What do 1,200 Kenyans tell a chatbot that they won’t tell another human being? Many people think generative AI was built to draft emails and automate spreadsheets. But nearly half of the Kenyans surveyed in a recent study are using it as a 24/7 digital confidant, sharing personal anxieties, relationship crises and unspoken secrets. In this edition of Delve Into AI, ADONIJAH NDEGE examines new research by Nendo. This Nairobi-based research consultancy reveals how chatbots are quietly crossing the line from productivity tools to emotional safe spaces, and why seeking life advice from code carries subtle new risks.
This finding offers a striking measure of how rapidly generative AI is evolving beyond its initial promise as a mere productivity tool.
Chatbots were designed to help users write emails, summarise reports, search for information, and automate routine tasks. While fulfilling those functions, an increasing number of users are now treating them as digital confidants.
In a survey published by Nairobi-based research consultancy Nendo, 46.3% of respondents admitted using AI for matters kept secret from others. Notably, 72% had turned to AI for emotional support or to navigate personal circumstances, with 54.1% doing so on a weekly basis.
These figures do not imply that nearly three-quarters of the broader Kenyan population are seeking formal therapy from AI. Nendo’s survey sample comprised 1,200 digitally literate, AI-aware individuals across 18 counties—a demographic noticeably younger, better educated, and more active online than the average Kenyan adult.
Of the participants, 1,161 completed the detailed AI module. Consequently, the findings are not nationally representative; nor did the survey classify every personal query as clinical therapy or evaluate the safety and factual accuracy of the responses provided.
What the dataset vividly illustrates is that among high-frequency users, a technology deployed for task automation is encroaching upon emotional territory historically reserved for human relationships.
This behavioural shift carries implications far beyond the debate over AI-driven job displacement. The more immediate transformation lies in the intimate nature of the interactions citizens are now willing to delegate to algorithms.
A digital confidant
Chatbots uniquely encourage personal disclosure because natural conversation serves as their primary interface.
Search engines demand that users compress complex problems into keywords. Traditional software obliges them to navigate menus and commands. Generative AI, by contrast, invites users to articulate thoughts in much the same way they would converse with a peer.
This structural distinction becomes profound when the underlying query is deeply personal. Searching for advice on a troubled relationship yields static web pages; engaging with a chatbot allows users to detail what transpired, answer clarifying questions, and seek tailored guidance on next steps.
While the underlying engine remains code, the user experience feels increasingly social and empathetic.
Nendo’s survey respondents demonstrate how swiftly this boundary dissolves. One male respondent aged 25 to 34 in Nyeri—a town 150 km north of Nairobi—reported using AI to draft meeting minutes, locate holiday rentals, and request emotional support.
These tasks once inhabited distinct domains: administrative duty, leisure logistics, and psychological well-being. Generative AI consolidates all three behind a single prompt box.
Other participants detailed using AI to draft business plans, seek financial guidance, complete academic coursework, translate regional languages, structure lesson plans, compose cover letters, and author research papers.
The underlying significance is that frequent engagement for mundane tasks renders the transition toward emotional intimacy practically frictionless. A user who has spent months asking an LLM to clarify complex concepts or refine documents requires no new platform when their inquiry pivots from “rewrite this email” to “how do I handle this crisis?” They simply continue typing.
A machine that listens
While the Nendo study does not definitively isolate why respondents confide in AI, the architectural mechanics of conversational platforms offer clear clues.
Generative chatbots are accessible 24/7. They demand no appointment scheduling, nor do they require emotional reciprocity. There is no risk of visible judgement when asking awkward questions, no mid-sentence interruptions, and zero social friction when revisiting the same dilemma repeatedly.
While these attributes were engineered for customer service efficiency, they prove extraordinarily well-suited to personal confession.
This trend is most pronounced among younger demographics. A striking 76% of respondents aged 18 to 24 and 81.4% of those aged 25 to 34 reported relying on AI for emotional support or personal guidance. This proportion dropped to 63.8% among 35- to 44-year-olds, falling below 50% for users aged 45 and above. Gender variance was also evident: women were more inclined than men to report emotional or personal AI use, at 76% compared to 68.4%.
Because the study does not unpack the root drivers of these disparities, concluding that younger demographics or women are actively replacing human relationships with AI remains premature. Nevertheless, the age correlation is vital: the cohorts establishing the deepest intimacy with AI are precisely those who will coexist with the technology for decades to come. For these digital natives, consulting an algorithm for personal counsel may soon feel as ordinary as querying a search engine does today.
Intimacy changes the risk
Public debate surrounding generative AI has overwhelmingly focused on hallucinated facts and disinformation. However, intimate personal usage introduces a far more subtle vulnerability: the real-world fallout from flawed AI output varies dramatically based on context.
An inaccurate document summary can be edited, a poorly drafted email revised, and a fabricated statistic cross-checked against primary sources. Evaluating personal life advice, however, is far more complex.
Nendo revealed that 57.9% of respondents placed high or complete trust in AI for career decisions. Similarly, 55.7% expressed significant trust regarding legal issues, 53.5% for financial choices, and 50% for health matters. By contrast, only 39.9% held equal trust in AI for family-related guidance.
Remarkably, respondents demonstrated greater willingness to trust algorithms with legal dilemmas than with domestic relationships. This may stem from the perception that legal and financial domains possess objective, rule-based answers, whereas family dynamics depend on nuanced emotional intelligence—a hypothesis the survey did not explicitly test.
Yet legal, financial, and medical decisions represent the exact domains where polished but inaccurate advice can trigger severe consequences. This danger is heightened by the mechanics of conversational interfaces. A chatbot does not merely fetch raw links; it synthesises information into an authoritative narrative tailored directly to the inquirer. Textual fluency thus becomes mistaken for institutional authority.
What does the machine truly understand about Kenya?
The risk multiplies when an AI system dispensing advice lacks genuine contextual understanding of the socio-cultural landscape in which the user operates. Here, Nendo identified a striking paradox.
While 72.5% of respondents believed AI models broadly understood Kenyan daily life, 25.1% reported receiving recommendations completely at odds with local culture or values. Furthermore, nearly three-quarters indicated they would increase their AI usage if models better comprehended indigenous languages and cultural nuances.
Linguistic capabilities remained a clear bottleneck: only 34.9% rated AI responses in Kiswahili or Sheng as “very good” or “excellent,” whereas 35.8% rated them as merely “fair” or “poor.”
Bridging the language gap, however, is only the superficial challenge. A model can easily translate a local phrase without understanding the social structures underpinning it. It can define a chama as an informal savings club, yet fail to grasp the complex social obligations, communal trust, and informal dynamics governing member behaviour within one.
This limitation becomes critical across family, financial, and career guidance, where logically sound Western advice may prove dangerously unsuited to local economic realities. As Nendo aptly concludes, true localisation requires AI systems to know “the difference between Kenya and a stereotype.”
True scale demands moving beyond surface-level integrations to robust execution. We’ve filtered the noise out of Moonshot 2026, optimising the conference strictly for high-calibre connections between startup founders, global financial operators, enterprise leaders and individuals rewiring Africa’s technical frameworks. Get 20% off Early Bird tickets for a limited time.
