What Is EmDash by Cloudflare

What Is EmDash by Cloudflare

By Stas Levitan, CEO · · 3 min read

Cloudflare launched a new CMS called EmDash, and this news is making a lot of noise. The new CMS is supposed to be the spiritual successor to WordPress, as they claim, and it is supposed to be agent-friendly, give agents skills for use on-site, not for the discoverability of the site in AI search, and many other features.

To be fair, there are a few things in EmDash that are genuinely interesting.

The first is the security model. Their core argument is that WordPress plugins are fundamentally too trusted, and EmDash tries to fix that by sandboxing plugins and giving them only the permissions they explicitly ask for. That is a real improvement on paper, and probably the strongest part of the whole pitch.

The second is that it is built in a much more modern way. TypeScript, Astro, serverless architecture, passkeys by default, cleaner schema handling, and better support for custom content types. Compared to the old WordPress stack, that part does sound more current and probably more pleasant for developers.

The third is that they are clearly thinking about agent workflows. Not AI discoverability, which is a different thing, but actual on-site agent usage. MCP support, CLI access, agent skills, and programmatic site management are all smart ideas if you believe more websites will be operated partly by agents.

The WordPress import story is also important. If they want adoption, they need migration to be easy, and at least conceptually they understand that.

So no, it is not that EmDash has nothing good in it. There are some smart ideas here, especially around plugin security, developer experience, and agent-facing tooling.

I have two immediate questions. Does the world need a new CMS? And would I, as an SEO, be excited about a new CMS from a company that has a long history of routinely blocking search engines even when you ask them not to, modifying your robots.txt by default, and deploying aggressive anti-LLM crawl tools?

That is the core issue for me. It is hard to get excited about a CMS from a company that has repeatedly made life harder for search visibility and crawlability.

Yes, I still use Cloudflare for many reasons, but they are clearly not friendly towards SEO, and I would not trust them with any site that I wanted to build a modern marketing strategy with.

This is my honest and short take on it.

Also, Who Is It For?

Newly built websites? Clearly not. If you just want to build a website, all these features can be easily built on Lovable, Replit, or any other vibe coding platform.

Is it for established businesses on WordPress or any other platform? I find it hard to believe that anyone would move their production site there.

If you are launching something new, you already have many flexible ways to build it. If you are running an established site, migrating your production website to a brand new CMS from Cloudflare sounds like a stretch.

My Guess

My guess is that Cloudflare, like many established companies, is out of ideas. Instead of listening to their user base about the many complaints around SEO and AI discoverability, they create a new CMS and expand horizontally.

So first they become a gatekeeper between sites and AI agents, and then they come out as the savior of us all.

I get it. You need to sell new products if you are Cloudflare. But I think this time it is a miss.

If you want a truly agent-friendly site for LLM discoverability, and if you are serious about winning in AI search, just install LightSite AI on any website. It literally takes a few minutes.

Related reading

From the Journal: What Is Google WebMCP and What It Means for Marketers · The Generative Search Revolution

Try the tool: AI Visibility Platform — free, no signup.

Key takeaways

Brand visibility in AI search is a different mechanic than ranking in classic SERPs. LLMs do not crawl, score, and rank documents the way Google does for the ten blue links. They retrieve passages, fuse them into a synthesized answer, and decide which entities deserve to be named in that answer. That decision is shaped by how often a brand co-occurs with the right topics across the public web, how clearly the brand defines what it does, and how machine-readable that definition is when an AI agent fetches a page directly.

Three things compound over time: structured data quality, citation density across authoritative third-party sources, and consistency of brand description. Sites that ship rich, current JSON-LD across product, FAQ, and article entities give LLMs the canonical version of the brand to repeat. Sites that show up in Reddit threads, comparison reviews, and journalist round-ups give LLMs evidence that the brand belongs in the answer set. And sites that describe themselves the same way everywhere — homepage, press, schema, social — train the model on a single, unambiguous entity.

The work is operational, not magical. There is no file you can add to your site to be cited. There is a discipline of being legible, defensible, and frequently mentioned in the contexts where buyers ask AI assistants about your category. AI Bot Analytics shows you whether the right crawlers are arriving. The LLM Discovery API makes sure they leave with clean, structured data. The GEO Checker gives you a baseline so you can measure whether any of this is working.