Job hunting with AI, without doing it alone

01The search got louder

The flood of applications is measurable. LinkedIn told The New York Times it was handling an average of 11,000 applications a minute, up more than 45% in a year, with generative AI contributing.¹ Greenhouse, whose software many companies use to hire, says the average job on its platform drew 28 applications in 2021 and 95 in 2025.²

Most of those figures are surveys, and Greenhouse and Gartner both sell hiring products, so read them as direction rather than precision. The direction is clear, though. When everyone can produce a polished application in seconds, polish stops being a signal. What still stands out is evidence that a real person did real work and can talk about it.

02What AI is for, and what gets you cut

Employers are drawing lines, and the lines are similar. Anthropic, which makes the Claude assistant, publishes guidance for candidates. It encourages AI for preparation, asks applicants to “create your first draft yourself, then use Claude to refine it,” and says live interviews are “all you” unless the company says otherwise.⁷ Amazon’s interview guidelines, as reported, warn that using generative AI during an interview may lead to disqualification.⁸ Google’s chief executive has said it is worth bringing some interviews back in person.¹⁰ Meta has gone the other way for some coding interviews, building an AI assistant into the interview itself, while still ruling out outside tools.⁹

Hiring managers are also getting better at spotting shortcuts. In Greenhouse’s survey, those who had caught deceptive AI use most often named candidates reading AI-written scripts, followed by instructions hidden in CVs to game screening software, and deepfakes.³ Gartner found 6% of candidates admitted to some kind of interview fraud, and predicts that by 2028 one in four candidate profiles worldwide will be fake.⁴ That is a forecast, not a count, but it explains why employers are adding checks. There is a market for cheating, too: a tool built to feed candidates answers in live interviews raised $5.3 million after its founder was suspended from university over it.¹¹

The workable rule is simple: use AI to prepare, not to perform. Use it to research a company, rehearse, find gaps in your stories and tidy your own writing. Do not use it to invent experience, answer for you in an interview or blast out applications you have not read. The employers above draw the line in the same place: at work that is not yours, presented as if it were.

03Build a story bank

Most good interviews are structured: every candidate gets the same questions about what they have actually done, scored against the same criteria. Research on hiring methods consistently ranks structured interviews among the best predictors of later job performance. A 2022 re-analysis by Sackett and colleagues placed them at the top, well ahead of unstructured conversations.¹²¹³ For a candidate, that means the questions are predictable in kind: a conflict you handled, a mistake you made, a result you drove, something you learned fast.

So prepare the answers in advance, from your real work. A story bank is eight to ten true stories, each written as the situation, your task, what you did and what changed. One story can answer several questions. The research does not show that any particular format wins interviews on its own. It shows that interviewers ask about specific past behaviour, and specific, true stories are what you need to answer well.

AI is genuinely useful here, as an interviewer that helps you remember. Ask it to question you, not to write for you:

You are helping me prepare for job interviews. Interview me about my last two roles, one question at a time. Ask about projects, problems I solved, disagreements, mistakes and results. After each answer, ask one follow-up for specifics: numbers, what I personally did, and what changed. Do not invent or embellish anything. When we have eight stories, list each one as Situation, Task, Action and Result, using only what I told you, and flag any gaps.

04Drill, then debrief

Knowing your stories is not the same as telling them well under pressure. Practice helps, and practising with others seems to help most. In a study of 213 candidates for police and fire promotions, those who prepared through study groups and role-play performed better in the interview than coaching alone would explain.¹⁴ It was not a randomised trial, but it matches what later studies found. In a small German experiment, students who practised with a virtual interviewer that gave feedback on their body language interviewed better and felt less anxious than a control group.¹⁵ Anxiety matters: in videotaped mock interviews, anxious candidates were rated lower, largely because they came across as less assertive and less warm.¹⁶

AI can play the interviewer so you can practise at any hour. The debrief is where the learning happens, and a person who knows you is still the best judge of whether a story sounds like you.

Act as an interviewer for a [role] position at a [type of company]. Ask me one behavioural question at a time, drawn from these skills in the job description: [paste skills]. Wait for my full answer. Then ask one probing follow-up, the way a real interviewer would. Do not give feedback or suggest answers until I type "debrief".

Debrief my answers. For each one, tell me: did I answer the question that was asked? Was it specific about what I did and what changed? Was it too long? Give me one thing to keep and one thing to change. Do not rewrite my answers for me.

