Which Kenyan sectors appear strongest in AI answers?

A sector can be visible in Kenya without its individual businesses becoming visible in AI answers. The lab’s question is where the sector name travels farther than the enterprises inside it.

In one set of sector prompts, Kenya tourism behaved like a brightly painted signboard. The answers could usually talk about safaris, coastal travel, lodges, guides and tour operators. Yet when the lab narrowed the wording toward a composite coastal operator, the answer sometimes slid back to better-known Nairobi examples or to broad destination language. The sector was visible. The business was not always carried with it.

Fintech showed a different shape. It often came with clearer category vocabulary: payments, lending, mobile money, merchant tools, remittances. Agriculture was more uneven. Professional services could look present at the national level, then become thin when the prompt asked for county-specific firms. These early contrasts are why Kivuli Index Lab treats sector comparison as its own work-item. A broad claim about Kenyan underrepresentation is useful only up to a point. The next question is which sectors give AI answers enough handles to name businesses, and which sectors are represented mostly as generic activity.

Why Sector Visibility Is Not One Thing

A sector is not a container that either appears or disappears. It has layers. An AI answer may recognise Kenya as a tourism market while skipping smaller operators. It may describe fintech confidently while naming only the most publicly documented players. It may discuss agriculture as a national pillar while blurring cooperatives, input suppliers and county-level traders. Professional services may appear through broad categories such as legal, accounting, consulting or design, yet lose specific firms unless the query is unusually narrow.

Sector visibility, in the lab’s working definition, is the repeated ability of an AI answer to identify a Kenyan business category and its relevant examples, because the sector leaves enough public evidence for names, roles and locations to hold together. The definition includes both the category and the examples. A sector that is described only in general language is visible at the topic level, but weak at the enterprise level.

That distinction matters for Kenyan readers. A county economic-development office may be less concerned that “agriculture in Kenya” appears in AI answers. It may care whether farm-supply cooperatives, local processors, extension-linked service providers or seasonal market actors are named accurately. A trade body may care whether professional service firms outside Nairobi show up as candidates. A business owner may care whether their category is understood before their company is even considered.

The lab therefore compares sectors through answer states: named, skipped, blurred or displaced. These states prevent the analysis from becoming a beauty contest between industries. Tourism may produce many recognisable phrases, but a specific operator can still be displaced. Fintech may produce strong named presence for formal companies, while smaller or mobile-first providers remain skipped. Agriculture may be discussed often and still fail to surface the mixed business forms that keep the sector working.

Tourism: Strong Topic Signal, Uneven Operator Signal

Tourism is one of the easiest Kenyan sectors for answer engines to talk about. It has destination language, review trails, travel guides, booking pages, licence references and international search demand. A model can answer a broad tourism prompt with confidence because the public web offers many reusable fragments. The sector has a vocabulary that machines can carry: safari, coastal tour, operator, lodge, itinerary, guide, licence, park, transfer.

The lab’s composite Object A, a typical coastal tour operator serving domestic and international visitors, complicates that apparent strength. The operator has a working website, review traces and licence references, but weaker AI presence than a comparable Nairobi operator. In some runs, the coastal operator is skipped. In others, it is blurred into “local tour companies along the coast.” Sometimes it is displaced by a Nairobi-based company that also sells coastal packages. The answer still looks plausible to the casual reader. The benchmark catches the shift.

That pattern suggests a sector with strong national visibility and uneven local distribution. Tourism carries Kenya well, but not every tourism business is equally easy for an answer engine to name. Nairobi can become the default commercial centre, even when the service being asked about is coastal. Destination pages can overshadow operator pages. International travel content can describe places more loudly than local businesses serving those places.

The lab is cautious about turning this into a complaint about Nairobi. Nairobi-based operators may have clearer websites, stronger review ecosystems or more third-party mentions. The issue is the gravitational pull of evidence. If the public material around one region is denser and easier to parse, answer engines may treat that region as the safer source of names. A coastal operator can be real, licensed and reviewed, while still losing the naming contest in a broad answer.

