How to Screen Virtual Assistant Applicants When Every Application Sounds Perfect

If you're working out how to screen virtual assistant applicants, you'll recognise this. The first application sounds excellent. So does the second. By the fifth, you notice they all say the same things.

"Highly organised."

"Excellent attention to detail."

"Proactive communicator."

"Experienced in Google Workspace, Canva, Asana, Trello, Slack, HubSpot and more."

None of it is false, probably. But none of it helps you decide who to talk to.

Learning how to screen virtual assistant applicants is mostly about one skill: turning claims into evidence. You're not trying to find the perfect person from a written application. That's impossible. You're trying to separate "plausible fit" from "obvious mismatch," figure out what you still don't know, and stop the best writer automatically beating the best worker.

This guide shows you a simple way to do that, with a scoring method you can use tonight.

A note before we start: here, a virtual assistant (VA) is an independent freelancer or agency you're buying a service from. Screening should assess that service, not dig into someone's private life or squeeze free work out of them.

Decide what "worth a conversation" means before you read a single word

This sounds backwards, but it's the most important step.

If you start reading applications before deciding what you're looking for, your standard quietly changes to match whoever impressed you most recently. The last confident application becomes the new benchmark.

So write it down first. Look at the job you're hiring for and ask:

• Which two or three skills actually matter for the first task?
• Which things are essential (non-negotiable)?
• Which are just nice to have?

For example, if your VA will maintain a supplier tracker, accurate source checking matters far more than design skills or how they come across on video.

Watch out for false essentials. Knowing your exact project management tool is usually a nice-to-have. Someone who's used a similar one can learn yours in a day. Don't let a long wish list knock out people who could do the real work brilliantly.

If you can't write those two or three skills down, the problem isn't the applicants. Your job description is too vague. Fix that first with our VA job description guide.

Every application has three columns: claim, evidence and unknown

Here's a simple trick that changes how you read applications.

Make a private spreadsheet. One row per applicant. Three columns for each important skill:

• Claim: what they say. "Experienced in customer support."
• Evidence: what backs it up. A sample, a specific example, a reference.
• Unknown: what you still don't know.

That third column is the secret weapon.

"Unknown" isn't the same as "bad." One applicant might write a short application that leaves lots of unknowns. Another might give evidence that directly shows they can't do what you need.

Those are completely different situations. But if you only use a gut-feel score, they both end up as "no."

Keep your notes factual. Write "no example of source checking provided" rather than "seems careless." The first tells you exactly what to ask next. The second is a judgement based on very little.

And keep this spreadsheet private. Applications contain personal information and sometimes details about other clients. Only collect what you need, and decide how long you'll keep it.

A polished sample can hide who actually did the work

A beautiful portfolio piece is impressive. But it doesn't tell you what the applicant actually did.

Did they design the whole thing? Upload text someone else wrote? Follow a process their previous client built?

All of those are legitimate experience. They're just different experience.

So when you look at a sample, look past the presentation:

• For a research sample, open a few sources and check they actually support the claims.
• For a customer reply, ask what policy and account information they had.
• For a spreadsheet or tracker, check whether unresolved items are clearly flagged or just quietly left blank.

Then ask the applicant one simple question:

"What was your part in this, and how did you check it was right?"

A strong answer names the source, the check and what happened when information was missing.

One more thing. If an applicant says they can't share previous client work because it's confidential, that can actually be a good sign. They're protecting exactly the kind of information you'll be trusting them with. Offer an alternative: a sanitised example, an explanation of their method, or a short paid task using invented data.

Use a simple 0 to 3 score, but never let the total decide

You don't need a complicated scoring system. You do need a consistent one.

For each essential skill, score each applicant like this:

• 0: no relevant evidence yet.
• 1: a believable claim, but not much behind it.
• 2: a relevant example showing a sensible method.
• 3: clear evidence of handling an awkward real-world complication.

Keep a separate flag for deal-breakers. If someone clearly can't work the hours you genuinely need, that's not a low score. It's a mismatch. Same for a serious confidentiality concern. Don't let a lovely sample average it away.

