Using AI won't set you apart anymore
Surveys differ on the exact numbers, but they agree on the direction. Fiverr, MBO Partners and others put freelancer AI use at roughly three-quarters or more. Fiverr's figure comes from its own users, so treat it as indicative rather than universal.
In Kenya the picture is even starker. Capital FM, reporting on the Digital 2026 Mid-Year Global Update, says Kenya ranked first globally for AI usage. 97.5% of surveyed internet users aged 16+ said they had used an AI tool in the previous month.
If almost everyone is using the same tools, using them isn't a selling point. The same Capital FM article says routine work such as basic customer-service queries, transcription, data entry and routine content production is increasingly open to automation. That is where the pressure on freelancers is greatest.
What clients can't easily get from a chatbot is someone who checks the output, protects their information and takes responsibility for the result. The rest of this post is a method for doing that.
Sort your tasks into three buckets: draft, check, never paste
Before you open any AI tool, decide which bucket the task falls into.
Let AI draft. First versions, rough outlines, transcription of your own recordings, meeting summaries, email replies and admin. Low risk, and you'll finish them yourself anyway.
Check carefully. Anything with facts, numbers, citations, quotes or named people. AI can help you get started, but nothing goes out until you've verified it.
Never paste in. Client data, material covered by an NDA, and personal data such as names, phone numbers, ID details or customer lists. If you're unsure, treat it as off-limits until the client says otherwise.
The buckets only work if you apply them before you start, not after you've already pasted something in.
How to check AI's work (it's the convincing mistakes that get you)
Obvious mistakes are easy to catch. Axios made the point in May 2026: the real trouble is false answers that sound convincing. Better techniques have improved accuracy, but not to 100%.
Forbes lists the kinds of errors to look for:
- Hallucinations, meaning invented facts
- False citations
- Outdated information
- Duplication
- Omitted information
- A mix of true and false claims in the same paragraph
How often this happens depends heavily on the tool and the task. Forbes, citing the 2026 Stanford HAI AI Index, reports hallucination rates from 22% to 94% across 26 top models, depending on the benchmark. Those are test results, not your everyday error rate, but they show why spot-checking isn't optional.
A short routine that works:
- Check every name, date, number and quote against a source you can open yourself.
- Click every citation. If you can't find the source, delete the claim.
- Look at what's missing, not just what's wrong. AI often leaves out the caveat that matters.
- Read it aloud or print it. Mixed true-and-false text is easier to catch when you slow down.
- Ask whether you'd stand behind it if the client called you about it tomorrow.
Protecting client and personal data
Putting confidential material into a public AI tool could breach a confidentiality agreement. Read the tool's terms on how it handles what you enter, and check what your client has agreed to.
On the legal side, the Office of the Data Protection Commissioner (ODPC) has published AI guidance dated July 2026. Reports disagree on its status: Eastleigh Voice describes it as issued, while other outlets call it a draft. Check odpc.go.ke for the current version.
The guidance reportedly says organisations remain responsible for personal data whether they build their own AI, buy software or use popular generative AI platforms. It also flags that generative tools can produce realistic but false content about real people.
It is aimed mainly at organisations. I haven't found anything confirming exactly how it applies to solo freelancers, so don't assume either way. If you handle personal data for clients, our post Data Protection Act for Kenyan Freelancers and Small Websites: What You Must Do in 2026 covers the basics. The safest habit is the third bucket above: don't paste personal data into AI tools unless you know it's allowed.
Tell clients what you do: a one-paragraph AI policy
One report suggests freelancers often disclose AI use only when asked, while clients prefer to hear it upfront. Whatever the legal position in your market (and this isn't legal advice), being open early is cheaper than explaining later.
Put a short paragraph in your proposals or contracts. Something like:
I use AI tools for tasks such as drafting, transcription and admin. I review and take responsibility for everything I deliver. I do not upload your confidential or personal information to AI tools without your written approval. If you have restrictions on AI use, tell me at the start and I'll follow them.
At onboarding, ask which outputs must be AI-free. A client who says "no AI" often objects to AI-generated final assets, not to grammar checking, transcription or code completion. Asking prevents a misunderstanding in either direction.
One more point: ownership. Under current US Copyright Office guidance, parts of an AI-assisted deliverable may not be copyrightable by anyone, so you may not be able to assign rights you don't hold. Kenyan law may treat this differently, so if ownership matters for a big contract, get proper legal advice.
Price your judgment, not your hours
If AI cuts a four-hour task to one, billing hourly means you earn less for being efficient. Plutio argues exactly this. Pricing by outcome fixes the problem.
When you do quote by components, itemise them. Kahawatungu, writing about AI video, suggests separating direction, generation, editing and review. The idea is that you sell the parts AI doesn't replace: deciding what to make, checking it and finishing it properly. That approach works beyond video.
For a writer or researcher, a quote might break down as:
- Brief and approach
- First draft (AI-assisted)
- Fact-checking and sourcing
- Editing and final review
This makes your checking visible, which is the part clients can't get from a free tool. For more on setting prices, see Freelancing as a Researcher: From Knowledge to Income.
The takeaway
AI is now ordinary equipment, like a laptop. Clients won't pay extra because you have it. They'll pay for someone who knows when to use it, catches its mistakes, keeps their information safe and says plainly how they work. Build those habits now, and put them in writing.