A copywriter quotes you €1,200 for a website rewrite. A second one quotes €350 and promises it in three days. Five years ago that gap would have told you something about experience. Today it mostly tells you how much of the job each of them hands to ChatGPT or Claude — and whether they are being upfront about it.
That is the new reality of buying outside help. Most freelancers and small agencies now use AI somewhere in their process: research, first drafts, image concepts, code, transcription, reporting. That is not a problem in itself. Used well, it means you get faster turnaround and better value. Used badly, it means you pay professional rates for unchecked machine output, with your confidential data pasted into tools you never agreed to.
This guide is for the buyer’s side of that relationship: the owner or manager hiring a designer, marketer, developer, bookkeeper or consultant. It covers the questions to ask, how to judge price, the red flags, and what to put in writing.
Why the old way of hiring outside help no longer fits
Traditional hiring rested on three signals: portfolio, day rate and turnaround. All three are now less reliable.
- Portfolios can be padded. A polished set of samples tells you less when a tool can produce a convincing “case study” or mock-up in minutes. What you need to know is how the person thinks, not just what the output looks like.
- Day rates can mislead. If AI turns a two-day task into a half-day task, a high day rate may still be cheap — and a low fixed price may still be poor value if nobody checked the result.
- Turnaround is no longer proof of effort. Fast delivery used to signal capacity. Now it can signal that the work was generated rather than crafted.
The owners we speak to tend to describe the same worry: “Am I paying for expertise, or for someone’s ChatGPT subscription?” The answer is that you should be paying for judgement — the brief, the choices, the checking and the accountability — and the hiring process needs to test for that directly.
Five questions to ask before you hire
You do not need to ban AI, and you should be wary of anyone who claims not to use it at all in 2026. Instead, ask these five questions on the first call. Good suppliers answer them easily; weak ones get vague.
1. “Where in your process do you use AI, and where don’t you?”
You want a specific answer: “I use Claude for research summaries and first-draft structure, but headlines, final copy and all client-facing claims are mine.” A one-word “sometimes” is not an answer.
2. “Which tools, and on which plan?”
This matters for confidentiality. Business and team plans from the major providers generally do not train on your inputs by default; some free consumer tiers may. If your supplier will see customer data, pricing or unreleased product details, you need to know whether it is going into a free consumer account.
3. “Who checks the output, and how?”
Ask for the review step in concrete terms. For a content writer: fact-checking against sources, with links. For a developer: tests and code review. For a bookkeeper: reconciliation against bank statements. “I read it through” is the minimum, not a process.
4. “Can you show me something you produced with AI and what you changed?”
This is the single most revealing question. A strong freelancer will happily show a raw draft next to the final version and explain the edits. That tells you exactly what you are paying for.
5. “Who owns the work, and can you guarantee it is original?”
Ownership of AI-assisted work is still a grey area in many jurisdictions. Our guide on who owns AI-generated content explains why this matters, particularly for logos, brand assets and anything you might want to register or defend later.
How to judge price when AI does part of the work
The fear many owners have is overpaying. The opposite risk — underpaying for work that then needs redoing — is just as common. Here is a practical way to think about it.
Prefer fixed-price or outcome-based quotes for repeatable work. If a supplier bills hourly and AI has halved their hours, either they pass on the saving or they are padding the timesheet. A fixed price for a defined deliverable (“10 product descriptions, 150–200 words each, two revision rounds”) removes that tension. Many good freelancers have already moved this way — see our piece on how to price services with AI for the seller’s side of the same shift.
Keep hourly or day rates for genuinely uncertain work. Strategy, discovery workshops, complex integrations and anything where the scope will change are still best bought by time, with a cap.
Compare value, not inputs. A worked example: you need 20 blog posts for the year.
- Supplier A quotes €150 per post. They use AI for outlines and drafts, then edit lightly. Turnaround: two days.
- Supplier B quotes €320 per post. They interview someone on your team for 20 minutes per post, use AI to draft from the transcript, then rewrite and fact-check. Turnaround: one week.
