Can AI tailor your CV? What has to happen first

If you've thought, "why wouldn't I just paste my CV and the job description into ChatGPT, Claude or Gemini and ask it to tailor things for me?", you're not missing something obvious.

Plenty of people already do exactly that.

And for some of what a CV needs, general-purpose AI can be genuinely useful. It can tighten a clumsy sentence, shorten something that's trying to do too much, suggest a clearer structure or help you find the words for something you know you did but haven't described particularly well.

There's nothing wrong with that. You don't need a specialist platform simply to rewrite a sentence.

The more interesting question is what has to happen before the rewriting starts.

In short: General-purpose AI can be an excellent writing and thinking partner. But tailoring an application properly involves more than producing better prose. Someone still has to work out what the role actually needs, which parts of your experience prove it, where the evidence is incomplete and whether the finished CV still says only what you can honestly stand behind.

The hard part happens before the rewrite

A job application looks like a writing task because the thing you eventually send is a document.

But the harder part is usually judgement.

Before anything gets rewritten, somebody has to answer questions such as:

• What is this job description really asking for?

• Which requirements look essential and which are merely desirable?

• Which parts of my experience genuinely support each one?

• Where is the evidence strong, partial or missing?

• Is there relevant transferable experience hiding behind a different job title or industry?

• Which achievements deserve more prominence for this particular reader?

• Is an impressive-looking claim actually supported by what I did?

• Have I given enough context for a number or achievement to mean anything?

• Is there something important I have simply forgotten to include?

• Does the wording reflect the employer's terminology where that is honestly accurate?

• Has anything drifted into keyword stuffing or exaggerated relevance?

• Does the finished version still sound like me?

• Could I defend every number, title, outcome and claim if somebody asked about it in an interview?

A good AI prompt can ask for all of those things.

But notice what has happened.

You're no longer simply asking:

"Please tailor my CV to this job."

You're designing an application-optimisation process.

And the better you want that process to be, the more you have to know what to ask for in the first place.

The prompt has to know what to look for

This is where experience with AI helps, but doesn't make the underlying problem disappear.

An experienced user can build a detailed prompt. You can tell the model not to invent anything. You can ask it to map your CV against the job description, separate strong matches from partial ones, flag missing evidence, preserve your tone, keep dates and figures consistent and explain every material change.

You can give it examples of your writing. You can build a project around your CV. You can maintain a separate document containing achievements the original CV doesn't mention. You can create your own rules for what the AI is allowed to infer and what it must ask you about.

All of that is possible.

But the quality of the process still depends heavily on what you remembered to include.

A model cannot follow an instruction you never gave it.

It cannot reliably work with an achievement that isn't in the context available to it.

And it cannot reliably distinguish between a genuinely missing requirement and relevant evidence that happens to be buried in experience you've never discussed with it.

That's not a criticism of AI. It's simply the difference between a general-purpose tool and a process built around one specific job.

General-purpose AI starts with what you give it and what you remember to ask. CareerIbis starts by working out what the role needs and what your evidence actually proves.

Memory helps, but it isn't the same as an evidence system

This distinction matters more now than it did a few years ago because AI tools are getting much better at carrying context.

ChatGPT can, when memory is enabled, carry useful information from previous conversations and other available context into later ones (OpenAI Help Center, 2026). Claude Projects let you keep documents, instructions and project knowledge together so Claude can use them across work inside that project (Anthropic, 2024), and Claude now has its own memory and past-chat search too, depending on your plan and settings (Claude Help Center, 2026).

Those features are useful.

But useful memory is not the same thing as a candidate-confirmed evidence record.

Something you happened to mention in a conversation six weeks ago is not automatically the same as:

"This is an achievement I have checked, confirmed as accurate and am comfortable using in an application."

Likewise, a remembered detail is not necessarily the detail that matters for this job.

The question isn't merely whether the AI knows something about you.

