A Kenyan lending platform broke into the top 3 AI citations in 15 days. Here's how

CodHob (codhob.sg) is a credit platform and white-label lending solution for the Kenyan and broader African market. The product is called CRYSTAMO. It's B2B fintech with a long sales cycle - partners are banks, integrators, and local financial institutions that take weeks or months to decide, not a single ad click.

That long cycle is exactly why AEO (Answer Engine Optimization) matters here. Partners research the market well before any contact with a sales team. If you're not part of that research, you're simply not on the list.

The baseline was stark:

• AI citations for codhob.sg on "credit platform Kenya" queries: 0
• No content hub on the site
• No monitoring set up in Semantica
• Site built on Tilda - fast to launch, but missing E-E-A-T signals, llms.txt, and pillar-page structure, exactly what AI systems look for when selecting a source

Pilot goal: reach 20-40 domain mentions by the end of September 2026.

1. Monitoring - 158 questions

Before writing a single article, we expanded the tracked LLM query set from 61 to 158. That's not "158 mentions" - it's monitoring coverage: the full range of what people are actually asking about the Kenya niche and the brand itself. Without that map, you can't pinpoint where the visibility gaps actually are.

2. Content across three channels - 33 articles in two months

Three channels: the company's own site (codhob.sg/news), LinkedIn, and Medium.

The logic is straightforward - one piece of content, multiple entry points. The site feeds AI crawler indexing. LinkedIn reaches the B2B audience that actually makes decisions. Medium builds an additional hub that AI systems index aggressively.

The top-performing piece - "Key Features of a Credit Platform in Kenya" - accounted for 8 of 35 AI-response citations at the checkpoint. That's over 20% of total domain citations coming from a single, well-structured article.

3. Checkpoint measurement

Between two checkpoints 15 days apart, citations went from 0 to 35 across AI responses. The domain climbed to #3 out of 1,216 in the niche. Share of queries returning a codhob.sg link: 22%. Average answer position: 2.7.

One detail worth noting: the /news section nearly caught up with the homepage in page views - 1,835 vs. 1,963. That's a direct signal the content hub has started carrying its own weight rather than sitting dormant on the site.

Here's a nuance that's easy to miss if you only look at the headline numbers. There's the domain codhob.sg, which AI systems are already actively citing as a source. And there's the brand name "CodHob," which barely appears in the actual text of the answers yet.

Domain ? brand. AI links to the articles but doesn't say the company's name out loud - and that's expected. A system first establishes a source as trustworthy, then starts associating that source with a specific name.

The next step in the pilot: named authorship on articles, links to CBK and the World Bank as trust context, and first-person publications via LinkedIn Articles.

Growth numbers are only half the story. The other half is the real constraints we're working through.

Tilda as a technical ceiling. The platform got the content hub off the ground fast, but it runs into limits on URL structure, author markup, and llms.txt - exactly the signals AI systems weigh when evaluating a source. The fix was a site rebuild.

Google Search Console is still ramping up. First week of data: 13 impressions, 0 clicks. Classic organic search is at a very early stage here - a separate track from the AEO results.

In this pilot, YandexGPT and Yandex Neuro accounted for 25 of 35 citations - 16% of the total AEO breakdown. That's notably higher than you'd expect for an English-language B2B product targeting the African market. For audiences with ties to the Russian-speaking information sphere - investors and partners operating through Russia and the CIS, for instance - that's a distinct competitive edge worth factoring into content strategy for similar niche fintech products.

The goal remains the same: a 20-40 mention corridor by the end of September. We're roughly 45% of the way there, and the citation pace is ahead of schedule.

Already in motion: a next batch of 30 articles (10 from external donor sites plus 20 rewrites of existing content), and further down the road, a dedicated AI-agent workspace to manage the company's content end-to-end.

How long does it take to show up in AI answers from zero visibility?

In this case, the first stable citations appeared 15 days after publishing began, given upfront monitoring and a precise understanding of which queries to target. The typical lag between publication and appearing in AI answers runs 4-8 weeks.

Why monitor 158 queries if you only published 33 articles?

Monitoring is a map of opportunity, not a report on what's already written. Without the full query list for a niche, you can't tell which topics actually get cited, which competitors dominate, and where the open gaps are. Content gets written against specific gaps in that map - not blind guesses.

Why does the domain get cited but not the brand name?

These are two different stages of trust for an AI system. First, the model recognizes specific pages as reliable sources of fact - that's already happened here. Then it starts associating those facts with a specific company name in its answer text - which requires additional signals: authorship, external mentions, links from authoritative sources.

Should you migrate off Tilda right away?

Not necessarily. In this case, quick fixes directly within Tilda produced measurable growth in 15 days with no migration required. A full site rebuild matters for long-term scale - especially for author markup and technical readiness for AI crawler indexing - but it doesn't block early results.

Why did Yandex drive nearly half the citations for an English-language fintech product in Africa?

The exact cause needs more observation, but the working hypothesis is the influence of the Russian-speaking information sphere across a number of adjacent markets and content distribution channels. It's a non-obvious result worth factoring into platform planning for similar niche B2B products with international - not exclusively Western - audiences.