Supplier A costs €3,000 for the year; Supplier B costs €6,400. If Supplier A’s posts read like everyone else’s and rank for nothing, that €3,000 is wasted. If Supplier B’s posts carry your actual expertise and bring in two clients, the extra €3,400 is the best money you spent all year. The question is never “did they use AI?” but “what did they add that I could not get myself in an afternoon?”
A useful test: before you hire for a writing or design task, spend 30 minutes trying it yourself with an AI tool. If your result is 80% as good as the samples you are being sold, the supplier needs to justify the remaining 20% — or you should be paying for editing, not creation.
Red flags that a supplier is cutting the wrong corners
None of these is proof on its own, but two or three together should make you cautious.
- Prices far below the market with no explanation. A 60–70% discount on the going rate usually means little or no human review.
- Generic proposals. If the proposal could be sent to any business in your sector, it probably was. Look for references to your actual website, customers or competitors.
- Confident facts with no sources. Statistics, legal claims or product specifications that cannot be traced back to anything are a classic sign of unchecked output.
- Reluctance to discuss process. A supplier who becomes defensive when you ask how they use AI is more of a risk than one who uses it openly.
- Requests for more data than the job needs. A copywriter does not need your full customer list. Someone asking for exports “to feed into the tool” should explain exactly where that data goes.
- Identical structure across deliverables. If every article opens with a question and closes with “In conclusion”, nobody is editing.
What to put in the contract
You do not need a 20-page agreement for a €500 job, but a short AI section saves a lot of trouble. For anything above a few thousand euros, or anything touching personal data, have a professional review the wording. The essentials:
- Permitted use. State that AI tools may be used for named tasks (research, drafting, transcription) and that final deliverables are reviewed by a named person.
- Confidentiality and data. Specify that your confidential information and any personal data may only be entered into tools on business plans that do not use inputs for training, and list any tools you have approved. Under GDPR, if the supplier processes personal data for you, they are likely a processor and you need a data processing agreement.
- Ownership. The supplier assigns all rights they hold in the deliverables to you, and warrants that the work does not knowingly infringe anyone else’s rights.
- Accuracy. The supplier is responsible for the accuracy of facts, figures and claims in the deliverables, regardless of how they were produced.
- Disclosure. The supplier tells you if a deliverable is substantially AI-generated (for example, an image or a full article), so you can decide whether to label it or use it at all.
If you sell services yourself, the same thinking applies in reverse — our guide to AI clauses for client contracts covers the clauses your own clients will increasingly expect from you.
How to brief so you get the AI speed without the AI sameness
A lot of generic output is a briefing problem, not a supplier problem. A thin brief gives everyone — human or machine — nothing to work with except the average of the internet. Give your supplier the raw material that makes the work yours:
- Your voice. Three to five pieces of writing you like, and one or two you dislike, with a sentence on why.
- Your customers’ words. Real phrases from reviews, sales calls and support emails. This is the single biggest upgrade to any AI-assisted copy.
- Your proof. Actual numbers, client results and examples the supplier is allowed to use.
- Your boundaries. Claims you cannot make, competitors you never mention, topics that are off-limits.
- Your definition of done. Word counts, formats, revision rounds and who signs off.
Then run a small paid trial before committing to a retainer. One article, one landing page, one month of bookkeeping. A €200–€500 trial will tell you more about a supplier’s AI habits than any number of calls.
The bottom line
Hiring outside help in the AI era is not about catching people using AI. It is about knowing what you are paying for. Ask where the tools are used and where the human judgement comes in, prefer fixed prices for repeatable work, protect your data in writing, and brief with the material only you have. Do that, and AI becomes a reason your freelancers and agencies deliver more for the same budget — rather than a reason you pay professional rates for something you could have generated yourself.
Where does your business stand on AI?
Take the free 3-minute AI Readiness Quiz and get a personalised score with your next steps.
Take the Free Quiz →