05Don't search alone

A job search takes a toll that most advice skips. A large review of studies found that unemployed people report psychological problems at about twice the rate of employed people: 34% against 16%.²² In a three-week daily study, job seekers’ mood swung from day to day, and their effort tracked how much progress they felt they were making, which is hard to see from inside a long search.²⁵ Searching while still employed carries its own strain.

The oldest evidence for doing it together comes from the Job Club, a programme built on job seekers helping each other: a buddy system, shared leads and daily support. Within two months, 90% of members had found work, compared with 55% of a matched comparison group.¹⁷ That was in the United States in 1975, and the comparison group was matched rather than randomly assigned, so the gap is likely overstated. Later trials were stricter.

The strongest evidence comes from the JOBS programme at the University of Michigan, tested in randomised trials. The first, with 928 recently unemployed adults, found participants got better-quality jobs, measured by earnings, and stayed more motivated if they were still searching.¹⁸ A larger trial found that two years later, participants had higher reemployment, higher monthly income and fewer symptoms of depression.¹⁹ A Finnish version with 1,261 people improved the quality of reemployment too.²⁰ The workshops trained people in job-search skills in a group, with practice, mutual encouragement and preparation for setbacks.

A review of 47 job search programmes found that participants had 2.67 times the odds of finding a job compared with those who did not take part. The programmes only worked when they built skills and motivation together. Social support was one of six ingredients linked to bigger effects, alongside teaching job-search skills, improving self-presentation, building confidence, encouraging proactivity and setting goals.²¹

Those programmes were run by trained facilitators, and none of the trials were in India. A group of peers is not the same thing. But you can borrow the ingredients that worked: a regular meeting, shared goals, practice, honest feedback and help getting through rejection. That is the idea behind Phyl Terry’s “job search councils”: a small group of people at a similar career stage, meeting weekly with a structured agenda. His programme’s site describes groups of four or five seekers plus a moderator.²⁴ It is a practitioner model rather than a tested one.

A group also widens your reach. A large LinkedIn experiment with more than 20 million people found that moderately weak ties, people you know but not closely, were the most useful connections for changing jobs.²³ Each member of your group brings a different set of those ties.

06If you're searching in India

Indian professionals report the same pressures. In LinkedIn research published in January 2026, 84% of Indian professionals said they felt unprepared to find a job, 76% said it had become harder in the past year, and 94% planned to use AI in their search. LinkedIn’s data showed applicants per open role in India had more than doubled since early 2022.⁵ These are self-reported figures from LinkedIn users, reported in the Indian press.

The market itself is not shrinking. Naukri’s JobSpeak index of new white-collar job postings rose 14% year on year in August 2026, with roles for people with up to three years’ experience up 15% and AI and machine learning roles up 31%.⁶ Postings are not hires, but the picture is of more competition for attention, not fewer jobs. A search group works the same way in Bengaluru or Pune as anywhere else. It can meet in person, on a video call or through a weekly thread, as long as it keeps a fixed time and a shared agenda.

07A four-week plan

One month is enough to change how your search runs. Keep applying to roles you would genuinely take while you do this.

1. Find your people and your stories

   Ask three to five people at a similar stage to search together, and agree a weekly time. Use the story finder to draft eight to ten true stories and add them to your bank.

   Output: a group and a story bank
2. Apply less, and better

   Pick fewer roles and read each description properly. Write your own first draft of each application, then use AI to tighten it. Share your target list with the group: someone may know someone there.

   Output: a short list of real applications
3. Drill out loud

   Run two mock interviews with the AI interviewer and one with a group member. Debrief each. Update the stories that fell flat.

   Output: three practised interviews
4. Review and adjust

   With the group, look at what is getting replies and what is not. Change one thing: the roles, the CV or how you reach out. Plan the next four weeks.

   Output: one change, and a plan

08Before you hit send

• Did I write the first draft myself, and is every claim in it true?
• Could this application only have come from me, for this role?
• Have I checked this employer’s rules on using AI?
• Do I have a specific story ready for each skill in the job description?
• Have I said my answers out loud to a person, not just rehearsed them in my head?
• Who else could look at this before I send it?

AI can make a search faster. It cannot make you less alone in it. Use the tools to prepare, and find the people to prepare with.