ChatGPT resume tailoring: what it does well, and what it makes up.
A general-purpose model tailors a resume better than most people expect and invents things more often than anyone admits. Both halves of that are worth understanding before you paste your resume into a chat window.
This is a comparison with a tool, not a company. Rezly runs on the same kind of large language model — ours is Claude — so the difference isn't intelligence, it's constraint. An open chat window will write whatever produces a good-sounding resume. A constrained pipeline is allowed to reorder and rephrase what your resume already says, and is stopped from adding anything it doesn't.
| ChatGPT (or any open chat) | Rezly | |
|---|---|---|
| Quality of prose | Excellent. It writes fluent, confident bullets and will happily produce ten variations. | The same underlying capability, pointed at a narrower job: rewording what's already true. |
| Fabrication | The core risk. Ask for improvement and it will add plausible technologies, invented percentages and responsibilities you never held — because it's completing a pattern, not reporting facts. | Blocked at two layers: the model is instructed never to add skills, employers, titles, dates or metrics that aren't in your resume, and a server-side check drops any rewritten bullet that introduces one anyway. |
| Reviewing the changes | You get a finished resume. Spotting what quietly changed means reading both versions line by line. | A diff. Original, revision, and the reason for it, accepted or rejected one at a time. |
| Honest bad news | Ask whether you should apply and it will usually encourage you, because agreeable answers are what it's shaped toward. | Returns apply, borderline or skip — and names which required things you're missing when the answer is skip. |
| Formatting | You copy text out of a chat window and reformat it yourself, which is where layouts break. | Exports back through the structure of your uploaded file, so the formatting survives. |
| Cost | Free tier available; paid tiers cost more per month than most of our credit packs cost once. | ₹99 / $9 for 15 tailorings, bought once, no expiry. |
Where ChatGPT is the better choice
For thinking out loud, it's better than we are and it isn't close. Working out how to describe a career change, drafting the awkward paragraph in a cover letter, rehearsing answers to an interview question, deciding whether a role is a sensible next step — a conversation is the right shape for all of that, and we don't offer one.
It's also free to start and already open in your browser. If you are tailoring one resume for one job and you will read every word of the output carefully, an open chat with a careful prompt will do the job. The failure mode shows up at volume, when you stop reading closely.
Is ChatGPT good for resumes?
It is very good at the writing and genuinely unreliable at the truth, and those two facts are easy to confuse because fabricated bullets read better than real ones. Asked to improve a resume, a general-purpose model optimises for a convincing document. It has no way to check what you actually did, so it fills gaps with what is statistically likely for someone with your job title: a framework you have never used, a percentage you never measured, a team you never led. None of that is malicious — it is what completion does when the prompt asks for a strong resume and the facts run out. Used deliberately, it is a good editor: give it a bullet you wrote and ask what is vague. Used carelessly, it hands you a document you cannot defend, and the interview is designed to find exactly that.
How do you stop AI inventing things on your resume?
Constrain it, then verify it. The prompt has to state what the model may do — rephrase, reorder, re-emphasise — and what it may never do: add skills, tools, employers, titles, dates, degrees or numbers not present in the source. Requirements the resume can't support have to go somewhere visible, as flagged gaps, rather than being quietly absorbed into a bullet. But prompts are instructions, not guarantees, so the second layer matters more: a programmatic check that compares every rewritten bullet against the original text and drops any that introduce a technology or figure that wasn't there. That is how this is built, and it catches real cases — including the model helpfully writing “add your real metric here” into a bullet. If you're doing this in a chat window instead, the equivalent is reading every line against your original and deleting anything you can't evidence.
Can you get caught lying on a resume?
Routinely, and usually not in the way people fear. The dramatic version — a background check uncovering a fake degree — happens, but the common version is quieter: an interviewer asks a normal follow-up question about something on your resume and you can't answer it. Someone asks how you got the 40% improvement, or what you found hardest about the framework listed in your skills, and forty seconds later everyone in the room knows. There is no confrontation; you simply don't progress. The cost is worse than not having the skill, because a gap you acknowledge honestly is a normal hiring conversation, while a claim that collapses under one question changes how everything else on the page is read. This is the whole argument for tailoring under constraint — where the line sits between emphasis and invention is worth being precise about.
ChatGPT and OpenAI are trademarks of their respective owners. Rezly is not affiliated with, endorsed by, or sponsored by any of them. Everything stated here was taken from their own public pricing and feature pages and is written in our words; no screenshots or copy have been reproduced. Pricing and features change — check the source before deciding, and tell us if anything here is out of date.