Job search · Updated 21 September 2026

How to Prepare for a Career Fair With AI

Autumn fair season is on, and the standard advice is to print more resumes and rehearse a thirty-second pitch. Two recent studies suggest neither is the thing being scored.

Point AI at the homework, not at the speech. When researchers interviewed employer representatives after a university career fair, what they described as making a student memorable was not technical skill or a polished introduction — it was preparation specific to their organisation and a clear sense of what the student wanted. When a separate survey asked students what they actually did to get ready, updating a resume was the most common answer by a wide margin and researching companies was a distant second. A chatbot is good at the half students skip and unreliable at the half they point it at, which is writing something for them to say.

The preparation students do is not the preparation employers count

In September 2025, Briana Lee, Kenny Kaʻaiakamanu-Quibilan, Samantha Limon, Anthony Peruma and Alyssia Chen published a survey of 86 students who had just walked out of a tech career fair at the University of Hawaiʻi at Mānoa — an event with 29 organisations, surveyed on the way out so the answers were fresh. Students picked about two preparation activities each. Here is what they picked:

Nearly a third spent less than an hour in total. Not one of the 86 asked the career office anything. And the activity that dominates the list is the one that produces a document rather than a conversation.

Now the other side. In June 2026, Yılmaz Hasret published a study in Current Psychology that interviewed both groups after the same fair — 13 students and 7 employer representatives, at a large regional fair in Türkiye, in 40-to-60-minute interviews conducted within two weeks of the event. The finding that matters is named in the paper as the preparation paradox: students "described preparation in operational terms, such as researching organizations, printing résumés, and arriving with questions", while employers "evaluated preparation through a qualitative lens", associating it "not with visible effort alone, but with demonstrated career clarity, foresight, and firm-specific intentionality".

The consequence is blunt: "Students who failed to signal these qualities were rapidly categorized as unprepared, legitimizing brief or procedural engagement." The short conversation is not the cause of a bad fair. It is the result of one.

The sentence that ends the conversation

The employer interviews in that study are unusually direct about what fails. One representative, an engineer, put it this way: "A student tells me, 'I'm an engineer,' but does not even know what our company does. For me, that ends the conversation." Another described the general texture of the day as "the mindset of 'let me take a look, maybe get a pen'", which "turns the fair into a crowd without quality".

What they described as working is the specific inverse. The ideal candidate, across the interviews, was characterised "less in terms of technical expertise and more through interactional and motivational qualities, including clarity of career goals, firm-specific preparation, and proactive engagement". The same representative who complained about the pens remembered exactly one exchange: "One student asked me, 'If I want to work in your company after graduation, what would my position look like in five years?' That kind of question makes you stop and listen."

That is a low bar and a strangely reachable one. It does not require a better GPA or a better project. It requires forty minutes of reading and one question that could only be answered by someone who works there.

The other thing now being asked about

There is a second reason to prepare a real answer rather than a script this year. In its Job Outlook 2026 Spring Update, the National Association of Colleges and Employers surveyed 185 employers between 12 February and 17 March 2026 and reported that "more than one-third of entry-level jobs require AI skills", which is "nearly triple the amount that indicated this just six months earlier". AI skills appeared in 16.5% of job descriptions in the spring, up from 10.5% in the autumn, and 28% of employers said they were seeking early career talent who can use AI in their work.

Its 2026 Student Survey, covering more than 17,000 students at 258 institutions, found the opposite mood. Thirty-one percent of graduating seniors said AI skills would be of little or no importance to their career, 50.5% said they were not building any, and 46% used AI in their job search at all. NACE president Shawn VanDerziel calls it "a striking disconnect: Employers are increasingly asking graduates to be ready for AI, while a significant portion of students are asking whether AI deserves a place in their work at all."

Useful detail: NACE defines the skill in question as "asking effective questions/prompts or checking AI outputs". Not building models. That means a thirty-second story about one real task, what you asked for and how you checked the answer is a complete response — and a student who used a chatbot only to generate a pitch has no such story to tell. Prompt 4 builds one.

The part AI will get wrong

Company facts are exactly the kind of thing a language model states fluently and gets stale: reorganisations, acquisitions, product launches, which office does what, who was laid off in March. A model answering from memory will hand you a confident paragraph about a company as it existed at some unspecified point in the past, and there is nothing in the tone to tell you which parts have moved.

So use a tool with live web search switched on, and apply one rule with no exceptions: anything you plan to say out loud at the booth must be traceable to the employer's own site or a named, dated news item. Everything else is background reading you do not repeat. Being wrong about a company in front of its recruiter is worse than knowing nothing, because it is not recoverable in the ninety seconds you have. Our hallucination checklist and source evaluation checklist cover the general habit.

Split the job

The workflow, step by step

Four prompts, for three or four companies rather than all of them — the median student in the survey spoke to five. None of them writes a line for you to say.

1. The dossier that has to cite itself. Run this the night before, per company.

I am going to a career fair and [COMPANY] will be there. Use web
search for this - do not answer from memory.

Give me, in this order:
1. What the company actually does, in one plain sentence a
   non-specialist would understand.
2. What someone in a [TARGET ROLE] role would work on there day
   to day.
3. Three things that changed at the company in the last 12
   months, each with a date.

Rules:
- After every fact, put the source in brackets: either the
  company's own site, or a named publication with a date.
- If you cannot find a source for something, write NOT FOUND
  instead of the fact. Do not estimate, do not infer, and do
  not fill a gap with what is typical for the industry.
- Mark anything you find that is older than 12 months as OLD.

Read the NOT FOUND lines carefully. They are not failures — they are the things you are allowed to ask about.

