How Webflow Built an AEO Assessment Tool in a Weekend
Guy Yalif built the prototype of Webflow’s AEO assessment tool over a weekend with AI. It now runs thousands of assessments a month and helps CMOs deal with a big shift: traffic from AI answer engines converts 6x better than traditional search, and most marketers have no plan for getting it.
This is what it looks like when AI lets operators build things. Someone with a good idea and real expertise in the area puts together a working prototype over a weekend, gets people inside the company using it, and then works with the team to turn it into a production tool. Engineers do get faster with AI, but the bigger change is that subject matter experts can now build the tools that solve their own problems, and run them for lots of people.
The Problem Worth Solving
“AEO started coming up in every conversation,” Guy explains. He’s Webflow’s Chief Evangelist and spends his days talking with marketing leaders, and more and more of them were stuck on the same question. How do you show up in ChatGPT, Perplexity, Gemini and Claude? Are the tactics for LLMs really different from SEO?
More is riding on this than many people think. Organic traffic has fallen 10-25% for many companies since large language models took off, but the traffic that comes from LLMs converts much better. At Webflow, 8% of self-serve signups come from LLM traffic, and that traffic converts 6 times better than unbranded search. Industry data from Ahrefs and SEMrush puts conversion rates 4x to 23x higher.
The reason is simple. A visitor who comes from an AI answer engine has already had a conversation. They asked longer questions (23 words on average, against 4 words in traditional search), narrowed down what they need over several rounds, and did a lot of research. By the time they reach your site they’re further down the funnel and ready to buy.
The catch is that LLMs rewrite the brand messaging you worked so hard on. They might boil your positioning down to a feature list, state it wrong with full confidence, or leave you out altogether. In traditional SEO, search results showed your words as you wrote them. AEO needs a different approach.
The Solution: An Actionable Assessment Framework
The answer was an assessment tool. You type in your domain name and Webflow looks at your site from the outside, crawling both the site itself and your presence elsewhere (LinkedIn, Reddit, Quora and so on). Then it sends back a personalized assessment with your maturity level in each category and specific next steps.
The framework scores AEO maturity in four categories:
- Content: Going from keywords, to answering questions, to staying relevant through personalization
- Technical: On-page SEO still counts, and now you also give LLMs more context through schema markup, sitemaps and site speed
- Authority: Backlinks still help, but being mentioned widely and offering experiences people engage with help more
- Measurement: Track mentions, share of voice and sentiment instead of leaning on keyword rankings
Each category has five maturity levels. CMOs can use that rubric to plan their team’s progress, and practitioners can use it to decide what to work on next. “There’s an awful lot behind this,” Guy notes.
How It Was Built: A Weekend Prototype to Production Pipeline
Guy has founded four tech companies. He’s mostly worked as a CEO and in marketing, but he’s more technical than most people in those jobs. Even so, he calls what he does now “coding in English,” and that’s how he built the first version over a weekend.
The V1 Process:
- Spent hours talking with ChatGPT and Claude about how the assessment should be structured
- Wrote a 600-line prompt document telling the LLM to crawl sites, analyze the data, calculate metrics and write summaries
- Tied it all together with Zapier (because “the amount of extra time it would have taken to figure out the API to Slack, the API to Marketo, to go access Google docs…that probably would have caused this to never happen”)
- Generated PDFs automatically from Google Docs templates
- Offered it through a web form and through a Slack channel where the sales team could ask for assessments
How did it go? “It was genuinely useful to people,” Guy says. So they kept improving it, and brought in engineers to turn the prototype into something production-grade.
What Production Looks Like Now:
- 87 pages of prompts (up from the original 600 lines)
- Separate assessors for each category: content, technical, authority and measurement
- An orchestrator that runs the logic, decides whether new analysis is needed, starts assessments in parallel and combines the results
- A Webflow CMS integration that builds a custom page for each assessment
- Connections to Marketo, Salesforce, Slack and batch spreadsheet processing
Guy is refreshingly honest about it: “When I started, I thought, hey, this is straightforward.” In practice it took constant rework. He split up prompts that produced “gobbledygook,” connected new data sources, and kept refining how the pieces talk to each other and share memory.
Coordinating all of that took real engineering skill. The main lesson still holds: Guy could spend his time on the part that created value (the assessment logic) instead of on delivery plumbing, because tools like Zapier and Claude connected everything.
The Results: 3,500 Assessments and Counting
In just three months the tool has run more than 3,500 assessments. It has become a useful part of sales conversations and has helped the market as a whole get more fluent in AEO.
The most telling detail: when Guy runs webflow.com through the assessment, Webflow scores a 3 out of 5. “It’s in that middle range,” he admits with a laugh. AEO is early for everyone, and that’s good news. In SEO you fight hard to move up a single position. Here there’s a lot of room. For example, 73% of websites cited by Google on the top page have schema markup, while 88% of websites have none at all.
How to Approach This at Your Organization
Webflow’s AEO assessment is a good model for any company that wants more of its people building their own tools:
Start with a real problem you hear again and again. Guy didn’t invent this on his own. AEO came up in every conversation he had with a CMO, and a complaint you keep hearing is a good sign there’s something worth building.
Build the rough first version yourself. Guy’s weekend prototype wasn’t pretty, but it worked, and it was useful enough that the sales team wanted it. That’s your signal to keep going.
Let AI cover the skills you don’t have. You can turn what you know about your field into a useful tool by building it in conversation with AI. Guy was “coding in English” and talked through structure and approach with LLMs. Some technical comfort plus deep knowledge of the subject now gets you prototypes that used to take an engineering team.
Expect to keep changing it. Guy’s V1 was 600 lines. The production version is 87 pages spread across several specialized agents. Your first idea will change a lot as you learn what users need.
Bring in specialists to take it to production. Guy built the assessment logic. Webflow’s design team made the polished pages, marketing ops built the form and the Marketo integration, and engineers wrote real code where it was needed. Know where your expertise stops and work well with the people who pick up from there.
Measure real use, not vanity metrics. 3,500 assessments in three months means thousands of conversations with potential customers, each one getting personal, specific advice. Care about whether it’s useful more than about how big the numbers are.
The Bigger Picture
Guy’s closing advice is simple and worth taking seriously: “The amount of time and resources this took compared to a year ago, two years ago? Unfathomable to be able to do this this quickly with this much.”
This is the change AI brings. People who understand a problem deeply can now build the fix themselves, try it with real users and improve it from their feedback, all before anyone commits engineering time to an idea that might not work.
Engineers still matter (Webflow’s EPD team is “phenomenal” and “busy building the core product”). What AI adds is a whole new class of internal tools, prototypes and experiments that can be tested quickly and then taken to production only when they’ve earned it.
Every company has operators with deep knowledge of their area and ideas for tools that would help customers or colleagues. What matters is whether you give them the permission and support to build them.
Want to check your own AEO maturity?
Go to webflow.com/solutions/aeo for a free assessment of your site.
Want to learn more about building AI tools that solve real problems?
That’s what we work on at elvex. We help teams turn operators into builders and ideas into production tools. Get in touch to keep the conversation going.