This approach borrows a principle from structured interviewing, which the US Office of Personnel Management describes as using job-related questions and consistent rating standards. You don't need their whole formal process for hiring one VA. You just need to decide what you're scoring before impressions take over.

And here's the crucial part: don't let the total make the decision.

Two applicants can reach the same total through completely different strengths. One might be excellent at the thing that matters most and weak elsewhere. The other might be average at everything. Read the scores, not just the sum.

Watch it work on three made-up applicants

Here's an invented example to show the method.

A small online shop wants a VA to prepare product catalogue updates. The job involves comparing supplier specifications, entering approved details, and flagging claims that can't be backed up. The owner still publishes everything at first.

The three skills that matter: source checking, handling exceptions, and clear handover notes.

Applicant A sends a gorgeous portfolio of product pages and says they know the shop's platform well. When asked, they explain that someone else researched the specifications, and they uploaded the approved text.

Scores: source checking 1, exceptions 1, handover 3. Total: 5.

Applicant B sends a plain spreadsheet built with invented data. Every specification has a source link. There's a column for uncertain details. There's even a note explaining why a reseller's claim wasn't treated as confirmed by the manufacturer.

Scores: 3, 3, 2. Total: 8.

Applicant C has relevant experience but can't share client work. They explain their checking process clearly and offer to do a paid test. Their available hours look like a fit.

Scores: 1, 2, 2. Total: 5.

Now look closely. A and C both scored 5. But they're totally different:

• A is strong at publishing but unproven on the most important skill.
• C sounds promising across the board but needs evidence.

So the owner:

• invites B and C to an interview,
• sends A one clarifying question about their source-checking experience before deciding.

That's not a rule that plain spreadsheets beat portfolios. If the job were mainly about publishing pages, A might be the strongest candidate. The method just keeps you focused on the job you actually need done.

Write the reason next to every score

A number on its own is easy to argue with, including with yourself next week.

So beside each score, write one short sentence explaining it.

For B's source checking: "Distinguishes manufacturer data from reseller claims and leaves unsupported fields unresolved."

For C's source checking: "Describes a sensible checking process. No shareable example yet."

Now you know exactly what to test in the next stage. And if someone else helps you review applications, you can compare evidence instead of arguing about feelings.

If you and a colleague disagree on a score, go back to the evidence. Often you'll find one of you gave points for something that wasn't on the list, like visual polish. Agree the standard, then re-check everyone affected, not just the person you were discussing.

Ask one short question before inventing a new test

When there's a gap in an application, the fix is usually a quick question, not a whole new assignment.

Ask A which facts they personally verified, and how. Ask C what a short test task would need to include to show their process. Keep it specific.

A few fairness rules:

• Ask questions with the same purpose to everyone. The facts you're checking may differ, but don't coach one person heavily while rejecting another for not guessing.
• Give a sensible reply window and mention time zones. A reply that arrives during your night doesn't mean they're slow.
• Don't judge on writing style unless writing is the job. If the role is internal data work, accuracy matters more than polished prose.

And if your "clarifying question" starts turning into actual work for your business, stop. That's a paid task now. Upwork's rules, for example, explicitly prohibit asking freelancers for free work, including work submitted just to win a job. Other platforms have their own rules too.

References are useful only when they're about the same kind of work

A reference is helpful when it talks about the kind of work you're hiring for.

With the applicant's permission, contact a former client and ask practical questions:

• What did they actually do for you?
• How did they handle missing information?
• What happened when something went wrong?

Don't contact a current client without permission. Don't ask for confidential project details. And make sure the reference is genuinely from who it claims to be, through a legitimate contact route.

A forwarded testimonial is a claim worth considering. It's not the same as a verified conversation.

Keep checks proportionate. For ordinary admin work, you don't need credit checks or a deep dive into someone's social media. More personal information doesn't mean better evidence of whether they can maintain your product catalogue.

If you're in the UK, the ICO's recruitment guidance covers contractors as well as employees. When checked in September 2026 it was marked as under review following the Data (Use and Access) Act, so treat it as context, not a final checklist.

