We ran 3 AI assistants through a Nairobi week

The demos always work. That is their job. So instead of trusting the stage, I spent a normal week in Nairobi leaning on three of the most popular AI assistants, ChatGPT, Gemini, and Claude, for the ordinary tasks I would have done anyway. Drafting messages, working out a budget, getting directions, checking facts, switching between English and Kiswahili the way people actually talk. I was less interested in which is smartest on a benchmark and more interested in which is least annoying when life is normal and the connection is not.

A note before the findings: this is a field test, not a lab. I used each assistant as a regular person would, on a phone, on regular data, over one week. Your mileage will vary with the version, the day, and your questions. With that said, here is what held up and what did not.

Everyday writing and thinking

For drafting and tidying up text, all three were genuinely good, and honestly close enough that preference came down to tone. Each could turn a rough WhatsApp rant into a polite message, summarise a long document, and rough out a plan. If your main use is writing help, you almost cannot go wrong, and the free tiers are already strong enough for most of it.

Local knowledge, the weak spot

This is where the gap between the stage and the street showed. Ask about a global topic and the answers were solid. Ask about a specific Nairobi neighbourhood, a local fee, a small Kenyan company, or a current matatu route, and the confident wrong answers crept in. None of them should be trusted on hyper-local detail without a check. They are world-class generalists and shaky locals.

Language and code-switching

Kiswahili was handled better than I expected, and basic code-switching between English and Kiswahili mostly worked. Sheng and very colloquial phrasing were hit and miss. For formal Kiswahili they were useful; for the way people actually text, results wobbled.

The unglamorous bit: data and connection

Here is the part no launch mentions. These tools live in the cloud, so they eat data and they need a signal. On a strong connection they felt instant. On a weak one, or when the network dropped, they stalled, and a long back-and-forth quietly chews through a bundle. If you are on a tight data plan, that is a real cost, and it shaped how I used them: shorter exchanges, fewer giant pastes.

So, the verdict

Treat any of the three as a sharp, fast assistant for thinking, writing, and getting started, and treat all three as unreliable witnesses on local specifics. Use them to draft and to reason, then verify anything local or anything that matters before you act on it. Pick the one whose tone you like, because on the everyday stuff they are closer than the marketing suggests.