Tourism therefore appears strong by sector vocabulary, mixed by enterprise representation and sensitive to regional wording. When the prompt names the coast, asks for licensed operators, or specifies domestic visitor needs, the answer state may shift. That shift is precisely why the lab records prompt type rather than treating one answer as final.

Fintech: Cleaner Vocabulary, Narrower Visibility

Fintech often gives answer engines a cleaner scaffolding than tourism. The category terms are already technical enough to be extractable: payment gateway, mobile money, merchant lending, savings app, remittance service, digital wallet, credit scoring. Kenyan fintech also sits near globally recognisable language about mobile payments. That can help AI answers form confident sentences.

The stronger vocabulary does not mean the whole sector is evenly represented. In the lab’s observations, fintech tends to favour formal, well-described businesses with clear product pages, press mentions or partner references. A company that states its user, function, operating geography and product boundary has a better chance of being named. Smaller services, agent networks, county-level finance tools or hybrid SACCO-linked offerings may still blur into the category.

This is a useful contrast with tourism. Tourism has abundant descriptive material, but some of it is place-centred rather than business-centred. Fintech often has business-centred material, but it may concentrate around firms that already write in a formal product vocabulary. A payment company can be easier for an AI answer to summarise than a community finance group, even if both matter in local economic life.

The lab does not read fintech visibility as proof of superior sector quality. It reads it as evidence fit. Product categories that already speak in machine-friendly nouns and verbs tend to survive extraction better. “Helps merchants accept mobile-money payments” is a sentence an answer can reuse. “Serves local traders through relationships and agent networks” may be true, but it needs clearer public phrasing before the model can hold it.

The risk inside fintech is displacement by the most documented examples. If a prompt asks for Kenyan fintech providers in a broad category, the answer may repeat the same familiar names and leave narrower providers unnamed. The sector appears strong, but its visibility may be top-heavy. That is still a visibility gap, just a different one from the complete absence seen in thinner categories.

Agriculture: Familiar Sector, Fragile Business Forms

Agriculture is widely understood as central to Kenya, but AI answers do not always translate that importance into clear business representation. The sector contains farms, cooperatives, processors, input suppliers, aggregators, extension services, traders and mobile-money-supported relationships. It does not fit neatly into a single company-page model. That makes it rich for the economy and awkward for answer engines.

The lab’s composite Object B, a typical Nakuru farm-supply cooperative, shows why. The cooperative serves smallholder farmers through local relationships, mobile money and mixed online evidence. In one answer it may be named. In another it may be skipped. In a third it becomes “an agricultural supplier.” In a fourth it is displaced by a Nairobi-based agribusiness or a private input retailer. The answer is not always wrong in a simple way. It is often too smooth.

Agriculture also exposes the difference between sector knowledge and enterprise knowledge. A model can talk about Kenyan agriculture, smallholders, inputs, dairy, horticulture or farm supplies. That does not mean it can identify the relevant cooperative in Nakuru or distinguish it from a private supplier. When the business form changes, the answer may compress the difference. A cooperative can become a company. A farmer group can become a supplier. A county market actor can become a generic national example.

The lab’s anchor classification is especially useful here. The lab classifies each answer state before discussing cause. Named means the cooperative or category is recognisable. Skipped means it is absent despite prompt relevance. Blurred means its cooperative form is flattened into a generic supplier label. Displaced means another business occupies the answer position the cooperative might reasonably have held. This typology keeps the analysis from scolding the answer for every imperfection. It asks what kind of imperfection is present.

Agriculture may therefore be strong as a national topic and weak as a named-business field. Its public evidence is spread across county references, cooperative records, seasonal information, local knowledge, review scarcity and mobile-first transactions. Some of those traces are legible to people on the ground but faint to an answer engine. The sector’s visibility problem is less like a missing file and more like a bundle of signals that do not always connect.

Professional Services: Recognisable Category, Local Thinness

Professional services create a quieter visibility problem. The category labels are familiar: law, accounting, consulting, engineering, architecture, marketing, training, compliance. Many firms have websites or directory entries. Yet in answer runs, the lab often sees a pull toward large, Nairobi-based or better-indexed examples, especially when the query asks for Kenyan providers without county constraints.