A mismatch in the CV deserves a question, not a verdict

Sometimes you'll spot something that doesn't add up. Overlapping dates. A tool they claim to know but which doesn't appear anywhere in their samples. A reference describing a smaller role than the application suggested.

Before you decide what it means, ask.

There's often an ordinary explanation: two clients at once, a renamed project, work shared within a team, or a misunderstood question. Sometimes there isn't, and it's a genuine misrepresentation.

You can't tell the difference by guessing.

Write down the specific mismatch, ask the applicant to explain, and then decide.

If the answer doesn't satisfy you, you can decline without accusing anyone of anything. Keep it factual: "Unable to confirm the source-checking responsibility needed for this role."

And if the mismatch turns out to come from a badly worded question on your side, fix it for everyone else. Screening should expose weaknesses in your process as well as in applications.

Build your interview questions from the gaps you found

Here's where all this pays off. Your interview questions practically write themselves.

For each shortlisted applicant, look at the "unknown" column and the low scores. Those are your questions.

Keep a common core so you can compare people fairly. In the catalogue example, every candidate might discuss what they'd do when a manufacturer and a reseller list different specifications. Then add a couple of tailored follow-ups based on each person's gaps.

Don't waste the interview re-reading their application aloud. Use it to walk through a real decision: what they'd check, what they'd do, what they'd avoid, and how they'd record anything unresolved. Our VA interview questions guide gives you a full script.

Tell finalists what happens next: whether there's a paid trial, when you'll decide, and what the agreement involves. Good people are deciding whether to keep time free for you.

Close the loop properly with everyone

Once you've decided, record the decision and the reason. Keep the stages separate:

• shortlisted,
• invited to a paid trial,
• offered ongoing work.

Those are different commitments. Don't let them blur together in your tracker.

Reply to everyone you're not progressing. You don't need to give detailed feedback, but a short, polite response is the decent thing to do. If helpful, mention the job: "Another applicant showed closer experience with this particular workflow."

Then tidy up. Delete applications, samples and notes you no longer need. Remove access for anyone who no longer needs it.

In the catalogue example, the result of screening is now clear: B and C have specific interview questions, A has one clarifying question to answer, and nobody has been given access to the live shop yet. The owner knows what's been shown and what still needs testing.

That's enough for this stage. Screening doesn't promise you a perfect hire. It just makes your next conversation far more useful than reading another polished application.

Your applicant screening sheet

Copy this into a spreadsheet:

APPLICANT SCREENING SHEET

Role and first task:

Essential skill 1:
Essential skill 2:
Essential skill 3:

Deal-breakers (hours, confidentiality, etc.):

For each applicant:
  Name / source / date applied:
  Skill 1 - claim / evidence / unknown / score 0-3 / reason:
  Skill 2 - claim / evidence / unknown / score 0-3 / reason:
  Skill 3 - claim / evidence / unknown / score 0-3 / reason:
  Deal-breaker flags:
  Mismatch to clarify:
  Clarifying question sent:
  Decision: interview / clarify first / decline
  Reason:

For the bigger picture, start with our main guide on how to hire a virtual assistant. Once you've got a finalist, use our paid trial guide to test them properly.

Frequently asked questions

How many virtual assistant applicants should I interview?

Only as many as you can assess properly, often three to five. A small shortlist chosen on real evidence beats a long list of people who simply wrote polished applications.

What are the red flags in a virtual assistant application?

Generic claims with no examples, samples where the applicant can't explain their own part, and pressure to skip written terms or platform rules. A short application on its own isn't a red flag. It might just need a clarifying question.

Should I ask virtual assistant applicants to do a free test?

No. If you want new work produced, pay for it with an agreed fee and scope. Some freelance platforms prohibit unpaid test work, and good applicants often skip roles that ask for it.

How do I check a virtual assistant's references?

With the applicant's permission, contact a relevant former client through a legitimate route. Ask specific questions about the work they did, how they handled missing information, and how they communicated when things went wrong.

What if an applicant can't share previous work because of confidentiality?

That can be a good sign. Ask them to explain their method, share a sanitised example, or complete a short paid task using invented data.