How Do You Analyze Recruiting Files and Data With ChatGPT, Claude and Grok?

What can each assistant do with your files?

Details as published in September 2026. Behind the scenes, all three write and run small programs to do the math, which is why their numbers are far more reliable on a spreadsheet than when you ask them to add figures in a normal chat.

Which recruiting analyses are worth trying?

These are the same numbers Module 5, Lesson 3 asks you to track. The assistant just makes them quicker to produce.

How do you ask for an analysis?

Describe the file, the question and the output. This template works in all three:

I have attached an export from our ATS. Each row is a candidate with columns [job, stage, source, applied date, stage date]. Before you analyze anything, tell me what you see in the file: number of rows, date range and any columns with gaps. Then calculate [the conversion rate from application to interview and from interview to offer, per job]. Show the method in a short table I can check by hand, then give me a chart and three plain observations. Do not guess where data is missing; list the gaps instead.

Asking the assistant to describe the file first catches most problems early, such as a date column read as text or a filter you forgot to remove before exporting.

Can you use it on resumes?

Yes, with care. Comparing an anonymized resume with a list of must haves and quoting the evidence is a good use. Asking the assistant to score, rank or pick candidates is not: it copies the bias problems covered in Module 4, Lesson 1 and may bring you under the hiring laws in Module 4, Lesson 3.

Before you upload a resume, take out the name, contact details, photo, date of birth and anything that hints at age, family, health or background. The screening prompts in the prompt library are written to organize evidence, not to judge.

How do you check the results?

Spot check three rows by hand. Ask the assistant to list the exact formula or filter it used. Compare one headline figure with your ATS’s own report. If a number surprises you, ask “which rows produced this number?” before you share it with anyone.

The most common errors are quiet ones: duplicates counted twice, a stage name spelled two ways, or blank dates treated as zero. Asking for a “data quality” section at the end of every analysis catches most of them.

What about privacy?

An ATS export is personal data even without names, because job, dates and location can identify people. Use a business plan where training is off by default, or strip the file down to the columns you need. Delete the chat when you are done if your employer’s policy asks you to. Module 4, Lesson 5 has the full checklist.

Quick answers

Can ChatGPT, Claude or Grok read a PDF resume?

Yes. All three read PDFs and Word files. Remove personal details first unless you are on an approved business plan.

How big a spreadsheet can I upload?

Big enough for most recruiting exports. OpenAI lists roughly 50MB for spreadsheets and Grok 150MB per file; very large files are better filtered down before upload.

Can it build a chart for my client report?

Yes. Ask for the chart and for the finished file, such as a PowerPoint slide or an Excel sheet, and check the numbers before you send it.

Is the math reliable?

On uploaded files, mostly yes, because the assistant runs code to calculate. Always spot check, and never trust figures it typed out without a file.