Sorting the Work: What a Model Does and What a Writer Does
This is usually framed as a purchase decision, which is why it rarely resolves. It becomes tractable if you split a resume into the separate jobs it requires and ask which are done by rephrasing text and which are done by knowing things about the world. A language model does the first well and cannot do the second at all. A professional writer’s entire product is the second. No prices appear here, and no claim is made about which produces better results, because neither claim can be honestly supported.
Two different resources, described accurately
A model is a text transformer with no access to your record. It knows what resume prose looks like in general, and it knows whatever you put in the prompt. It has never seen your performance review, does not know what your team actually shipped, and cannot find out. Its confidence is unrelated to whether it is right, so its assertions about your work are generated rather than known.
A professional resume writer is buying you two things. One is a conversation: an outsider asking what you did, why it mattered, what the hard part was, and refusing the first vague answer. The other is calibration — someone who has read many resumes for roles like yours and has a sense of what lands. Neither is a writing skill exactly. Both depend on the writer actually having relevant exposure, which is worth establishing before engaging anyone.
The tasks, sorted
Rephrasing something you already wrote. Model. Fast, unlimited attempts, no scheduling. This is genuinely the strongest use, and it is available whether or not you also hire someone.
Compressing a long paragraph into a bullet. Model. Mechanical, and easy to check because the source is in front of you.
Consistency across a document. Model, with your review. Tense, verb forms, date formats, parallel structure across sections you edited at different times.
Deciding what goes in and what is cut. Human judgement — yours, or someone’s you trust. This requires knowing what the target reader values, and a model has no access to that beyond generic patterns.
Extracting facts you have not articulated. A person. This is the interview effect, and it is the thing most people are actually buying when they hire a writer. Being asked “what would not have happened if you had not been there” by someone who waits for an answer produces material that no prompt does, because the material was never written down to be pasted in.
Judging your market. A person with relevant recent exposure, or nobody. A model will answer this question fluently and without grounding, which is worse than not answering.
Verifying that a claim is true. Only you. Neither a model nor a paid writer knows your record. A writer will use what you tell them, and if you overstate it to them, they will write the overstatement up more persuasively. That is the mechanism behind a claim you cannot back up surviving into a finished document.
The failure modes are not symmetrical
A model’s characteristic failure is fluent invention. Ask it to improve a thin bullet and it will supply scale, scope, and an outcome, because that is what a strong bullet contains and it has no way to know yours did not. The additions are plausible, unmarked, and yours to defend later — the general pattern being where drafted bullets drift from your record.
A hired writer’s characteristic failure is different: a polished document in a voice that is not yours, describing your work at a level of confidence you did not authorise. It reads well and then you have to speak it aloud in an interview. Because the writer worked from what you said, any exaggeration you offered comes back amplified and formalised.
Both failures land in the same place. Something on the page that you will be asked about and cannot substantiate. Whichever route you take, the last pass is the same: read every line and ask what happens when someone follows up, which is what the interview questions your resume invites is about.
The third option people skip
Write it yourself, use a model for rephrasing, and get one honest reader — a colleague in your field, a former manager, someone who has hired for the role. That reader supplies the calibration and the awkward questions, which are the parts a model cannot provide, and you keep the voice.
The awkward questions matter more than the polish. If a stranger reading your document cannot say what you do and what you are applying for, no amount of professional wordsmithing fixes it, and if they can, most of the remaining gap is small.
How to decide
Ask what your bottleneck is. If you have your material written down and it reads clumsily, that is a rephrasing problem and you already have the tool for it. If you cannot get the material out of your head at all, or you cannot tell what matters, that is the problem a person solves — and the person could be a paid writer or could be a well-chosen friend with an hour.
What is not a good reason to buy either is uncertainty about whether the resume is good enough. Neither a model nor a writer can tell you that, because neither knows what the employer will do with it. What both can help with is making the document say clearly and accurately what you actually did, which is the only part under your control — and starting from written-down facts rather than a blank prompt shortens the job whichever route you pick.