Ugly Pages Win: Clarity Beats Beauty

Clarity converts better than polish. At a healthtech company, an outdated landing page beat every polished redesign our team A/B tested, converting at 10–11% against the redesigns’ 7–8%. Our read is that the plain page answered a visitor’s first questions faster: what this is, whether to trust it, and what to do next.

The Redesigns That Kept Losing

By our own design standards, the company’s main landing page was ugly. The design was outdated, the layout was clunky, and the page felt years older than the product it was selling. The obvious move was to fix it.

So we invested in proper redesigns, with clean layouts and modern visuals: the kind of page a team shares internally as the new standard. Each version looked far better than the original. Each version lost.

Test after test, the ugly original won. It converted at 10–11%, while the redesigns converted at 7–8%. That is a gap of roughly three percentage points, and a gap that size adds up across all the traffic the page receives.

It was frustrating, and a little embarrassing. We kept the ugly page live anyway, because the test results, not our taste, decided what shipped.

Why Did the Ugly Page Convert Better?

As we read the results, visitors do not judge a page the way designers do. They are not asking whether it is beautiful. They want answers to three questions:

  • Do I understand what this is?
  • Do I trust it?
  • Do I know what to do next?

The original answered all three quickly. It was blunt: no clever copy and no visual flourishes competing for attention, just a clear offer and an obvious next step. The redesigns, for all their polish, added friction. They had more elements to process, more choices about where to look, and a slightly longer path to the call to action.

Clarity is a conversion mechanism. When beauty competes with clarity, beauty becomes a liability.

How Should You Test a Redesign Before It Ships?

  • Run it as a variant, not a replacement. Keep the current page live and make the redesign earn its place in an A/B test.
  • Judge it on the conversion that pays. Measure sign-ups, booked calls, or purchases, not clicks, time on page, or internal opinion.
  • Check the first screen against the three questions. If a hero image, a slider, or a secondary link competes with the main call to action, cut it or move it lower.
  • When a redesign loses, keep the original. Then test one change at a time to learn which element helped and which one hurt.

We treat every redesign as a hypothesis until it wins a test. That applies to landing pages and ad creative as much as to each step of a paid acquisition funnel. For how we apply the same rule to brand creative and app-store assets, see AI and brand storytelling: what to automate, what to test.

What Does This Mean for Pages That AI Agents Read?

The audience for a page now includes software. On June 25, 2026, Worldline, Crédit Agricole, and Mastercard announced what they called France’s first agentic payment in production: an AI agent shortlisted festivals and started a ticket purchase on Weezevent, and the payment went through only after the customer explicitly approved it (Worldline, The Paypers).

An agent doing that job has to match the facts it was given, in this case a budget, an event type, and a location. Polish does not help it if those facts are missing or hard to find.

The platforms’ own guidance puts the emphasis on clear, consistent facts. OpenAI’s merchant docs call for a regularly refreshed product feed with descriptions, pricing, inventory, and fulfillment options, so ChatGPT can surface products accurately in search and shopping. Google says pages need no special AI files or markup to appear in AI Overviews or AI Mode. Its advice includes keeping important content in text form and making structured data match what the page shows.

The discipline that won with visitors carries over. State the offer, the price if there is one, and the next step in plain text, and keep any product feed or structured data consistent with the page. The principle has not changed: clarity wins. The audience is just broader.

Sources: Worldline, “Worldline, le Crédit Agricole et Mastercard réalisent la première transaction de paiement agentique en France” (in French), press release, June 25, 2026. The Paypers, “Worldline, Crédit Agricole, Mastercard complete first agentic payment in France,” June 29, 2026. OpenAI, “Key concepts – Agentic Commerce,” accessed October 23, 2026. Google Search Central, “AI features and your website,” updated December 10, 2025. The landing-page conversion rates come from our team’s own A/B tests.

Updated October 23, 2026: we rewrote this post, re-checked its facts, and added sources.

Picture of Alex Piliavsky

Alex Piliavsky

Alex Piliavsky is a co-founder of Alchemy Avenue, an AI-driven growth and lifecycle marketing agency. He helps businesses build growth that actually lasts instead of chasing quick wins. With over a decade bouncing between countries and industries, Alex learned the hard way that real growth isn’t about luck! It’s about patience, systems, and boring stuff like clarity. He specializes in strategic digital marketing and automation, turning chaotic operations into predictable growth machines.
Picture of Alex Piliavsky

Alex Piliavsky

Alex Piliavsky is a co-founder of Alchemy Avenue, an AI-driven growth and lifecycle marketing agency. He helps businesses build growth that actually lasts instead of chasing quick wins. With over a decade bouncing between countries and industries, Alex learned the hard way that real growth isn’t about luck! It’s about patience, systems, and boring stuff like clarity. He specializes in strategic digital marketing and automation, turning chaotic operations into predictable growth machines.

Related Articles

Hands moving gold coins from several small jars into one larger jar

Cut Fast or Burn Budget: A UA Lesson

Scaling UA depends on cohort-level measurement and cutting losing channels fast. In our Eyecon case study, monthly spend grew from about $10,000 to roughly $600,000, and payback fell from over a year to 2–3 months. Here are the rules.

Read More
Man in a server room frowning at a printed attribution report with one line circled in red; a tablet shows a pie chart

Google’s AI Max Can Break Your Attribution

Google’s AI Max for Search can send a click to a different page than the final URL you set. Here’s Google’s guidance on tracking templates, URL exclusions, and testing, plus our three-step checklist for keeping CRM attribution intact.

Read More
Man in a dark office marking a large printed spreadsheet with a red pen, a laptop open to a product roadmap

User Demand Isn’t a Business Case

At a caller-ID app, around 25% of surveyed users said they’d want VoIP calling. We didn’t build it: a flat “unlimited” price couldn’t cover per-minute costs. Here are five checks to run before you build.

Read More