It is whether the right evidence has been deliberately surfaced, tested against the current role and checked before it becomes part of the application.

Good prose can still contain a bad claim

There is another complication.

AI is very good at making language sound coherent.

Sometimes that is exactly what you want.

But polished wording can also make a weak or unsupported claim sound surprisingly convincing.

OpenAI itself has written about why language models can produce confident but incorrect information (OpenAI, 2025). Applied to a CV, the risk is not necessarily an outrageous fabrication. It can be much subtler.

A sentence such as:

"Led the strategic transformation of..."

may have started life as:

"Helped implement..."

Or a model may fill in an apparently harmless piece of missing context because it makes the sentence flow better.

The result can read beautifully.

That does not make it true.

So one of the most important checks in AI-assisted CV work is not:

"Does this sound impressive?"

It is:

"Can I actually stand behind this?"

Plausible is not the same as true.

And when the document may end up being discussed in an interview, that distinction matters.

One excellent result is not the same as a repeatable process

There is also a difference between getting a strong result once and running the same disciplined process across application ten, fifteen or thirty.

You might write an excellent prompt today.

Next week, you adjust it.

Then you start a new chat.

Then you upload a newer CV.

Then you forget to mention an achievement that appeared in the previous conversation.

Then you ask for a slightly different style and one of the earlier evidence checks quietly disappears from the process.

None of those things means you're bad at using AI.

They are simply normal consequences of running a multi-step workflow manually.

The challenge is not only:

Can I produce one good tailored CV?

It is:

Can I apply the same evidence discipline every time without having to rebuild the process from memory?

That is a different problem.

That is a lot of process for one prompt

And this isn't the same as asking AI to rewrite an email or tidy up a paragraph for a presentation.

That is a lot of process to leave to a prompt you wrote in five minutes when the application may be for a £40,000, £60,000 or £100,000 job.

The point isn't that the AI will inevitably get it wrong.

It may do an excellent job.

The point is that the stakes justify understanding what is actually being delegated.

If the model is handling the wording, fine.

But somebody still needs to own:

• the interpretation of the role;

• the evidence;

• the gaps;

• the factual checks;

• the relevance decisions;

• the voice;

• and the final judgement about whether the document genuinely represents you.

That somebody is still you.

An easy-to-miss one: what you're pasting in

There's also a more practical consideration.

Your CV contains personal history. A job description may contain information an employer expects to remain inside its hiring process.

Pasting both into a third-party AI tool means sending that material to another company's systems under whatever privacy settings and data-handling terms apply to your account at the time.

For most people, most of the time, that may be a perfectly reasonable trade for the convenience.

It should still be a deliberate choice rather than an automatic one.

That matters particularly where a job description refers to an unreleased project, internal restructuring, sensitive client work or anything else that clearly wasn't written for unrestricted circulation.

And then there's the elephant in the room

Let's talk about the part people often notice after they've created their shiny new CV or resume masterpiece.

Does it actually sound as though they wrote it?

There is an entire cottage industry of supposed "AI tells".

Em dashes get blamed. So do perfectly balanced sentences, suspiciously tidy paragraphs, Oxford commas, excessive headings and words nobody remembers using before AI tools became part of everyday writing.

None of those proves anything.

Human beings use em dashes.

Human beings use Oxford commas.

Some British publishers prefer -ize.

Good writers occasionally produce annoyingly neat sentences all by themselves.

The problem isn't one punctuation mark.

It's accumulation.

If your old CV sounds like you, your LinkedIn profile sounds like you, your emails sound like you, and then your new application suddenly sounds as though it has been edited by the world's most enthusiastic management consultant, something can feel slightly off even if nobody could identify one sentence and explain why.

So the final question shouldn't merely be:

"Does this sound professional?"

It should be:

"Would I actually say this?"

If the answer is no, change it.

AI should help you express your experience more clearly, not quietly replace your voice with its own.

