Redesigning for humans, search, and AI: What worked and what didn't

Not because the old one was broken — it was, honestly, beautiful. Animations that made the page feel alive, content that revealed itself as you scrolled, interactions that rewarded curiosity. It was everything a modern website was supposed to be.

The problem: a lot of what made it great, the machines couldn't see.

That realization is what set the brief for the redesign. And it's the same realization that's quietly reshaping how smart marketing and web teams think about their sites right now.

We pulled back the curtain on this full redesign over on Flow TV. Watch it free.

The homepage started with one job

For most of the web's history, a homepage had one audience: the human being looking at it. You optimized for clarity, conversion, and brand. You made it look good, feel right, and tell your brand’s story the way you intended. You tested headlines and swapped hero images and debated button colors.

Then search engines came along, and you added a second audience — crawlers that needed structure and signals to understand what you were talking about. Most teams got pretty good at that balancing act through SEO.

But something changed. AI entered the picture, and it created an entirely new audience with different rules and behaviors.

Now your homepage has three jobs.

1. It has to represent your brand visually to a human. 
2. It has to signal to AI who you are and who you're for. 
3. And it has to convert visitors who are arriving more informed than ever because AI already gave them a summary of you before they clicked.

The perception problem hiding underneath

Here's what makes this harder than it sounds: AI isn't always getting the summary right.

When we first started checking how AI portrayed Webflow, we were consistently described as a “no-code website builder.” While not inaccurate with how Webflow started, this framing is a relic compared to where our platform is today and how we define Webflow as the agentic web marketing platform for high-performing brands on our site.

AI systems synthesize answers from signals across the web: your homepage, third-party reviews, Reddit threads, YouTube, Wikipedia, publications that have written about you. The homepage is one input into a much larger consensus. But it's a critical one. And if the structure of your site doesn't clearly communicate who you are, what you do, and who you serve, AI will fill in the gaps with whatever it finds most legible.

That's the perception gap. And it's the reason the three-job brief got so complicated.

When design meets discoverability

The redesign started with a principle that sounds obvious in retrospect: content present on the page from the start.

Our previous site leaned heavily on movement. Content that revealed as you scrolled. Elements that animated as you engaged. It felt alive in the best way…for a human. 

But LLM crawlers don't scroll. They don't execute JavaScript or watch hero animations. They read the code underneath. If your most important content only appears after an interaction fires, as far as the machine is concerned, it might not exist at all.

So we simplified. 

Fewer reveals, fewer gates. More structure, more semantic clarity. We rewrote copy to be outcome-oriented moving from what we do to what we help teams accomplish. When an LLM is synthesizing an answer about what a platform does, outcome language maps much closer to how people actually ask questions.

What surprised us: the changes that helped AI also helped humans. A more direct page. A cleaner read. Less noise between the visitor and what they came to find.

We also added schema markup and more structured content — the kind of explicit signals that help AI systems understand your positioning, your audiences, and your differentiators with confidence rather than inference.

Not everything worked the way we expected. An llms.txt file had less impact than anticipated while refreshing content, restructuring page architecture, and answering real questions clearly had more.

Measurement is harder than tactics

The honest answer on tracking is that this space doesn't have clean answers yet.

We think about it in three layers. 

Visibility: how Webflow shows up in AI-generated responses across a tracked set of priority prompts, and whether the brand sentiment and messaging reflect who we actually are. 

Citations: whether we're being referenced as a source, not just mentioned. 

Business impact: whether AI-referred visitors are signing up and converting.

That last one matters more than it might seem. Early evidence shows AI-referred visitors convert at a significantly higher rate than other sources (e.g., 37% vs. 14% for non-brand SEO). They arrive informed, with context, often already past the comparison phase. Your site isn't educating them from scratch — it's validating a decision they're already leaning toward.

That changes what the homepage needs to do. And it changes how you measure whether it's working.

The loop has to close

The thing we keep coming back to: insight without execution isn't a strategy. If you're tracking how AI portrays you, try changing the structure, updating the schema, or testing a new framing.

What we built on our own site, and what we're now making available in Webflow, is a closed loop. You see how AI sees you. You get recommendations. You make changes. You measure what moved. And you do it again.

That's what an AI-era website actually looks like. It’s not just a page that's been optimized for AI search, but a system that keeps getting better because the feedback is built in.

The window to define this for your category is narrow. If you're not shaping how AI describes you, someone else — or the agent — will do it for you.

Want to see the full picture? Watch the complete session, including a live demo of the new Webflow AEO tools.

‍