Webflow MCP Now Reads Your Traffic, A/B Test and AEO Data

On October 15, Webflow added Analyze, Optimize and AEO Analytics data to its MCP server. Teams that use these tools can now ask the same AI agent about traffic, A/B tests and AI crawler activity. A Gemini connector launched the same day, but for now it only works in the US.

What happened

On October 15, Webflow announced that its MCP server now reads Analyze, Optimize and AEO Analytics data. MCP (Model Context Protocol) is the layer that lets AI assistants like Claude or ChatGPT work directly with your Webflow site. Agents can now read:

• Page-level traffic, goal data and filtered queries from Analyze. Previously agents could only read site-level data.
• A/B tests and experiment results from Optimize, including which variation is winning. Creating and editing experiments through an agent is still in development.
• From AEO Analytics, the pages LLM crawlers request most (filterable by provider, such as OpenAI, Anthropic and Perplexity), plus your tracked prompts and how answer engines responded to them.

The update also brings Team Administration: inviting people, changing roles and removing access to workspaces and sites can now be done from an agent or a script.

According to Webflow’s blog announcement, agents respect the roles and permissions your team has already set and only see the data the person prompting them can access. There is one important condition: an agent can only read data from these three tools if you have access to them in Webflow. If you do, reading them through MCP costs nothing extra. The announcements don’t say which plans include Analyze, Optimize and AEO Analytics; if these tools aren’t enabled on your site, the update doesn’t give you this data.

The Gemini connector: free, but US-only for now

The same day, Webflow released its MCP connector for Google Gemini, which joins Claude and ChatGPT. From inside the chat you can build pages and components, manage CMS collections, add interactions, and run SEO/AEO and accessibility audits. It’s free to install and, according to Webflow, doesn’t use Webflow AI credits at this time.

There is an important limit for teams in Turkey. Webflow’s help article says Google only offers the connector in the US and in English for now. You need to be 18 or over, based in the US, and signed in with a personal Google account; work and school accounts aren’t supported yet. Gemini Enterprise support is “coming soon,” and Gemini support for Webflow Enterprise plans is in development. Given these conditions, you most likely can’t use the Gemini connector from Turkey today.

Webflow says the same MCP server also works with the Claude and ChatGPT connectors. We could not confirm that these connectors are available from Turkey or for your account type; we recommend checking during setup.

What it means for your brand

Webflow’s example goes like this: traffic to a landing page dropped this week and you ask your agent why. The agent reads traffic from Analyze, running experiments from Optimize and crawler activity from AEO Analytics, and lists them in one response. For teams that have these tools, the three were already in Webflow, just on separate screens. What’s new is pulling them together with one question. Deciding the cause of the drop is still up to you: an experiment starting or crawler activity changing in the same week doesn’t show that it caused the drop.

If you use Analyze, the most concrete gain is page-level goal data. One of Webflow’s examples is asking which landing pages produce goal completions and which only collect visits. You can also compare the weeks before and after a page change. That comparison can’t tell you for certain whether the change worked, since traffic can shift for other reasons in the same period. An A/B test set up in Optimize is a more reliable way to measure that.

AEO (answer engine optimization) is about showing up in AI answer engines like ChatGPT and Perplexity. AEO Analytics does the measuring; the MCP update opens the same data to agents. According to Webflow, this data helps you decide which pages to work on first. Webflow’s own caveat is worth repeating: “Crawler requests aren’t citations, though.” A crawler requesting your page doesn’t mean an answer engine used it. A list of the most crawled pages shouldn’t be read as a list of the pages that show up most in AI answers.

For brands and agencies running several sites, Team Administration can help too. In Webflow’s example, adding a new teammate to three sites with the same role takes a single prompt, and the agent can only make access changes your role allows. We still recommend keeping a record of access changes.

What it doesn’t change

An agent can only read data that exists. With no experiment set up in Optimize, there are no results to read; with no tracked prompts in AEO Analytics, you won’t see what answer engines say about you. Which questions matter to your brand, which hypothesis an experiment should test, and how a page should change based on the result are still decisions for your team.

Practical steps for your Webflow site

The first three steps happen in Webflow’s own dashboard and don’t need a connector.

1. Check which of Analyze, Optimize and AEO Analytics are enabled in your workspace, and who on the team holds which role. What an agent sees is capped by the permissions of the person prompting it.
2. If you use Analyze, define your goals. Without goals such as form submissions, quote requests or bookings, page-level goal data won’t tell you much.
3. If you use AEO Analytics, pick the questions to track. Write a short list of questions your customers are likely to ask AI assistants; you see what answer engines say through these prompts.
4. Once you’ve confirmed the Claude or ChatGPT connector works for your account, ask Webflow’s example question (traffic, experiments, crawlers) about a page you care about, and compare the figures in the answer with what you see in the Webflow dashboard.
5. Keep publish approval with a person. Connectors can create pages, make bulk CMS changes and publish content to the live site.

How we approach it at Klyne

As a Certified Webflow Partner, we think this update helps most on sites where the data is set up properly. On a site with no defined goals, no experiments or no tracked questions, the agent has little to read. That’s why, for Webflow web design projects, we recommend planning goals and measurement together with the site structure, and keeping the final call with your team if you bring in an agent.

Frequently asked questions

Can I use Webflow’s Gemini connector from Turkey?

Most likely not yet. Webflow’s help article says it’s offered only to users who are in the US and 18 or over, in English, with personal Google accounts. Webflow’s MCP server also works with the Claude and ChatGPT connectors; check during setup that they’re available from Turkey and for your account.

Do these features cost extra?

According to Webflow, if you already have access to Analyze, Optimize and AEO Analytics, reading their data through MCP costs nothing extra. The Gemini connector is free to install and doesn’t use Webflow AI credits for now. If these tools aren’t on your site, you can’t reach this data through MCP either.

Can an agent make changes to my site I didn’t want?

An agent is limited to the Webflow role of the person prompting it; everything it can see and change depends on that role. Since connectors can edit pages and CMS content, we recommend assigning roles carefully and keeping publish approval with a person.

If you’d like to plan how to set up traffic, experiment and AI visibility data for your Webflow site, get in touch.