What AEO readiness actually looks like (from the people doing the work)

But there's a gap between talking about AEO and actually executing it — and that gap is where most teams are quietly struggling.

The gap stems from how quickly AI platforms are changing, making AEO an ill-defined target. OpenAI, Google, Perplexity, and Anthropic are all approaching this differently, and none have settled on a final answer. Instead of asking whether teams are ready, the better question is: what have people who've actually done the work learned? 

To answer this question, we sat down with five practitioners to talk about what they actually ran into when they started building for AEO: Alexander Diner (Brand and Web Design at Webflow), Josephine Cahill (Web Lead at Oyster HR), Josh Jacobs (Staff Brand Designer at Jasper), Uzair Dada (CEO at Iron Horse), and Mason Poe (Founder at Edgar Allan). Let’s take a closer look at what they shared. 

Want to go deeper on these tactics and all things AEO? Flow TV has it all and is free to watch.

AEO as content architecture vs. technical infrastructure

AEO is forcing brands to reckon with something most have quietly avoided. Alexander explains: "We've been doing decades of SEO where we're essentially gaming for keywords and specific phrases around how we position," he says. "And now we're reorienting, perhaps for the first time, to human language, to the ways that we really think as people and the questions that we actually ask." 

Many brands are discovering that their SEO game was never as sharp as they thought. They could hide behind optimized phrases before, but AEO doesn't offer that cover.

When an LLM surfaces your brand in a response, it's not rendering your homepage. It's not seeing your logo, color palette, or carefully designed hero section. It's reading your words, structure, and claims and deciding whether they're worth citing. Strip away the visual layer, and what's left has to stand on its own.

That puts pressure on an organizational gap that has existed for years. Brand work has historically lived in decks, guidelines, and beautiful PDFs, while performance work has lived in keyword spreadsheets, traffic dashboards, and SEO audits. The two have rarely had to speak the same language. 

AEO is what finally forces that collision and co-owns the discipline together. What agents actually read is built by both teams, and neither team can own it alone. If you're not sure where this work lives in your organization, that's not a secondary problem to solve after you fix your schema. That's the first problem.

Retrofit vs. rebuild: three paths, one question

Once teams accept that AEO is both a content and technical problem, the next question is: do we need to start over, or can we work with what we have? Most teams usually weigh three options: rebuild, iterate, or systematize. Here’s how our practitioners approached each: 

Rebuild: Alexander rebuilt Webflow.com as part of a full rebrand. Even with that, the web team was able to launch the new website quickly, because the goal was velocity, not perfection. "The website should be treated as a living, breathing thing," he explains. “Rebuilding only buys you something if it leaves you with a system you can change fast, not a monolith that requires a project plan every time something needs updating.”

Iterate: Josh saw that Jasper had no dedicated content about their own AI agents, so they were invisible in searches most relevant to their product. LLMs are more likely to cite you for topics where you have clear, structured content; without it, even a category leader can get overlooked. Using internal tooling and templates, Josh filled that gap. In six months, he went from zero to 130 AEO keyword mentions. Josh learned that you need a clear owner and a defined gap before you touch a single page. 

Systematize: At Oyster HR, content accuracy has legal implications since it shapes how companies hire and treat employees globally. To ensure content freshness, Josephine’s team built a workflow to monitor content six months after publication, and anything showing a 15% month-over-month decline in organic traffic is immediately queued for a refresh via agentic workflows. 

Measure carefully 

Every practitioner had a measurement philosophy, and they converged more than you'd expect.

Josephine's approach is a scientific method applied to content: form a hypothesis, run a controlled experiment, measure the impact. In practice, that means tracking citation rate, off-site mentions, competitor share of voice, and sentiment across Reddit and G2. For attribution in a world where GA4 can't track agent-driven traffic, her team scrapes Gong call transcripts weekly, monitoring competitor mentions, integration questions, and how prospects say they found Oyster. It's one of the most reliable attribution signals available right now.

Josh pushed back on something many practitioners are quietly getting wrong: using LLMs to self-assess brand perception. "I would strongly discourage people from taking that as a source of truth," he said. "These models are so biased on everything else you've prompted them with." A model primed by your previous prompts reflects your assumptions back at you, which feels like validation but tells you nothing about how your brand actually surfaces when a prospect asks a neutral question cold. Aggregating results across many independent prompts reduces that bias. A dedicated tracking tool does it at scale. A single ChatGPT conversation about your brand is an echo, not a signal.

And even with rigorous measurement, there's the risk of moving too fast without the right guardrails. Jess Fain, Senior Director of Product at Webflow, named the failure mode directly: "Agents are now members of your team. But they are really high-risk members — entry-level employees who hallucinate, who make mistakes, and who sometimes don't follow the rules."

Where to actually start

Every practitioner on this panel began somewhere different — a content gap, a broken workflow, a brand story that didn't survive contact with an LLM. But their advice on where to begin converges more than you'd expect.

Establish co-ownership across disciplines first. AEO doesn't belong to content or to dev — it belongs to both. Get brand, SEO, demand gen, and web in the same room before anyone touches a page.

Fix your data fragmentation before you produce content at scale. Josephine's team runs every channel off a single brand source-of-truth document — events, demand gen, product marketing, partner content. It's trained, updated, and shared. Fragmented inputs produce fragmented outputs, and an LLM synthesizing your brand across owned, earned, and social will find the inconsistencies faster than you do.

Implement measurement before you scale content. As Josh put it: "If you have the number one position on the SERP for a keyword, you're 3.5 times more likely to get referenced in an LLM." Start with SEO fundamentals, then pull your call transcripts to understand what conversations you're supposed to be showing up for before you measure whether you are.

Build guardrails before you hand anything to an agent. Mason's advice is direct: tell the agent exactly what to check. "Here is our review guide. Here are the 20 things that you as an agent have to look at before anything is moved to the next step." Don't auto-generate keyword-stuffed content hoping it sticks. The models grading your visibility are the same ones you're trying to impress.

The bottom line

AEO is a moving target — but that doesn’t mean that you can’t move past it. The companies making real progress named the problem clearly, measured before they scaled content, and kept humans in the loop when the stakes were high. Mason puts it perfectly, "There is this notion of there being magic wands — that there's this one thing you can just do and it's automatically perfect and done. There are no magic wands."

If you're ready to put that into practice, Webflow's Spring release includes the latest AEO tools to help you get there — and if you want to build the technical foundation first, the AEO courses in Webflow University are a good place to start.