Tailoring a Resume With AI: Re-Emphasising vs Acquiring

Tailoring is the one AI resume task where the model holds both documents, and that changes its behaviour. Asked to match your resume to a posting, it will close the distance between them — and there are two ways to close that distance. One is re-emphasising what you already have. The other is writing the posting’s requirements onto your page. A model cannot tell which it is doing, because both produce a resume that lines up with the posting, and only one of them is true.

The two moves, side by side

Say a posting asks for experience running vendor relationships, and your resume mentions that you handled purchase orders for one supplier.

Re-emphasising promotes that line up the page, gives it its own bullet instead of a clause, and names the thing plainly: managed ordering and delivery scheduling with our packaging supplier. Nothing was added. The reader now sees it first.

Acquiring produces: managed vendor relationships across the supply base, negotiating terms and monitoring performance. Three things arrived that you never supplied — a plural supply base, negotiation, and performance monitoring. It reads better. It is also a claim you would have to survive a conversation about.

The second version is what you get by default, because you asked the model to make your resume match and the fastest route to matching is to write the requirement into the page. Nothing in the process distinguishes “you already have this, stated obscurely” from “you do not have this.” Both look like a gap in the text.

What to ask for instead

The fix is to never ask for a tailored resume. Ask for the comparison, and do the tailoring yourself.

Useful requests, all of which leave the writing with you:

• Here is a posting and here is my resume. List the requirements in the posting that my resume does not currently address. Do not rewrite anything.
• Which of my existing bullets is the closest match to each requirement?
• Reading only my resume, what would you guess this person’s main strength is? Then compare that to what the posting is asking for. A mismatch tells you the page is arguing for the wrong thing, which is a reordering problem, not a rewriting one.
• Which requirements in this posting look central and which look like boilerplate? You will often disagree, and disagreeing is informative.

Each of these returns a list you evaluate. None of them returns prose you might paste. That is the whole design: the model becomes a reader, and a reader cannot fabricate your experience because it is not being asked to produce any.

Then classify every gap yourself

Take the gap list and put each item in one of three buckets. This is the step that does the actual work.

Have it, buried. The evidence is somewhere on the page but stated in your internal vocabulary, or three lines down in a role that is fourth from the top. Fix by promoting and renaming, using words a reader outside your old employer will recognise.

Have it, not on the page at all. Real experience that never made the cut. Add it, in your own words, from your facts file. This is why the file exists — the material is already written down and already true.

Do not have it. Leave it alone. Not softened, not gestured at, not implied by adjacency. A posting is a wish-list assembled by several people, and no candidate satisfies all of it. A resume that answers most of a posting honestly beats one that answers all of it unbelievably, and the gap you leave open is a question you can answer in an interview rather than a claim you have to defend. Where the missing thing is a skill, a skill you barely know covers how to represent partial familiarity without overstating it.

The keyword trap

The advice to “use the posting’s keywords” is where tailoring most often turns into acquiring. Reusing the employer’s vocabulary for something you genuinely did is just clear writing — if they say customer success and you say account management, say customer success. Importing their vocabulary for something you did not do is a claim, and it does not stop being a claim because it was described as a keyword.

The test is whether the word survives a follow-up question. You mention stakeholder management — tell me about a time that was difficult. If you have an answer, the word was yours. If the word arrived from the posting via the model, you are constructing an anecdote in real time under observation.

A model asked to add keywords will not apply that test, because it has no way to. It will place the terms where they fit the sentence.

Watch what happens to the summary

The professional summary is the most tailored part of most resumes and the part that drifts fastest, because it is prose with no fixed facts to anchor it. Ask for a summary tailored to a posting and you get one that describes the ideal candidate for that posting, in the first person, about you. Read those three lines against your record more carefully than anything else on the page — when an AI summary promotes you has the specific patterns to look for.

The version-control problem nobody mentions

Tailoring per application means several files that differ slightly, and the differences are exactly the kind a model introduced. Two months later you cannot remember which version went to which employer, and you are asked about a bullet you no longer have in front of you.

Keep one master document holding everything, honestly written, and treat each application as a trimmed and reordered copy. Save what you actually sent, per employer, and note the date. The point is not tidiness — it is that when the interview comes you can read the same words the interviewer is reading. Keeping a single consistent account across your resume, your profile, and the conversation is the subject of one story across resume, profile and interview.

A workable pass, in order

1. Read the posting yourself first, and write down what you think the job actually is.
2. Ask the model for the gap list, with an explicit instruction not to rewrite.
3. Sort every gap: buried, missing, absent.
4. Promote and rename the buried ones. Add the missing ones from your facts file, in your words.
5. Leave the absent ones absent.
6. Ask for tighter phrasing only on lines you have now written, and read the result against your record with the bullet-checking method.
7. Read the summary last, and separately.

The general boundaries of AI-assisted resume work are in which resume fields AI can touch. The point specific to tailoring is that it is the one task where the tool is actively incentivised to cross them.