Copywriter

Agency job postings in the UK, measured as a share of all agency postings — not employment, and not in-house or freelance copywriting. A share can fall because other roles grew; here total postings rose while writing rose far less.

Verifiable change in hiring, headcount, hours or job scope. Highest weight — but causal attribution still has to be argued, not assumed.

Measured across Claude chat, Cowork, Claude Code and Anthropic's first-party API during the report's sample period (spring 2026, covering the 15 April US tax deadline). It counts conversations, not people and not employers: 81% of conversations producing a blog or article and 80% of those producing marketing content were classified work-related, but nobody asked whether an employer commissioned, paid for or accepted that output. Higher delegated autonomy on the agentic surface appears across 26 of 31 output types — 0.37 points on a 1-5 scale across all conversations — and it survives holding the model constant: among Sonnet-served conversations the gap is still 0.26 points, which is why the authors read it as a property of the product rather than the model. Anthropic operates the tool being measured and sells it, and it is the only party that can see this data at all - both facts belong beside the numbers. Nothing here is occupation-specific: the mapping from conversation to occupation is a classifier's, not an employer's.

Measured, large-scale use of a tool for real work, where the decision to use it was the worker's rather than an employer's. It is more than a capability record — the work is real, not a demo — and less than a deployment record, because no employer put it into production, required it, or built a process around it. Weighted `cautious`: `automating` means the machine can do the task AND there are adoption signs, and this is an adoption sign — but usage can be experimental, and much of the measurement comes from a party with a stake, so one record is never enough and two independent ones are. Note who is counting. Vendor telemetry sees this directly and sells the tool, so such a record names that stake in its scope; a statistics agency asking firms whether their workers use AI in tasks sees the same channel with no stake at all, and that is the better source where it exists.

The UK's advertising regulator ruling on paid search ads for an accident claims management company, found through the ASA's own AI-based ad monitoring. The advertiser said the claim '100% Free Service' in one ad had been generated by AI and had since been removed, that it had strengthened its internal approval process and that any future ad copy would be approved by one of its directors. The ASA found the ads misleading on the nature of the service, cost and savings, in breach of the CAP Code's rules on misleading advertising, qualification, prices and price comparisons, and ruled that they must not appear again in the form complained of. The director sign-off is the advertiser's own commitment, recorded in the ruling, not something the ASA required; the ruling is about one small advertiser's pricing claims.

Failure, rollback, regulation or cost is suppressing adoption. Can lower an assessment or widen its uncertainty.

One global consumer fintech's in-house marketing; company-reported figures with no independent audit. Applies to volume marketing copy, not to positioning work or regulated claims.

An employer has put it into production. Can move the baseline — weighted by scale and how similar the setting is.

A paper by Meta researchers about the ad-text generator in Meta's own ads product. The model was trained on pairs of ads in which advertisers changed only the text, treating the variant with the higher click-through rate as preferred, so the results of copy tests became the training signal; in a randomised test from February to April 2024 across 34,849 US advertisers and about 640,000 ad variations, it raised click-through by 6.7% against the earlier model, and advertisers using it created 18.5% more variations. Advertisers must still select each variation before it runs. Meta sells both the ads and the tool.

An employer has put it into production. Can move the baseline — weighted by scale and how similar the setting is.

An online retailer's engineering blog about its own use of generative AI. For product descriptions, its human experts wrote several hundred high-quality examples and the model was fine-tuned on them to learn the company's language, style and template, producing descriptions written in the Stitch Fix brand voice; for ads, the model generates headlines and copywriters review and edit them to align with the brand's tone and messaging, an approach the company says it now uses for all Facebook and Instagram ad headlines. It gives no volumes beyond the several hundred examples, and it is the company's own post.

An employer has put it into production. Can move the baseline — weighted by scale and how similar the setting is.

Entering as a volume writer is no longer a plan; that market has been repriced to near zero and it will not recover. What is left is a smaller, better-paid job about knowing customers, owning a voice and reading results. The way in is to attach yourself to a product and its customers rather than to writing in general: one company, one audience, learned deeply, with the work you did and the numbers it moved written up.

If your income depended on volume, it has already fallen or will. If it depended on strategy, voice and results, it has probably held, and the demand for someone to own coherence across machine-produced output is new demand in your favour. The uncomfortable move is to stop selling words and start selling judgement, priced accordingly.

Reshape the role

Sell the message, not the words

Clients still cannot work out what to say. If you can, the drafting is a detail and the price of the drafting is irrelevant to you.

Reshape the role

Own the brand voice system

Someone has to write the guidelines that tools and colleagues follow and catch the output that drifts. That job did not exist at most companies three years ago.

Adjacent move

Product marketing or lifecycle marketing

These roles are message-and-results jobs with a writing component, rather than writing jobs. The skills transfer directly and the seats are more stable.

Not on volume, and not by competing on price with tools that cost cents. Freelancers who are still doing well have narrowed to a niche where they know the customer better than the client does, sell positioning and messaging rather than word counts, and often review machine-produced copy for clients who found out the hard way that fluent is not the same as effective. It is a smaller market with fewer people in it, and it rewards specificity over range.

Watch your own brief inbox. The change shows up as briefs that arrive with a draft already attached and ask you to 'polish' it. When most of your work is editing someone else's generated first pass, the repricing has already happened whether or not your rate has changed yet.

Volume copy — product descriptions, variants, the routine output that filled the week and paid the bills. Finding the message and high-stakes copy are not in the same position, but they are bought in small, senior quantities, often by a founder or head of marketing rather than from a writer.

Neither answer is safe by itself. What has held up is copy attached to consequence — regulated claims, pricing, launch messaging — because someone has to be accountable when it is wrong. Specialising into a domain where being wrong is expensive is a more defensible move than specialising into a format.