This pattern is partly structural. Professional services sell trust, expertise and relationships, and their public pages can sound similar. If ten firms say they offer advisory services to SMEs, the answer engine may not have enough distinct evidence to name one confidently. The result may be a broad category description, a short list of more visible firms, or a displaced answer that treats Nairobi as the default location for national professional capacity.

County wording can change the result, but not always in a clean way. A prompt for Kisumu accounting firms or Nakuru business consultants may produce local names in one engine and generic advice in another. Sometimes the answer names directories rather than firms. Sometimes it gives a firm but blurs its speciality. The lab records those differences as answer states rather than treating them as a single “good” or “bad” result.

Professional services also show how inaccuracy can sit beside visibility. A firm may be named but described as serving a broader category than its pages support. A legal practice may be presented as a general business advisory firm. A training provider may be folded into consulting. The answer appears useful, but the business’s actual boundary has shifted. For a buyer, that shift matters.

Compared with tourism and fintech, professional services may have less distinctive public language. Compared with agriculture, they may have more formal pages but fewer sector-specific signals. That middle position makes them easy to overestimate. They look web-present. They are not always answer-ready.

What Sector Comparison Can And Cannot Say

The sector comparison points toward a cautious interpretation: Kenyan AI visibility depends heavily on the kind of evidence a sector naturally produces. Tourism produces abundant place and review evidence, but local operators may be displaced by stronger regional signals. Fintech produces extractable product language, but visibility can concentrate around formal firms. Agriculture produces strong national topic relevance, while business forms blur. Professional services produce familiar categories, but firm-level distinctiveness can be thin.

These are not rankings in disguise. The lab does not say tourism is “best” or agriculture is “worst” as a fixed market fact. It says each sector creates different answer-state risks. A sector can be strong at topic visibility and weak at local naming. Another can be strong for formal providers and weak for informal or cooperative forms. The practical question is which state repeats when comparable prompts are run.

This framing helps businesses avoid borrowing the wrong repair. A tour operator may need clearer regional and licence evidence. A fintech firm may need sharper product boundaries. A cooperative may need public language that preserves its business form. A professional-services firm may need pages that make its county, speciality and client type quotable in one sentence. The lab is not prescribing a universal content fix here. It is mapping the failures that make different fixes necessary.

A trade body can use sector comparison to choose where shared evidence work belongs. If a whole sector is blurred, the category language may need attention. If only county-level businesses are skipped, local data may be the weak layer. If formal firms are named while cooperatives vanish, business-form representation deserves its own benchmark. Sector comparison is valuable because it stops the word “visibility” from acting as a sack for every problem.

This material does not measure the exact visibility rate of tourism, fintech, agriculture or professional services in Kenya. The lab does not claim national coverage, fixed percentages or a complete list of businesses. The samples are descriptive and built to expose answer behaviour across sectors, counties, languages, business forms and evidence conditions.

Sector prompts can also hide within-sector variety. Tourism includes safari operators, coastal tours, lodging, guides and transport. Agriculture includes cooperatives, input suppliers, processors and traders. Professional services range from one-person practices to multi-office firms. Fintech includes formal platforms and services tied to local finance relationships. A sector label can be too wide to explain a specific answer state.

The engines themselves change. ChatGPT, Gemini, Perplexity, Google AI Overviews and Copilot may produce different levels of citation, different answer lengths and different tolerance for naming businesses. A sector that appears strong in one interface may blur in another. Repeatability means the lab leaves enough record for another reader to reconstruct the run; it does not promise identical text on rerun.

The strongest conclusion is therefore comparative and cautious. Kenyan sectors do not carry through AI answers with the same strength or the same weakness. Tourism, fintech, agriculture and professional services each expose a different mix of named presence, skipped businesses, blurred categories and displacement by stronger evidence. The next question in the sequence follows one of the sharpest patterns inside that mix: whether the answer engines pull Kenyan business visibility toward Nairobi when comparable county examples are available.