How One Paragraph on Wise’s Website Ended Up in Millions of AI Recommendations

Most people have heard of Wise even if they’ve never transferred money abroad. It’s one of those fintech brands that crossed over into mainstream name recognition, the kind of company people reference when they talk about how banking should work, even if they’ve never actually used it to send a payment.

We came across something interesting about Wise recently while running an AI visibility check for one of their competitors. AI assistants across the board put Wise at the top for money transfer queries, which is mostly a reflection of strong branding and years of good marketing. But the way they recommend Wise is what caught our attention: ChatGPT doesn’t just say Wise is a good option. It quotes specific numbers. “The mid-market interbank rate.” “6.58 EUR for a 1,000 EUR transfer.” That kind of precision means AI is pulling from a concrete source, and when we traced it back, we found it on a page that most Wise users probably scroll past without reading.

The page nobody reads

Wise has currency conversion pages for dozens of currency pairs. Most visitors come for the calculator: type in an amount, pick two currencies, get a result. The calculator does its job and the visitor moves on.

But above the calculator there’s a small block of text:

100% transparent pricing. Banks and other providers inflate the mid-market exchange rate to hide fees and make you pay more. Not Wise. You’ll always get our stand-out rate, like the one on Google.

Below that, a comparison table with exact fees broken down by provider. One line in that table says it costs 6.58 EUR to send 1,000 EUR to a US dollar account.

About eighty words and one small table. The kind of content that probably took someone half an hour to write, including the time it took to look up the competitor fees for the table.

And that’s almost exactly what ChatGPT says when someone asks it about sending money abroad. The mid-market rate language, the specific fee, the comparison with banks. AI found that paragraph, decided it was useful, and now cites it to millions of people every day.

The competitor’s page had the same calculator and nothing else

The fintech company we were checking had its own conversion pages, structured the same way, one per currency pair. The calculator was well built, the design was clean, and a visitor could get everything they needed from the page in about ten seconds.

The difference was that there was no text around the calculator. No headline about pricing, no explanation of how rates work, no comparison table. Just the widget.

For a person visiting the page, that’s perfectly fine. The calculator does the talking. But for an AI assistant trying to decide whether to recommend this company to someone asking about money transfers, the page offered nothing. No claims to evaluate, no numbers to cite, no language to borrow for a recommendation.

The same product, a similar page, a completely different outcome in AI results, and the gap between them was a paragraph of text.

Why calculators are invisible to AI

This is the part of the story that extends well beyond Wise and fintech.

AI assistants don’t browse the web the way people do. When a person visits a currency conversion page, they type a number into the calculator, pick a currency, click a button, and read the result. AI can’t do any of that. It can’t type into input fields, it can’t click buttons, it can’t trigger JavaScript, and it can’t read dynamically loaded results. What it can do is read text that’s already sitting on the page: headings, paragraphs, table data.

A page with a great calculator and no surrounding text is a useful page for humans and a blank page for AI. The calculator might be doing exactly the same math as Wise’s calculator, converting the same currencies at the same rates, and AI would have no way to know.

This is true across the entire web. Mortgage calculators, shipping cost estimators, BMI tools, ROI calculators, tax estimators, loan comparisons. Millions of pages built around interactive tools that work beautifully for the person using them and give AI nothing to work with. Every one of these pages could become a source that AI cites in recommendations, and all it would take is a few sentences of text describing what the tool does and what results a typical user would see.

Eighty words and ten minutes

Adding a headline and a short descriptive paragraph to an existing page is about ten minutes of work. The people who built the product already know what to write. For a currency conversion page, that means a line about the pricing approach, a sentence about how the exchange rate is determined, and a small table with a concrete example.

The only thing that matters is specificity. “Competitive rates” gives AI nothing to quote. “A flat fee of 0.43% on EUR to USD transfers” gives it a fact it can cite in a recommendation and link back to the source. Wise got this right, whether by intention or instinct, and the result is that AI assistants across every major platform parrot those eighty words back to millions of users who are trying to figure out how to send money to another country.

The whole industry is looking for complex theories about what drives AI visibility and generative engine optimization. Sometimes the answer is just that somebody wrote a paragraph on a page, and their competitor didn’t.

---

Reachd.ai tracks how AI assistants recommend products and services across ChatGPT, Gemini, Perplexity, and other platforms. The Wise example in this post came from a real visibility check where missing page content was the first thing our system flagged.