2. The question generator that refuses to answer. This is the one that changes how the conversation goes.

Here is what I found out about [COMPANY]:

[paste the dossier from prompt 1]

Write 8 questions I could ask their representative at a career
fair booth.

Hard rule: reject any question whose answer is already on their
careers page, in their job postings, or in what I pasted above.
If reading could answer it, it wastes the two minutes I get.
Every question must need a human who works there.

For each question add one line: "needs a person because ___".

Do not answer any of the questions. Do not rank them or tell me
which is best. I will pick.

Pick two and memorise them. If you only do one thing on this page, do this one.

3. The clarity drill. The direct answer to what employers said they were actually reading.

Interview me until I can answer one question clearly: what kind
of work do I want, and why that kind.

Ask me ONE question at a time and wait for my answer. Start
broad and get more specific based on what I say. Push back when
I give an answer that would fit any student - "I like problem
solving" tells a recruiter nothing.

Use only what I tell you. Do not suggest careers to me, do not
write my answer for me, and do not tell me what employers want
to hear.

Stop after 10 questions and do one thing only: list back the
specifics I gave you - actual courses, projects, tasks I
enjoyed - then name the two answers that were vaguest, so I
know what I still cannot explain out loud.

4. The AI answer, from something real. Thirty seconds, built from a task you actually did.

Employers are asking early career candidates how they use AI. I
want to answer in 30 seconds from something I really did.

Here is a real task where I used an AI tool:

[describe it - what the task was, what you asked the tool for,
what you did with the output]

Turn this into a short spoken answer with exactly three parts:
the task, what I asked the tool to do, and how I checked the
output before I used it.

Rules:
- Use only what I wrote. Do not add a tool I did not name, a
  result I did not report, or a number I did not give you.
- If I did not say how I checked the output, do not invent a
  check - tell me that is the missing half and ask me for it.
- No buzzwords. It should sound like me describing an
  afternoon.

The follow-up almost nobody is confident about

In the Hawaiʻi survey, 16 students said they were not at all confident of hearing back from anyone and another 26 were only slightly confident — close to half the sample, and the authors note that many "left the event feeling uncertain about the outcome". A follow-up message within a day or two is the cheapest way to change that, and it works for the same reason the good question did: it names something specific from the conversation. Write down, on your phone, the person's name and one thing they said, before you walk to the next table. Our guide to drafting outreach emails with AI covers the message itself; the only difference here is that you are not a stranger, so the first line should be the thing they said, not your major.

Where the line is

Nothing here touches an academic integrity rule, but two honesty limits matter more at a fair than in an essay, because you are saying it out loud to someone who can check. Do not claim a skill, a tool or a project you cannot discuss unprompted for two minutes, and do not present a chatbot's summary of an industry as your own reading. Both fall over on the first follow-up question. If the recruiter turns out not to be one, our guide to spotting fake recruiters and job scams covers the warning signs.

Related reading

• Before the fair. Tailoring your resume and cover letter and getting a resume past AI screening, for the document half.
• After the fair. Behavioural interview practice with STAR, preparing for a live AI interviewer, and negotiating a first offer.
• Being findable. Writing a LinkedIn profile AI search actually finds, which is where the recruiter looks next.
• Saying it out loud. The public speaking and presentation workflow, if the talking part is the part you dread.

FAQ

Should I use AI to write my elevator pitch?

It is the lowest-value thing you can use it for. In the survey of 86 career fair attendees only 10 preparation responses mentioned preparing an elevator pitch at all, and when the employer interviews described the ideal candidate they named clarity of career goals, firm-specific preparation and proactive engagement rather than a polished opening. A generated pitch also tends to sound like every other generated pitch, and it collapses the moment the conversation goes anywhere you did not script.

What should I actually ask a recruiter?

Something only a person who works there could answer. One employer in the Current Psychology study remembered a single question from a whole fair: a student asking what their position would look like five years after joining. The filter in prompt 2 is the practical version — if the answer is on the careers page, it wastes the two minutes you get. Questions about how a team actually spends its week, what surprised them about the role, or how a recent change has landed internally all pass that test.

Will a chatbot get company facts wrong?

On recent specifics, often. Reorganisations, acquisitions, launches and layoffs are exactly the details that go stale, and a model answering from memory will not flag which parts have moved. Use a tool with live web search on, ask it to cite each fact, and only say things out loud that trace back to the employer's own site or a dated news story. Everything else stays as background.

Do employers really care whether I can use AI?

A growing share say so. NACE found more than a third of entry-level jobs now require AI skills, nearly triple the figure six months earlier, with AI mentioned in 16.5% of job descriptions and 28% of employers seeking early career talent who can use AI. Meanwhile 31% of graduating seniors said AI skills were of little or no importance and half said they were not building any. NACE defines the skill as asking effective prompts and checking outputs, so one honest story about a real task clears the bar.

Is it worth going if none of the companies are a fit?

Often yes, for information rather than applications. Just over half the students surveyed said they learned about a career path or technology they had not known about, with several naming roles they had never considered. Set the goal as finding out what work exists rather than getting an offer, and a fair with the wrong companies still pays for the hour.

Bottom line

Students prepare a document; employers read for direction and firm-specific homework. That mismatch is the whole problem, and it is fixable in an evening. Use AI for the research, make it cite every fact, and turn that research into two questions no website could answer. Let it interrogate you until you can say what you want in a sentence, and build one true thirty-second story about how you use AI, because a third of entry-level jobs now ask. Then leave the pitch unscripted and write down one thing each person said, so the follow-up has something to be about.