So what is CareerIbis actually doing differently?

CareerIbis is not trying to out-chat ChatGPT, Claude or Gemini.

It is doing the application groundwork that a general-purpose AI prompt otherwise leaves to you.

Rather than starting with:

"Here's my CV. Please improve it."

CareerIbis starts by comparing your actual CV with the actual job description.

It looks at what the role is asking for, where your existing evidence already supports it and where something appears missing or unclear before anything gets rewritten.

Where evidence is missing, the goal isn't to manufacture a plausible substitute.

It is to find out whether the evidence genuinely exists.

That may mean surfacing something you did but never thought to put on the CV. It may mean recognising that experience from another role, industry or environment is genuinely transferable. Or it may mean concluding that there really is a gap.

All three are useful answers.

Only one of them involves rewriting.

That is why the process starts with evidence rather than prose.

You can build this yourself

None of this means you need CareerIbis in order to use AI well.

You can absolutely build the process yourself.

You can create a project, upload your CV, maintain a master evidence document, write detailed instructions, tell the model what it may and may not infer, map requirements, interrogate gaps, preserve your voice, check every claim and repeat the sequence for the next role.

If you enjoy doing that, or you've already built a workflow you trust, there is no reason to pretend otherwise.

The question is not whether you're capable of it.

The question is whether this is where you want to spend your expertise.

Using general-purpose AI well is a skill.

Using it to run an evidence-led application process is several skills layered together: prompt design, role analysis, evidence retrieval, CV judgement, factual verification, editing and version control.

Applying for a job shouldn't require you to become a prompt engineer, evidence analyst and CV editor at the same time unless that's genuinely how you want to spend the evening.

A few checks worth doing, whichever tool you use

Before you send an AI-assisted CV, ask:

• Does every specific number, title or outcome match something you can genuinely stand behind if somebody asks about it in an interview?

• Have you checked the job description itself, or mainly asked the AI to make the CV "sound more relevant"?

• Do you know which requirements your evidence actually supports and which ones remain genuine gaps?

• Has anything become more impressive in the rewrite than it was in reality?

• Would you actually say this sentence?

• Is there relevant experience missing simply because it wasn't on the original CV or in the prompt?

• Is anything in the job description something the employer probably wouldn't want sitting in a third-party chat log?

• If you tailor another application next week, do you have your strongest evidence saved somewhere you control, or are you relying on the AI to remember the right things?

• Could you repeat the same checking process on the next application without having to remember how you did it this time?

General-purpose AI can do a remarkable amount.

The part worth paying attention to is everything that still has to happen around it.

Frequently asked questions

Can ChatGPT, Claude or Gemini tailor my CV to a job description?

Yes. General-purpose AI can compare documents, suggest stronger wording, restructure content and help you think through relevance. The quality of the result depends on the evidence you provide, the instructions you give and the checks you perform afterwards. The model can help with the process, but it does not automatically know whether every relevant piece of your experience has been supplied or whether every claim in the final version is genuinely supported.

Is it safe to paste my CV and a job description into AI?

That depends on the tool's current privacy and data-handling settings and on what is actually in the documents. Check the controls that apply to your account rather than relying on a blanket rule. It is also worth avoiding pasting information from a job description that is clearly internal, confidential or not intended for wider circulation.

Can a recruiter tell if I used AI to write my CV?

There is no reliable universal answer and no single punctuation mark or writing pattern proves AI involvement. What can stand out is a sudden change in voice. If the CV sounds very different from the way you normally write or speak, that is worth correcting whether or not anybody could technically identify why.

Should I use general-purpose AI and CareerIbis together?

You can.

They are doing different jobs.

General-purpose AI is a flexible writing and thinking tool. CareerIbis is designed around the application process itself: comparing the role with your evidence, surfacing alignment and gaps and helping you work from what is genuinely true before the writing stage takes over.

For some people, one will be enough.

For others, using both will make perfect sense.