Sooner or later, every agency that sells content hears the same question from a client: “Why are we paying for this when ChatGPT can write it for free?”
It’s a fair question, and the usual defensive answers don’t land well. “AI content gets penalised” isn’t quite true. “Our writers are better” is hard to prove in a proposal. The honest answer is more interesting. AI has made words almost free to produce. That is exactly why the parts of copywriting AI can’t produce have become the parts worth paying for.
What happens when the cost of words drops to zero
The supply side of content has already changed. Graphite’s analysis of English-language articles from Common Crawl estimated that AI-generated articles overtook human-written ones in November 2024, after which the share plateaued at roughly half. The same research noted that AI-generated articles largely don’t appear in Google and ChatGPT results.
That second point is the one agencies should pay attention to. When anyone can produce a competent 1,500-word article on any topic in seconds, a competent article stops being an advantage. It becomes the baseline, and a crowded one. The question shifts from “can we produce content?” to “why would anyone, human or machine, choose this content over the thousands of near-identical versions of it?”
Google’s position is about value, not the tool
It’s worth being precise here, because a lot of the advice agencies pass on to clients is wrong. Google has said plainly that using AI to create content isn’t against its guidelines in itself. What it targets is the behaviour. Its spam policies define scaled content abuse as many pages being generated mainly to manipulate search rankings, however they’re produced.
So the real risk for clients isn’t “Google will detect the AI”. It’s that content produced cheaply, at volume and without anything original in it is exactly the kind of content that has nothing to rank on. It gives AI answer engines nothing to cite, and readers no reason to trust the brand that published it.
What clients are actually paying for now
If the words themselves are commoditised, the value moves to everything around them:
- First-hand knowledge. Interviewing the client’s subject-matter experts and pulling out the details, opinions and examples that only exist inside that business. No model has access to these until someone writes them down.
- Judgement. Deciding what to leave out, which claim needs a source, and when a topic isn’t worth writing about at all. AI will happily write anything it’s asked to. Good writers push back.
- Accuracy and accountability. A confident wrong statistic in a client’s content is a liability, especially in regulated industries. Someone has to own the facts.
- Voice. Brands that sound like everyone else are forgettable. AI drifts towards the average by design.
- Structure that works for answer engines. Leading with the answer, writing passages that stand on their own, and backing claims up. This is increasingly what decides whether content is cited in AI answers. We covered why in AEO vs SEO.
- Persuasion. Landing pages, service pages and emails exist to move someone to act. That still depends on understanding the customer better than the competitor does.
Where AI genuinely belongs in the workflow
None of this is an argument against AI. Agencies that refuse to use it will simply be slower and more expensive. The useful line is between AI as an assistant and AI as the writer of record.
| Good use of AI | Still needs a human writer |
|---|---|
| Research summaries and first-pass outlines | The angle, the argument and what the piece is really for |
| Variations of meta descriptions, ad copy and product attributes at scale | Thought leadership and opinion pieces published under a real name |
| Reworking existing approved copy into new formats | Health, legal and financial content where accuracy carries risk |
| Spotting gaps against competitor content | Expert interviews and original insight |
| Proofreading and consistency checks | Final editorial sign-off and fact-checking |
Stop selling words
The biggest change agencies can make is commercial, not editorial. Pricing content per word, or per post, frames copywriting as a production line. That is precisely the part AI has made cheap, so the client’s question about ChatGPT becomes perfectly reasonable.
Pricing and reporting on outcomes changes the conversation. That means briefs informed by search and customer research, expert input, pages that rank and get cited, and copy that converts. A client paying for that isn’t comparing you to a free tool. They’re comparing you to the result they’d get without you.
The agencies most exposed right now are the ones that quietly replaced writers with prompts and kept charging the same. Clients will work that out. The ones best placed are using AI to remove the low-value hours and putting those hours back into the thinking.
FAQ
Does Google penalise AI-generated content?
Not for being AI-generated. Google’s guidance says using AI isn’t against its guidelines in itself. Its scaled content abuse policy targets pages produced mainly to manipulate rankings, whatever tool made them. The practical risk is low-value content that has nothing to rank on.
Can AI content rank and get cited in AI answers?
It can, if it’s accurate, useful and contains something that isn’t already everywhere else. Purely generic AI output rarely clears that bar, because thousands of near-identical versions already exist.
How should agencies explain the value of copywriting to clients?
Focus on what the client gets that a prompt can’t give them. That means their own expertise turned into content, accuracy they can stand behind, a distinct voice, and pages built to rank, be cited and convert. Then report on those outcomes rather than on volume.
If your agency needs content with real research and editorial judgement behind it, see how our white label copywriting works.