Two Candidates Have the Same AI Resume Score. Who Should You Interview?

Two applicants both received 9/10 from your AI resume reviewer. You only planned to invite one more person to interview. Do you choose the person listed first, the one with more years of experience, or ask the AI to break the tie?

Do not break the tie from the score alone. Compare their evidence against the job's existing requirements. If neither has a clear, job-related advantage, collect the same small piece of missing information from both. Do not treat the order on the screen as a reason to close an application.

This guide is about that tie—not the whole hiring process. It includes a comparison sheet you can use without buying a tool.

The same total can hide different evidence

In our five-resume shortlist check, Candidate A and Candidate C both received 9/10. A had worked in online-shop support. C had worked at a hotel front desk.

These are the same fictional candidates and the same recorded run from 3 September 2026, not a new experiment. The scores are actual outputs from that demonstration. The comparison and proposed follow-up below are our judgments; no follow-up interview took place.

The fictional vacancy involved answering order questions, following problems through, keeping case notes and applying refund rules. The employer explicitly accepted transferable experience and would provide software training.

Here is what the source resumes actually gave us:

A's industry experience is closer. C's tasks still match much of the work. Neither resume shows a sample customer email or proves how well the person handled those situations.

We would not use the exact previous job title to settle this tie. The employer had already said that hotel experience could count. Nor would we turn Freshdesk knowledge into a new requirement after promising software training.

Separate a useful difference from an easy shortcut

Read the written reviews and open both originals. Then ask: Which difference matters to a requirement we set before seeing these applicants?

A difference may matter if one applicant provides clearer evidence of a task the hire must perform independently from the start. It needs a source line and a connection to the actual job. A longer resume, a more senior title or a familiar employer is not that connection by itself.

For A and C, we would want to know how each explains an unresolved problem to a customer. Both resumes give enough relevant work to justify asking; neither settles the quality of the written explanation.

If you discover that your criteria were incomplete, fix them openly. Apply the corrected criteria to everyone affected, including candidates below the current cutoff. Do not add a rule solely to distinguish these two.

The UK advisory service Acas recommends assessing applications against the job description and person specification using a consistent process. Our comparison sheet applies that principle to an AI-score tie; it is not a validated hiring assessment.

Copy this side-by-side tie check

Put source evidence beside each other before debating who seems stronger. Repeat the requirement block only for the few requirements that matter to this decision.

TIED-SCORE CHECK

Role and criteria version:
First candidate/file:
Second candidate/file:

Required task:
First — exact resume evidence:
Second — exact resume evidence:
Does either have a relevant advantage?
Why does it matter to this job?

What neither resume settles:
Same core follow-up for both:
What a sufficient answer must show:

Evidence after the follow-up:
Human next step and reason:
Reviewer and reply date:

Do not add decimal places to make the sheet look more precise. “Better evidence of handling refund exceptions” tells you something. “9.14 beats 9.08” does not explain a decision unless those numbers come from a defensible assessment you understand.

Ask one question that could change the decision

For this customer-support example, we would ask both candidates:

Describe a customer problem you personally followed through. What information did you check, what did you tell the customer, and what did you leave for the next person if it was still open?

Before hearing either answer, write down what you are listening for:

• Their own actions, distinguished from what colleagues did.
• A check of the relevant facts before making a promise.
• An understandable customer update.
• A resolution or a clear next owner and follow-up.

Ask the same core question and use the same standards. Clarifying a detail is fine; coaching one person toward the answer and not the other is not a useful comparison. If two people are reviewing, write separate evidence notes before discussing the candidates.

This follows the approach in the US Office of Personnel Management's structured-interview guidance: predetermined job-related questions and common rating standards. You do not need a long panel interview to apply the basic idea.

If written replies are the real uncertainty, use the same short fictional customer message for both instead. Tell them what you will assess, provide the same facts and make appropriate adjustments so they can participate. Do not ask them to handle a live customer or produce free work for your business.

What if you only have one interview slot?

First decide whether you truly need a full interview to answer the open question. Two short preliminary calls may be more useful than one long interview chosen from an unexplained score tie. Tell both people that this is a preliminary stage and when they will hear back.

If one person's evidence is stronger against the established requirements, record that reason and choose the next step yourself. If both remain credible and indistinguishable on the information you have, say so. Consider making room for both rather than inventing a new filter to force a winner.

Do not settle the question by application order unless your process genuinely uses an appropriate, clearly communicated rule. Do not search personal social profiles for a reason to prefer someone. The unresolved problem is missing job evidence, not a shortage of personal information.

Do not keep rerunning until someone wins

A fresh AI response that breaks the tie is not new evidence about either applicant. If an input or extraction error needs fixing, rerun the affected group under the same corrected setup and keep a record. That is different from asking repeatedly until a favorite gets the higher number.

Likewise, the two top scores are not automatically the only candidates worth checking. In the original demonstration, another candidate received 7/10 while the written review recommended an interview. The general shortlist-check guide explains that cutoff problem.

Resume Ranker can organize the reviews and original PDFs for this comparison. The number helps you find a place to start. The source evidence and the follow-up answer should explain why you invite someone.