- Elliot from Ghost Team
- Posts
- A new AI discovery hack!
A new AI discovery hack!
AEO is the start... but what's next?
Hey builders,
We just ran an MCP event in Lisbon! Was amazing gathering the tech community together there. I gave a talk on what I think is the most critical problem for almost every business right now: AI discoverability.
Everyone's talking about AI discoverability right now - how do you show up when people ask ChatGPT or Claude for a recommendation or to take an action instead of Googling it?
And one of the best answers today is to have an AI App. Not a website the AI might cite, but an actual app (a ChatGPT App or Claude Connector) that lives inside the platform and does the job right there in the conversation.
But here's the part almost nobody's talking about: once you have an app, the app itself needs to get discovered. Agents choose which tool to call - and if they never pick yours, it doesn't matter how good it is.
That question is the whole story. So today I'm breaking down the talk I gave:
→ Why AI platforms are becoming the new browser
→ The shift from chatbot era to agent era (and why it changes distribution forever)
→ The black box problem every builder hits after go-live
→ AEO vs ADO - and why the bigger game is being the tool the agent calls, not the answer it cites
Let's dive in.
First, why should you care? Look at the numbers.
ChatGPT is at 1B+ monthly users. Gemini is at 472M. Claude is at 56M.
These platforms are becoming the new browser. It's where people ask for recommendations, run comparisons, and get their work done, every single day.
If your customers are making decisions inside these interfaces, that's where your product needs to show up.

We're moving from the chatbot era to the agent era
Here's the shift that matters.
A chatbot gives you text. Ask ChatGPT to book you a hotel and it used to hand you a list - useful, but you still had to go do the booking yourself.

Agents actually do things. They choose tools, call apps, and finish the workflow inside the chat. ChatGPT and Claude are both doing this today.

ChatGPT Apps and Claude Connectors are AI Apps that live inside these platforms. Ask to book a hotel now and instead of a text response, an actual app loads in the interface, ready for you to complete the booking without ever leaving.
This is a big deal. It means your product can show up at the exact moment of intent.

Not a search result later. Not a website visit first. Right in the conversation, the second the user is ready to act
Every major platform now has an app store for this. And the directories are growing fast - we're tracking 1,700+ ChatGPT apps, ~500 Claude connectors, and Codex plugins just getting started. See the live tracker here: agentdiscoverability.com/track

But here's the problem: go-live = black box
You build. You test. You submit. You go live. Great.

And then... it's a black box.
→ Will an agent ever find you?
→ Will it pick you over a competitor?
→ How do you actually get users?

You have no idea. There's no Search Console for agents. No App Store analytics. Nothing.
I've watched this pattern before. Every new platform creates a new optimization surface:
→ Web 1.0 gave us SEO
→ Web 2.0 gave us social & growth marketing
→ Mobile gave us ASO
→ The agent era gives us... AEO AND what we call, ADO (Agent Discovery Optimisation)

AEO vs ADO: the distinction everyone's missing
Everyone right now is optimizing to get mentioned in an answer. That's AEO - Answer Engine Optimisation. It's a good start, but it's not the full picture.
We think the bigger game is ADO - Agent Discovery Optimisation:
→ AEO = citations. Getting mentioned in answers and recommendations. → ADO = tool use. Making sure it's YOUR tool the agent actually calls, above your competitors.
One gets you talked about. The other gets you the transaction.

And here's the thing: this ranking is already live. When someone asks Claude for a hotel, the agent surfaces a shortlist of connectors and picks one. That's a ranking surface. And someone wins it, every single time.

We're already seeing clear category leaders emerge. Booking.com appears in 94% of tracked hotel prompts. ClickUp leads project management at 93%. HubSpot owns CRM at 91%. These positions are being claimed right now, while most companies don't even know the game exists.

So, how do you actually know your visibility?
This is the question that matters after everything above: when someone asks Claude or ChatGPT for a tool in your category, do you show up? Do you get picked? You can't improve what you can't see.
The platform gives you a Discovery Score: four measured buckets that tell you how likely agents are to find and choose you:
→ Reachability - can Claude actually reach your product?
→ Relevance - does your listing match the intents people ask agents to handle?
→ Recall - do you appear when agents search for options?
→ Rank - do agents choose you to get the job done?

Measure → score → fix → lift. Then agents find you, use you, and keep coming back.
Your Discovery Score will decide if agents use you
Here's the way I'd frame all of this.
For 25 years, your Google ranking determined whether people found you. For the last 15, your App Store ranking determined whether people downloaded you.
In the agent era, your Discovery Score determines whether agents find you, choose you, and call your product or service - or your competitor's. As more of the internet's decisions get made inside ChatGPT and Claude, this becomes the determinant of whether your product or service gets used at all.
The good news: it's day one. The rankings are still being claimed, and most companies don't even know the game exists yet. That's the window.
Here's how to get started:
→ Check your Discovery Score today - every company can get their score to see how they rank today. Just reply ‘SCORE’ to this email and I’ll send you your score.
→ Want to go deeper? Everything we're tracking is at agentdiscoverability.com/track
→ Read the full guide on organic discoverability — I wrote a deeper breakdown of how agents actually surface and rank connectors here: https://www.agentdiscoverability.com/blog/organic-discovery-in-claude-guide/
→ Have questions? Just reply to this email. I read every reply.
Happy building,
Elliot
P.S. One plug: we're hiring. If you're an engineer who wants to build at the frontier of the agent discovery, reach out → [email protected]