AI has half-decided before it ever recommends you

A few weeks ago I wrote about the four names ChatGPT gave me when I asked it to recommend a wedding venue near me. One of our clients runs one of the loveliest venues in the county, and it wasn't on the list.

I focused on the list. What I skipped over is the more interesting bit: what happened in that chat before the four names appeared. The list wasn't the start of the decision. It was nearly the end of one.

There's a conversation before the shortlist

When someone uses an AI assistant to find a supplier, they rarely open with "give me a name". The request usually moves through three rough stages:

• "I need a way to do X." They describe a problem, not a purchase.
• "What should I be weighing up?" They ask what matters, what to look for, what the trade-offs are.
• "So who should I actually go with?" Only now do they ask for real names.

The assistant behaves differently at each stage, and here's the part that matters: most of the shaping happens in the first two. By the time it hands over names, it's working from a picture it has already built, largely from what it already knows about your field rather than a fresh look at the web that minute.

The visible bit is the bit that's already decided

You can't see those early stages at all. You don't get to watch someone describe their problem, and you certainly can't see how the assistant decides what "good" looks like in your line of work. By the time a recommendation appears, the shortlist is half-settled.

So the one part you might eventually glimpse, your name in a list, is the part where the thinking is already done. That's a different problem from the one most people picture. It isn't "how do I get to the top of the answer". It's "how do I get into the thinking, long before the answer".

What actually moves you upstream

Three things move it, and I'll be straight about the limits of each.

• Be clear about who you're for. An assistant matching a problem to a supplier needs to understand who you genuinely serve. "We help businesses grow" tells it nothing. "We rebuild slow WordPress sites for professional-services firms in the South West" gives it something real to match against.
• Say things only you can say. The models already know the generic version of your subject. What they can't get for free is your specific experience: the real numbers, the trade-offs you've hit, the things you'd tell a client over a coffee. (The term for this is information gain, which just means adding something the model doesn't already have.) Restating what's on a thousand other sites adds nothing.
• Earn a reputation off your own site. What other people say about you, reviews, mentions, being cited elsewhere, feeds the picture too. You don't fully control it, which is rather the point of it being worth anything.

Nobody, me included, can give you the exact weighting an assistant puts on these. Anyone who claims to is guessing with confidence. But the direction of travel is clear enough to act on.

Where our CMS helps, and where it can't

Last time I wrote about the AI-search tools we've built into Motley CMS: the technical groundwork, and a checker that tells you whether an assistant names you. Those handle the visible, decisional end, the bit nearest the answer.

The upstream work is a different thing. Being clear about who you're for, writing what only you could write, earning a reputation worth citing: that's content and graft, not a feature you switch on. No platform does it for you, ours included. What a good one does is make sure the effort isn't quietly thrown away by plumbing that gets the basics wrong.

And match the effort to the size of the decision. Nobody asks AI which toothpaste to buy and agonises over the answer. Choosing a builder, or committing to a platform for the next few years, is a heavier decision, and worth a heavier amount of work.

Where to start

Before any of this, one honest question: is AI even a channel your buyers use? For some businesses it's already shaping enquiries. For others it's barely a trickle, and the time is better spent elsewhere. Worth knowing which you are before you pour effort in.

If it does matter for you, the next questions are practical. What are people actually asking on the way to needing what you do, and at which stage? Then you do the work: sharper positioning, content only you could write, a reputation worth quoting.

If you'd like to work out whether AI is sending people your way, and where you sit in that conversation before your name ever comes up, that's exactly the sort of thing I'm happy to look at with you. Get in touch and we'll take a look together.