Background
Arkose Labs protects some of the world’s largest platforms from bots, fake account creation, and account takeover. Their buyers are CISOs and heads of fraud at enterprises where a single attack costs real money, and those buyers increasingly open their research by asking an AI assistant which vendors are worth a look.
Arkose came into this from strength. Models already knew the brand and treated it as credible, which is not a given in security, where AI systems are cautious about recommending anyone. The site was clean and permissive to crawlers, and Arkose publishes original threat research — exactly the kind of material models prefer to cite.
The problem was placement. On many of the prompts their buyers were asking, Arkose surfaced around tenth. Most AI answers name five to eight vendors, so tenth is the equivalent of page two. Underneath that sat a description problem. Arkose can accurately be called bot mitigation, fraud prevention, account security, or challenge enforcement, and when all four run together without a hierarchy, a model has a harder time deciding which question Arkose is the answer to. Competitors with simpler stories were easier to summarize, so they got summarized.
All of this was about to get harder, because Arkose was mid-rebrand into a platform company. When a business changes what it sells, AI keeps describing the old version of it for months.
What they did
Arkose’s marketing and security teams know their category and their buyer at a level very few companies match. What they needed was a precise reading of how AI systems were actually describing them, and where that sat against the company they were becoming.
ChatRank built that picture. We modeled 4,500 personas from Arkose’s own ICP data, enriched with employment history, geography, and the software those companies genuinely run, then simulated thousands of buying conversations to see which vendors came back and which sources the models pulled from. The citation analysis turned strategy into a work order. Rather than guessing, the team could see the specific pages winning the answers they wanted and why.
From there the work ran on two tracks. ChatRank wrote the core pages of the new site — including the homepage, platform page, and resource center — so the new positioning read cleanly to a model on first pass. Alongside that came monthly articles built around real buyer questions: how attackers bypass MFA, how to stop fake account creation at scale, and how to let AI shopping agents through without letting bots in. Arkose’s own experts reviewed every piece, and that review is what kept the work technically credible enough to be worth citing.
What happened
The new site launched with the AI-search work built into it rather than bolted on afterward. Every core page shipped answering a real buyer question in its opening lines, naming the buyer and the environment explicitly, and using one consistent set of words for what Arkose does.
That last piece mattered most. The rebrand gave the team a chance to settle the category question once and carry the same language across the homepage, the platform page, and the resource center, so a model reading any of them comes away with the same story.
Most companies go through a rebrand and then spend the following year discovering that AI is still describing the business they used to be. Arkose went into theirs knowing exactly how models were reading them and what to change, which turned the relaunch into the moment the new positioning landed rather than the moment the confusion started.
The part that matters
A rebrand is already expensive. The last thing a company wants is to spend months rebuilding its positioning, messaging, and website only to have ChatGPT keep describing the business it used to be. That is especially important in enterprise software, where a CISO researching a category may ask an AI assistant which vendors to evaluate before they ever visit a vendor website. If Arkose’s new positioning isn’t present in that answer, the sales team may never get the chance to explain it.
The goal wasn’t simply to get Arkose mentioned more. It was to make sure that when AI mentioned Arkose, it described the company Arkose was becoming.
Why it works
Arkose was never fighting for credibility. Models already trusted them. The work was making the new story easy for a machine to understand and repeat, which comes down to naming the buyer early, answering the actual question first, creating a hierarchy between products and capabilities, and using the same language everywhere.
If your company is heading into a rebrand, a repositioning, or a new site, this is the cheapest moment to do it. Fixing how AI describes you while the pages are already being written costs a fraction of what it costs to go back and unpick it a year later.