Ari Yablok, Head of Brand at Island, has been having the same argument for five years around what happens when data says one thing and brand instinct says another, and he's landed somewhere more useful than "it depends." Data and brand are built to solve different problems, and the mistake most companies make is asking one to do the other's job.

As part of our ongoing series that included our talk with Anthropic's Corey Moen and, more recently, a conversation with Stu Ross from OFF+BRAND, I sat down with Ari to talk about how Island's brand has evolved since we started working together. 

The timing couldn’t have been better: we'd just wrapped three years of data on what that evolution produced. The conversation and the numbers turned out to make the same argument from opposite directions.

Why brand can’t wait around for data to catch up

Island's founding challenge wasn't just building an enterprise browser. It was explaining what an enterprise browser was to three different audiences who had never had a reason to think about a browser as anything other than a place where work happens to occur. Ari's team couldn't point to a category playbook, because there just wasn't one.

Data is what everyone talks about. And what it can do well is tell you when a customer journey is broken: when visitors are leaving confused, when a page isn't answering the question someone came to your site with. What it can't do however, is validate a brand decision that, by definition, hasn't been tried before. Ari draws a clean line between those two categories and uses each one for what it's built for.

A brand decision that hasn't been proven yet is exactly the kind of decision data can't validate. Data is a record of what already worked somewhere else, which is precisely why the best brands can't wait for it before doing something no one else has done.

Why Island stopped doing quarterly rebrands and started doing continuous ones

A few years into building Island's brand, Ari's team hit a wall that a lot of fast-growing companies hit: the traditional rebrand cadence of stop, think for three months, build for three months, publish, etc. became impossible once the product itself was shipping a new capability every few weeks. Waiting for a clean, finished version of the story meant publishing a story that was already out of date.

So they split the brand into what stays fixed and what's allowed to move. 

Island kept a small set of durable principles as the foundation, including its visual language, its tone, and one rule: never talk about work in the language of danger. (No fear-based framing, because Island's whole premise is that if you build work the right way from the start, you're not in danger in the first place.) Everything else from the messaging to the site structure, to which industries get their own dedicated conversation, was free to evolve month to month.

That's the model Island's brand design lead recently quipped about as the “continuous brand launch,” and it's a useful reframe for any company whose product moves faster than its marketing team can responsibly rebrand around it.

AEO is turning websites into a judgment call

The most interesting part of the conversation was Ari's read on what AI answer engines are doing to the website's job description. For years, a company website had two functions: tell people who you are, and answer their questions. Ari's argument is that the second function is quietly migrating to ChatGPT, Perplexity, and Claude, and it isn't coming back.

That shift makes the website's job more specialized. If a buyer already has their factual questions answered before they ever click through, then the visit itself becomes a judgment call: does this company feel like a partner I'd want to work with? Ari's team has responded by leaning harder into video, in-person events, and anything that reveals something about Island that a language model can't summarize into a bullet point.

The practical implication for anyone building a Webflow AEO strategy is this: don't treat AEO as a reason to strip personality out of a site in favor of pure extractability. Treat it as a reason to be more deliberate about the parts of the site that AI can't answer on your behalf.

The proof is in three years of Island's own numbers

Ari's philosophy is compelling on its own, but philosophy is easy to go on about and hard to verify. 

Here's what happened on Island's site over the three years Edgar Allan has worked alongside Island’s brand and product teams:

Impact & Results Overview

_{First-touch leads: HubSpot, first-touch attribution on the production domain, excluding offsite syndication.}

^{*Compounded growth from 2023 - 2026}

Edgar Allan's own ag-nts data, which tracks AI agent and answer-engine traffic on client Webflow sites, shows the organic search line in this table wasn't a one-time spike; it built steadily as AI-driven referral traffic to Island's site grew.

Island kept its brand principles fixed while it let the website underneath them get rebuilt and tested constantly, running a brand clarity assessment at the level of principle and keeping execution moving underneath it.

Performance data never overrode the brand decisions that made Island recognizable in the first place. The 457% jump in conversion rate happened because the site kept getting easier to use without the brand ever blending into the category it was trying to stand apart from. This is the core of what we call Visibility Engineering and Optimization: brand and performance work run as one continuous loop.

What changes when fewer, better visitors show up.

As AI answer engines absorb more of the top-of-funnel research, the traffic that reaches a website skews toward higher intent. I.e., someone who shows up after already asking an AI system about you has more intent to connect or buy from you than someone who clicked a banner ad. That means the visit has more gravity, and a clunky last mile now costs more than it used to.

Ari's estimate on this is blunt: a polished, on-point experience at the point of decision can be the difference between a 50/50 shot at winning an RFP and a 65/35 one. That's a real business outcome carried entirely by conversion rate optimization, the least glamorous part of the work and, according to Island's own numbers, one of the parts of the work with the biggest return.

If you're running a similar playbook, that's the sequence worth borrowing: get the brand principles fixed enough that everyone in the company can recognize them without a style guide, then treat everything downstream, the site, the CRO program, the AEO structure, as a live system you keep tuning rather than a project you finish.