How AI Scoring Works
Last updated
HireSalt uses AI to read each resume and evaluate it against your job's scoring criteria. Scoring runs in the background, so you can keep working while resumes are processed. Every candidate gets a score from 0 to 100, and every part of that score comes with a short note you can read.
The Process
- Resume text is extracted from the PDF or DOCX file.
- The AI reads that text next to the job description and your criteria weights.
- Each of the 7 categories (33 sub-criteria) is scored from 0 to 100. Five categories carry weight by default.
- The weighted average produces the final candidate score.
- After scoring, Career Story reads the career arc. It adds context but never changes the score.
How the final score is built
The score is a weighted average of the category scores. Under the default weights, Skills Match, Experience, Education, Semantic Fit, and Stability & Trajectory count toward the total. Practicalities and Engagement Profile are scored and shown for context, with zero weight, so they only change the total if you give them weight.
If your job also has criteria the AI drafted from the job description, or custom criteria you added, each set is averaged on its own and the final score blends the sets by the split shown in the criteria editor.
Reading the reasoning
Open a candidate and expand a category to see its sub-criteria. Each one has a score and a note of one or two sentences saying what the AI found and what it missed. The note is the AI's summary, not verified evidence, so check it against the resume before you act on it. A gap the note points out often makes a good interview question.
How long it takes
In three timed runs on 16 October 2026, the first score arrived 18 to 28 seconds after scoring started, and the runs scored 252 to 392 resumes per hour. Your times will vary with load. The full method is in the article linked below.
What the AI cannot read
- Scanned resumes saved as images are not read. HireSalt does not run text recognition, so the file fails and its credit is returned.
- Tables inside Word files are not read, so skills listed only in a table are missed.
Accuracy Tips
- Longer, more detailed job descriptions give the AI more to score against.
- Adjust criteria weights before scoring. They lock when the first candidate is scored; to change them after that, reset the criteria (scores are deleted) or duplicate the job.
- To see how the ranking would move under different weights without re-scoring anyone, use the weights check on the Shortlist check tab.
See also: Criteria and Sub-Criteria Explained — Shortlist Check: Weights and Who Moves — Understanding Career Story