to stand out, because everyone around them can look competent too.

If you do apply through formal rounds, two things matter. First, make sure your application reflects your own judgment and thinking rather than what the best model suggests you say. Reviewers can often tell, and if they can’t, your interview or work trial will. Second, assess honestly how competitive you can be for a given role. If you don’t have the context, AI won’t fake it for you. There will very likely be candidates who use AI and have the context and have their own strategic takes.

Formal public applications are the most crowded channel, and for many roles, they’re not the highest-yield place to put your effort. The better bet is often to invest in things that compound: visible work, real relationships, and legible thinking, well before you need them. That investment opens up closed hiring rounds, warm referrals, and the chance to create your own role (see my previous post about unofficial work that gets you hired and building your surface area for serendipity).

It’s harder than polishing your CV, but I think it’s a more reliable way to get work in an increasingly competitive space.

---

Hi, I’m Sofia Balderson. I lead Hive, a global community for people working to end factory farming. I started Notes from the Margin to share the messier, more personal reflections that don’t fit in formal updates. If you care about leading, belonging, or building something that matters (especially from the edges), you might enjoy sticking around.

Got forwarded this newsletter, or a new visitor? Subscribe to get updated about my new posts.

1. These tasks are likely to change as fast as AI develops - my knowledge here is based on what I’ve recently seen some recruiters talking about and is not exhaustive.

2. It’s worth acknowledging that these tasks can introduce more biases than written output.

3. These tasks can be tricky to organise for a large number of applicants, so I expect they will likely appear in the final stages of applications.