AI Search Optimization, Kenya

What AI assistants show when Kenyans ask about businesses

Questions about Kenyan businesses range widely: a hospital that accepts a particular insurer, a solar installer in Nakuru, a firm that can audit a donor-funded project, a safari operator for a family trip, a supplier of cold-chain equipment. The tools behave differently. Google's AI Overviews and AI Mode lean on Google's own index, ChatGPT search, Perplexity and Copilot pull pages at the moment of the question and link them, and Gemini mixes live lookups with what it absorbed in training. The rules for choosing sources are published only in outline.

Observation still gives a usable picture. For Kenyan commercial questions, answers often cite Kenyan news and business publications, directories, travel and review platforms, forum discussions, NGO and development sector listings and company websites that explain their services in plain terms. Smaller Kenyan firms with thin websites are frequently described from whatever a directory or social page says, which may be years out of date.

My starting point is a recorded baseline. With your sales and customer service staff, I list the questions buyers ask, including town names, estate names and Kenyan terms such as SACCO, chama or county, and run them through each tool. Each reply goes into a log: mentioned or not, the sentence used about you, the links offered and any mistake. Patterns across many replies, not one answer a director happened to see, decide what gets fixed first. The general method is on my AI search optimization page.

English, Swahili and Sheng: how answers change with the prompt

Questions reach assistants from Kenya in English, in Swahili, in a switch between the two mid-sentence and, among younger users, in Sheng. AI tools respond to each differently. An English prompt about a Kenyan service usually draws on English pages from Kenya and beyond. A Swahili prompt may be answered in Swahili built from English sources, or from a smaller pool of Swahili pages that includes Tanzanian content, where vocabulary, place references and prices may not suit a Kenyan reader. Mixed and Sheng prompts are the least predictable: some tools handle them well, others misread slang or answer in a register that sounds foreign.

I test these variations only where your customers actually use them. A clinic, retailer or consumer service may see real Swahili and mixed prompts; a corporate supplier selling to procurement teams may see almost none. Results are recorded by language, so a weak Swahili picture does not hide behind a strong English one.

If Swahili pages earn a place, the engagement runs in English and fluent Kenyan writers from your staff or a local agency own the wording, so it sounds like Nairobi or Mombasa rather than Dar es Salaam. My briefs list the buyer questions, the Kenyan vocabulary to preserve and the English page each one pairs with. Swahili pasted from a translation tool usually reads stiffly to Kenyans and is poor material for an assistant to quote. Google's AI summaries are treated in more depth under AI Overview SEO.

Registered names, phone numbers and other entity facts

Kenyan businesses often reach customers through several identities at once: the name registered with the Business Registration Service, a shorter trading name, a brand used on social media and sometimes a person's name that customers associate with the business. Phone numbers multiply too, with a landline on the old website, two mobile numbers on directories and a WhatsApp line on social profiles. AI tools stitch these fragments together, and when they disagree, answers quote the wrong number or merge you with a similarly named firm.

I prepare a fact record that management approves:

• the registered name, trading names and how each should be written
• branches, towns and counties served, with one address format for each location
• the official phone, WhatsApp and email contacts, and which old ones should disappear
• for regional groups, which entity serves Kenya and which serves neighboring countries

Once approved, the record shapes the about and contact pages, the footer, the Organization schema with registered and alternate names, and sameAs links pointing only at accounts the business itself runs. It also produces a correction list for directories, chambers and associations, supplier registers with corporates and NGOs, and old press pages, each with an owner. Payment details deserve a check as well: where customers pay by M-Pesa, the business name shown on the paybill or till should match the name they expect, which supports trust when buyers verify you. This is entity SEO for a Kenyan setting.

Pages and Kenyan sources that give AI tools better material

AI tools quote content that states facts clearly. For a Kenyan service firm or supplier, a few reworked priority pages usually do more than a long run of blog posts:

• an opening that tells the reader what you provide, who it is for and which towns or counties you reach
• delivery, installation or response arrangements described honestly, including areas you do not serve
• cost drivers explained in shilling terms without quoting figures you cannot honor
• licenses, registrations and prequalification papers that corporate and NGO buyers request, with how to get copies
• questions buyers ask by phone or WhatsApp, answered on the page in plain language

Structured data repeats those facts and is validated against the visible page. I also check that your hosting, firewall or security plugin does not block the crawlers you choose to allow, and that priority pages are indexed in Bing, since Copilot relies on it.

Independent sources confirm what your site says. Coverage in Kenyan business media, chamber and association directories, event and conference listings, partner pages and supplier registers where you genuinely appear all help. I list where you are already mentioned and flag outdated details for correction. I do not buy sponsored articles or links. Because Perplexity shows its sources openly, I also review its citations for your prompts, as explained on my Perplexity search optimization page.

Measuring AI visibility under Kenya's Data Protection Act

There is no rank to track in an AI answer, so I combine several imperfect signals and explain each one:

• the question log repeated on a fixed rhythm, each reply marked for mention, correctness and links
• assistant-referred sessions in a channel of their own, knowing many such visits carry no referrer
• indexing and click data for key pages from both Google and Bing webmaster tools
• a how-did-you-hear-about-us question on forms and in the first WhatsApp reply, stored against the CRM lead
• searches for your brand name, which sometimes climb after assistants start naming you

The WhatsApp question matters in Kenya, where many inquiries never touch a web form. A short, consistent question from whoever answers the chat gives a rough but useful picture of how buyers found you.

All of this involves personal data. Under Kenya's Data Protection Act, forms, cookies, analytics tags and WhatsApp histories each need a lawful basis and a retention period, agreed with whoever handles data protection for you and checked by your legal advisor. Anything not needed for a decision is left out.

Delivery is remote: findings in a shared workbook, video reviews in overlapping working hours and a ranked change list your developer or agency can follow. For branch listings and map results, see the local SEO consultant Kenya page. Broader search strategy sits on the Kenya SEO consultant page, and the Kenya hub lists my other work there.