A few weeks ago I opened Claude and asked it a question about my own business, the same way a stranger might.
It guessed. Not wildly wrong, just vague. It pieced together an answer from whatever it could find lying around the internet, some of it old, none of it exactly right.
That bothered me more than it probably should have. I had already written the real answer down, on my own website, in my own words. The AI just had no way to get to it.
So I fixed that. I connected vedbhanu.com to Claude directly. Now when someone asks an AI assistant about my work, it can read the actual page instead of guessing. You can see it for yourself at vedbhanu.com/mcp, it is live right now.
What you will get out of this:
- What this actually is, with no technical words you have to look up
- Why this applies to almost any business, with real examples
- The real cost and the real risk of leaving it as it is
What AI does with your business right now, whether you set anything up or not
People are already asking ChatGPT, Claude and Grok about businesses, communities and newsletters every day. That is happening today, with or without you.
What the AI does with that question depends entirely on whether it can reach your real information.
Most of the time it can't. So it does one of two things. It scrapes whatever old, half-loaded version of your page it can find and hopes for the best. Or it just makes something up that sounds plausible.
Think of it like this. Someone calls a business and asks the receptionist a question. If the receptionist has the real file in front of them, you get the real answer. If the receptionist has never worked there and is guessing from a five year old brochure, you get something that sounds confident and is often wrong.
Right now, for almost every small business, the AI is that second receptionist.
This isn't just for communities and newsletters
Those are just the easiest examples to picture. The actual problem shows up anywhere a business has real, current information sitting on a website or in a document, that an outsider would otherwise have to guess at.
A few examples of what that looks like.
A paid community
Someone hears about your community from a friend, a post, a podcast mention. Before they join anything, they do what everyone does now. They ask an AI assistant about it. "Is this worth it." "What do you actually get." "How much does it cost."
Without this set up, the AI has nothing solid to work with. It might find an old post, guess at the price, or just tell them it does not have enough information and to go check the website themselves. Most people will not go check. They will just move on to the next thing in their feed.
With this set up, the AI can answer with your real pricing, your real description of what members get, straight from your own site. The person gets a real answer in the conversation they are already having, not a dead end that sends them nowhere.
A newsletter or content writer
Someone finds one good piece you wrote. They want to know what else you have said on the topic, or whether you have covered something specific before. Normally that means digging through an archive by hand, if there even is a searchable one.
With this set up, they can just ask their AI assistant directly. "What has this person written about pricing." "Summarise everything they have said about hiring." The assistant pulls the real answer from your real writing, not a guess about what a newsletter like yours probably says.
This is the exact idea Pieter Levels used on his own blog, which is what got me testing this on my own site in the first place. If you write regularly, it is worth a look at what he set up.
A local service business, like a clinic, a law firm or a broker
Hours, pricing and what you actually handle are exactly the kind of thing that changes every few months and gets stale on the internet fast. Someone asking an AI assistant "do they handle this" or "what does a consultation cost" before ever calling you can easily get an answer that was true two years ago and is not true now. That is not a lost click, it is a lost booking, and you never see it happen.
A course or coaching program
Curriculum, pricing tiers, what is actually included, all of it changes as the program improves. A prospect asking "is this worth it" or "what is different about the higher tier" is asking the exact question your sales page was written to answer. If the AI cannot reach that page, it guesses, and a bad guess at this exact moment is the one that loses the sale.
An online store
Stock, pricing and specs change constantly. An AI assistant answering from a months-old cached version of your product page will confidently tell someone something that is no longer true, an old price, a discontinued option, a feature the current model does not have. That is worse than saying nothing.
The pattern across all of these is the same. The businesses that need this most are the ones whose real information changes the fastest, because that is exactly where an AI is most likely to answer confidently and be wrong.
The Ask, Answer, Act loop
Once you strip the technical language out, the whole thing is three steps.
The part worth sitting with is step two. That is the only step that changes. You are not asking anyone to behave differently. You are just making sure the AI they were already going to ask has the real answer to give them.
What it actually looked like to build, in plain words
I did not buy a tool for this. I added a small helper to the site I already have, one that does four simple jobs when an AI assistant asks:
- Look up what I have written, by topic
- Pull up the full details on one of those, when asked
- Show the real pricing, not an estimate
- Hand over the real link to book a call
That is the whole thing. No new subscription, no dashboard to maintain, nothing your visitors have to download. You can read the connection steps yourself on the MCP page, they take a couple of minutes for anyone technical enough to help you set it up.
The real cost, and the real risk of skipping it
Let's do the maths honestly, the way I would for a client.
The cost of doing this is close to nothing. It reuses information you have already written down. Building it took an afternoon. There is no ongoing bill for running it.
The cost of not doing it is a conversation you never find out happened. Someone asked an AI assistant about your community or your writing. It answered, badly or not at all. They moved on. You will never see that in your analytics, because they never reached your site to be counted.
You might be thinking, does this actually matter yet, isn't this early. Maybe for some businesses. But the direction is not in question, more of these conversations happen every month, not fewer. And it matters most for exactly the kind of business where the real answer changes often, because that is where a stale guess does the most damage.
The short version:
- People already ask AI about businesses, right now, whether you have set anything up or not
- Without this, the AI guesses or goes quiet. With it, the AI answers from your real page
- It matters most for businesses whose real information changes often, pricing, stock, availability, curriculum
- A community owner gets real pricing and details in front of someone before they lose interest. A writer gets their archive answered, not just searched. A clinic, a course or a store stops being described by an outdated guess
- The cost to set it up is close to nothing, the cost of skipping it is invisible by design
You can see the working example at vedbhanu.com/mcp, connect your own AI assistant to it and ask it something. If you want to know what setting up the same thing for your own business would look like, that is exactly the kind of thing I cover in a fulfillment diagnostic.
If you are still working out where AI actually fits in your business before this, start with how to find AI use cases in your own business, this article picks up right after that one.
Want to remove founder dependency from fulfillment? Start with a fulfillment diagnostic. I will show you where delivery waits on you, and if nothing is broken I will say so.