AI search: ranking in Claude

Hey builders,

Two weeks ago, Claude turned on organic app discovery. If you missed the breakdown of what that actually means and why it's a big deal, I covered it in last week's post.

Quick recap: Claude is now surfacing connectors (MCP & MCP Apps) inside conversations at the moment a user expresses intent. Users don’t have to go to a directory visit or know the app exists ahead of time. The user types what they want, Claude pulls in the apps that match. It’s a brand new discovery field, and a brand new form of SEO every company will have to learn to optimise for.

Two weeks in, we've been deep in the data. Tracking every connector daily on agentdiscoverability.com. Running tests on customer apps. Speaking with founders building connector-first companies, and with enterprises trying to figure out how to win this new channel.

We have a lot of interesting insights to share soon a detailed view of the data, but not quite yet as the data is still cooking 🙂 But what I wanted to share today are some raw thoughts and reflections on what we’re seeing on the ground.

SEO 2.0: Claude can now organically surface your Claude connector

For those that want to watch me breakdown what we know in a video, you can watch it just below!

How apps are actually showing up

There are a few different patterns we're seeing live right now:

→ Sometimes Claude returns a list of connectors to pick from. This is typically 3-5 but we have also seen lists as large as 8 to 10 connectors where the user has to scroll through to choose

→ Sometimes when one connector is a clean fit for the intent, it just surfaces a single option

→ Rare but also happening: connectors can appear inside clarifying prompts

All variations are happening every day. Which one you land in depends on the intent, the prompt, and how clean the match is.

It's happening across every category

We’re not just seeing this happen in consumer apps, it’s happening across B2C and B2B. In fact, we’re seeing ever app get surfaced if you run the right prompts so it is across all categories and all apps.

Opportunity for anyone who gets their connector in the store to start appearing for their target prompts.

Only 386 connectors live in total

The Claude connector directory is growing quickly, but it is still early.

• 386 Claude connector listings live in total as of May 8, 2026

• 118 added in the last 30 days (April 9-May 8), meaning 31% of the current live directory was added in the last month

• 71 added in the 30 days before that (March 10-April 8), so growth is up 1.7x month over month

We’re now at 386 live public directory entries, up from 268 on April 8. So even though the ecosystem is growing fast, it is still relatively early and there is still room for distribution advantage.

For many target intents, there are still only a handful of apps realistically in the running. Categories remain thin, which creates a real opportunity for apps that get into the directory now and optimize for the prompts their buyers actually use.

It surfaces for action-driven prompts, not just connector queries

Some people have been asking whether you need to mention the word ‘connector’ them to appear e.g. "what connector should I use to book a hotel?"

In short, no, this is happening for real intent prompts — "book me a hotel in Lisbon", "create a shareable document", "find me a job in product" — are pulling apps. These are the prompts real users actually type.

Ranking matters

It is not enough to just appear in the list. Where you sit in the order shifts user behaviour. Just like google, the higher the position, the higher the likelihood of being selected. Same dynamic everyone learned with Google search rankings.

Two games you have to play: "are you in the list" and "where do you rank on it".

Brand new apps can absolutely appear at the top

This is the bit I want every business and founder reading this to internalise.

You can have zero SEO presence. Zero prior ranking. Zero brand recognition. And still appear directly for your target prompts above incumbents.

Here you can see how Send is appearing the highest in the rank for this prompt. And thats even with TWO other apps I’m already connected with also appearing as well. Why is this? It’s because they have the strongest alignment to the intent.

If you build a connector and really go after specific intents you really want to win, you can out compete.

The second you're in the store, you can start ranking

There is no warm-up period or domain-authority delay. Get in the directory and you are competing immediately.

Launch partners do get some preferential treatment

Something that I am seeing is Claude is giving some advantage for launch partners or partners that are extremely well known in their domain for some prompts. If you ask for a hotel and you'll get Booking, Expedia, Tripadvisor and Trivago at the top. That is partner weighting and historical precedence at work.

This is not SEO. It’s a new discovery layer to optimise for

This is a brand new field with a brand new opportunity.

It is still very early to determine as a matter of fact what influences rank but based on all our testing across tens of thousands of prompts, traditional SEO does not matter for getting your connector to rank.

To win in this space, you need to compete on tool descriptions, schema design and intent fit, meaning the playing field is genuinely open. That has not been true a new distribution surface for a long time.

And the prize is bigger than a citation

This is the part most companies are under-pricing.

A citation is the LLM equivalent of an organic impression. May or may not click. May or may not convert. Same lossy funnel SEO has always been.

An organically surfaced connector is a product-led tool call. Claude pulls your product directly into the conversation and uses it.

Impression → click → site → convert collapses into a single in-LLM action.

You're not competing for an impression, you're competing for the actual conversion.

That is a meaningfully higher-value placement than anything traditional SEO has ever offered.

Every major LLM is heading the same way

Claude got there first, but this is not a Claude-only story. ChatGPT is expected to turn organic discovery on very soon. Grok just launched its character store. Copilot is already testing organic. Gemini and every other major model company are visibly heading down the same path: bringing external capabilities directly into the conversation through some form of app or connector layer.

The good news for builders: MCP & MCP Apps are an open standard. What you build for Claude today is largely portable to every other client as they each turn this on. Get the playbook right in Claude now, and you're set up to win the next launches before they happen.

In short

→ Apps are surfacing in lists, as single picks, and inside clarifying prompts, across B2C and B2B
→ The store is still small (386 live) but growth has tripled month-over-month — the window is closing
→ Action-driven prompts pull apps in just as well as connector queries
→ Rank, not just appearance, is the game
→ Brand new apps can out-rank incumbents AND already-connected apps if their intent fit is stronger
→ Launch partners get a boost; outside of those, the field is open
→ This is not SEO. It's a new layer with new rules: tool descriptions, schema, intent fit
→ The prize is conversion, not citation
→ Every major LLM is heading the same way. Claude is just first to roll out at scale

What we're doing about it

We've been tracking every connector in the Claude directory daily since launch with our MCP App Intelligence platform on agentdiscoverability.com. If you want visibility on whether your app is appearing for the prompts that matter, where you rank, and which competitors are taking your slots, sign up at agentdiscoverability.com.

If you have any questions on what we're seeing, drop me a reply. I read every one.

Happy building,

Elliot