When AI changes the work · 5 min read

If AI can write the first draft, what is a copywriter for?

AI has changed the price of a first draft. Here is what still calls for a writer, which adjacent directions to explore and a way to test them.

Written by Nivaan, founder of OffLadder · Last reviewed 10 October 2026 · Sources listed at the foot of this page

Separate the writing from the job of writing

A freelance blog writer paid for undifferentiated articles faces different pressure from an editorial lead who interviews experts, interprets evidence and is responsible for accuracy. A brand writer testing how a message lands also does different work from someone filling a content calendar. Treating all of them as one occupation hides the change.

The World Economic Forum's 2025 employer survey projects disruption across many roles and skills by 2030. It is a survey of expectations, not a verdict on your next contract. Better evidence comes from looking at the briefs customers still commission and what they refuse to let a tool do alone.

The useful skills may be upstream or downstream

Upstream: discovering what an audience needs to know, interviewing a reluctant source, finding the one true claim and choosing the angle. Downstream: checking accuracy, editing for clarity, designing a quality rubric, testing copy with users and deciding whether to publish. Some of this work is done by strategists, content designers, UX writers and AI evaluators. Some job titles overlap. Do not assume a role called 'model evaluation writer' is a standard vacancy everywhere.

A particularly revealing test is editorial judgement. It is easy to say AI copy sounds wrong. Can you specify the problem well enough that someone else could identify and fix it? That turns taste into work other people can use.

First separate writing output from writing judgement

A writer can now produce a competent-sounding first draft quickly. That does not mean the draft has a point of view, contains the truth or works for the intended reader. The commercial question is whether the organisation values those differences enough to pay for them. For bulk product descriptions, the answer may increasingly be no. For a complex service, sensitive public information or a distinctive brand, editorial thinking may still matter greatly.

List the work you did beyond typing: interviewing someone to find the real story, discarding a weak claim, turning customer language into a headline, deciding what to leave unsaid, editing a founder's vague opinion into a usable argument, or proving that a message changed behaviour. Those are different from generating paragraphs. Build your case around the choices and results, not the word count.

Four routes, with their costs

Brand and messaging strategy rewards research and hard decisions about positioning. You need to show why a claim is believable, who it serves and what you deliberately excluded. UX content requires understanding users, error states and product constraints; the output is not just friendlier buttons. AI-content evaluation involves rubrics, edge cases and repeatable quality checks rather than liking or disliking prose. Technical or regulated communication rewards accuracy, domain learning and careful reviews. These are options to investigate, not guaranteed growth categories. Each has competition and its own training requirements.

Here is a practical test: take a vague landing-page claim such as 'We empower teams through innovative solutions.' Interview a real user, or use genuine public user complaints about a comparable product. Rewrite the claim so it names a problem and an outcome you can substantiate. Produce three headline versions and a short note explaining why each might work. Then ask a reader unfamiliar with the product what they think it does. The answers are more useful than whether the copy sounds clever.

Create a portfolio of decisions, not only finished copy

For each sample, show the brief, the evidence you used, what you changed and what you learned from feedback. If AI assisted you, disclose its role appropriately and show your judgement. Hiring managers do not need a hundred adjectives describing your creativity. They need to see how you make a decision when the brief is poor, the facts are uncertain and the first draft is ordinary.

If your favourite part was interviewing, investigate research and positioning. If it was resolving complex product language, investigate content design. If it was finding holes in AI output, investigate evaluation. If you hated all three, that is important evidence, too.

Where OffLadder fits

OffLadder can put adjacent directions on the table without telling a writer to 'just become a prompt engineer'. Explore gives you small tasks, and Laddie helps examine what your choices reveal. Try a research, UX or evaluation experiment, then record which part made you curious enough to keep going. Begin at https://offladder.com/. You are testing your relationship with the work, not defending a job title.