98% of Website Visitors Buy Nothing – Here's the Actual Fix

01 /The Concept

The core problem with most websites is that 98% of traffic doesn't convert. Ninety-eight percent of e-commerce visitors don't buy anything. Ninety-eight percent of SaaS visitors don't even sign up. And so on.

As Kenobi argues, one of the biggest reasons is that every website looks the same to every visitor – whether that visitor is a solo founder or a Fortune 500 CEO. When a site tries to speak to everyone with identical language, it usually ends up connecting with no one.

Kenobi's solution is a tool that lets you adapt your website for each specific visitor – swapping generic language for more precise framing that actually has a chance of landing.

All that's required is a single snippet of JavaScript on the site. No database changes, no additional platforms.

Here's the nuance: the startup calls this automatic, but it's more accurately semi-automatic. After the JavaScript is installed, a small personalization widget appears at the top of the page, prompting the visitor to enter their company name or website URL. Only after that does Kenobi automatically rework the page to better match what the visitor's company actually does.

As an example of the effect: instead of a generic headline like "Adapt your website to every visitor," a visitor from a startup news platform might see "Show every visitor the startups that interest them" – with a custom subheadline tailored to that audience's specific lens.

Kenobi claims that visitors who fill in the personalization widget spend three times longer on the site and convert to a target action at twice the rate of those who don't.

The site owner also gets a notification for each personalization event, including the visitor's company name and website – enabling direct follow-up outreach.

The founders went through Y Combinator in winter 2022, but only recently posted about Kenobi's launch in the YC blog.

The reason: for the past four years, they were running Verdn, a donation platform, and only pivoted in May to start building Kenobi.

Kenobi's initial form was a tool that let salespeople auto-generate custom landing pages tailored to individual prospects – so they could send a targeted URL rather than a generic homepage.

Once the AI behind it was generating those pages at the right quality and speed, the founders opened up the same engine for live website personalization. The logical next step – currently still ahead – is automatically enriching anonymous visitor data to surface their company URL, so personalization happens fully without any visitor input.

02 /Why It Matters

Another startup digging in the same direction is Fibr ([related review](/review/chem-tochnee-sootvetstvie-tem-luchshe-prodazhi)), which has raised $3.8M. Its platform is already more advanced – automatically adapting site pages based on the search query that brought a visitor in, the ad they clicked on, their geographic location, or even how many times they've visited. Results include a 10% reduction in bounce rate and a 30% reduction in customer acquisition cost.

Beyond text, there are other powerful personalization vectors to explore.

Supersonik ([related review](/review/kuj-prodazhi-poka-gorjacho)), for example, raised €4.2M in September for a platform that generates real-time personalized product demos – showing each visitor exactly how a SaaS product could work for their specific role, industry, and use case.

The homepage is only the first conversion choke point. The second – and often steeper – drop-off happens at checkout and payment pages.

PrettyDamnQuick ([related review](/review/esli-ne-hochesh-terjat-uzhe-gotovyh-pokupatelej)) argues that generic checkout experiences are a major culprit: the same checkout flow doesn't work equally well for different buyer segments. Personalizing those pages – at precisely the moment when you're trying to close the sale – can move conversion rates meaningfully. PrettyDamnQuick built a checkout personalization platform for Shopify merchants that lifted purchase conversion by 17% and average order value by 19%. The startup has raised $38M, including $25M this year.

Helium ([related review](/review/kak-lego-no-dlja-produktovikov-i-marketologov)) built a paywall and payment page personalization platform for mobile apps and web services, raising $7.9M – including $5.4M in August. One client reports a 77% increase in subscription starts after enabling Helium.

Some of these platforms are built around automation; others around manual experimentation. But the distinction is converging – manual experiments will increasingly be run by AI, and automated systems will increasingly expose manual controls. At a conceptual level, this is all the same idea at different stages of maturity.

03 /Opportunities

The macro trend is clear: automatic, per-visitor personalization of websites and apps.

Right now, the workflow is: build a "generic" site first, then layer on personalization rules. Which raises the obvious question – why bother with the generic intermediate step at all? Why not let the site generate itself according to rules and algorithms from the start?

That line of thinking points toward an "ideal" future platform: not a static page builder, but a knowledge base with a set of initial rules and target metrics – from which the CMS assembles the right page, on the fly, for each visitor.

It would also monitor performance and run experiments continuously, adding new rules to the knowledge base to progressively improve page generation and metric outcomes.

So in a sense, today's personalization tools are all transitional scaffolding on the way to that "ideal" platform.

And the most ambitious direction of travel is building that platform.

First versions are already buildable – all the components exist. As AI capabilities improve, these platforms will keep getting sharper – until they fully displace the static builders where sites are made today.

The platform is buildable today – all the components exist. The practical starting point is a single vertical: pick a category (e-commerce, B2B SaaS, or professional services), assemble the minimum viable knowledge base, and prove the personalization loop works before broadening scope. That's the only sequence that avoids getting lost in abstraction.