AI resume screening in India: what recruiters actually need from it

A recruiter with an open role and a full inbox has three jobs to finish before a shortlist reaches the hiring manager. They have to get through the applications without missing good people. They have to rank candidates against the role in front of them, not against a vague idea of a strong candidate. And they have to explain the order to whoever asks, whether that is a founder or a compliance lead. Screening software should be judged on how well it helps with all three. Speed alone does not settle it.

Most of the numbers here come from US and global sources, and India-specific figures are thin. Where a figure has no Indian basis, we say so.

Get through the pile without missing good people

The volume is the obvious problem. Recruiters often describe receiving hundreds of resumes for a single listing and spending days reading them. We have no survey figure for India, so we do not quote one; the complaint repeats often enough to take seriously.

The less obvious problem is what a fast filter throws away. Keyword matching misses candidates whose experience is described in other words, or whose background is non-traditional. Creative resume formats break parsers, especially ones with tables or multi-column layouts. Scanned PDFs need OCR, which lowers the quality of what the tool reads. A filter that handles only clean text will quietly drop people whose resumes look designed.

Look for parsing that reads achievements and infers seniority from the language a candidate uses. Matching job titles alone is not enough. Ask whether the tool shows what it extracted, so you can catch a misread before it affects the ranking.

Published time-saving figures vary widely, and several that circulate in vendor blogs trace back to no named study, so we do not quote them here. We found no figure broken down for India.

A quick test helps. Run a resume with a two-column layout and a scanned page through the tool, then check whether the skills and dates come out correctly. If they do not, the tool is not reading that candidate, whatever score it gives. For what HireSalt does with a file it cannot read, see what happens to unreadable resume files.

Rank against this role, not a generic best candidate

A ranking only helps if its criteria match the job in front of you. Check whether the tool applies your definition of fit or a generic one.

Weighting is the practical fix. Each skill or category gets an importance, and the score follows those weights. A customer-facing role should weight communication heavily. A data role should weight technical depth. The same resume can be strong for one job and weak for another, and the ranking should show that. Start with criteria written in the words a hiring manager would use, then adjust the weights after the first batch of results. Skill adjacency helps too. A candidate who lists Docker and container orchestration can match a role that asks for Kubernetes experience, even though the words differ.

Campus placement is the part of this piece most tied to the Indian calendar. A placement office can be handling dozens of company relationships at once, and each company wants a different balance of skills. One workaround is a separate spreadsheet for each company, and that setup breeds errors and mix-ups. The window we have been working with runs from September to October and again from December to January. Campus calendars vary, so check yours. A window of about two months leaves little room to re-tune criteria by hand for each company.

Explain the decision and keep the record

Rankings get challenged in meetings, not inside the software. A recruiter who ranks a candidate differently from their own instinct has to say why, to a hiring manager or a client. That is the job in practice: a decision that holds up on evidence rather than instinct.

A score nobody can explain tends to get argued with, not used. What helps is showing the features behind each score, such as the skills matched and how recent the experience is. Some tools also show how confident they are in a score. That practice is still emerging.

The explanation matters to candidates too. A clear reason tends to leave candidates with a better impression of a rejection, though we have no India-specific figure to quote.

Keep a record as well. Basic applicant tracking systems include an activity timeline that logs each action, email and score change for each candidate. Teams need that record to agree on decisions and to answer questions later. Compliance pressure is rising too. New York City’s Local Law 144, the Colorado AI Act and the EU AI Act each set rules for AI hiring tools. Check what applies to you with legal counsel.

On the Indian side, the law to ask about is the Digital Personal Data Protection Act. We have not verified what it requires of screening tools, so we are not making claims about it here. Check the Act’s text and ask your legal team before you assume either way.

What to ask a screening tool

HireSalt is built around this problem. Each candidate’s score comes with the reasoning behind it; how HireSalt scores a resume walks through one example. The recruiter sets the criteria and their weights. The score is the start of the decision, not a replacement for it. Plans are priced in rupees on the pricing page.

Measure your own pipeline before you compare tools. Count the applications that arrive, the ones that reach a manager and the days the shortlist takes. Those numbers are yours, and they will tell you more than a published average.

Then run one test. Ask for the reasoning behind the candidate ranked fifth, not the first. The top of any list is easy to defend, because the score is obviously high. The middle is where a recruiter gets asked questions. Then ask which criteria moved that candidate up or down, and whether you can change one weight and watch the order shift. A tool that answers those questions helps you decide. One that cannot is only sorting.