The Proxy Buyer

There's a new buyer in every deal now, and it's not human. It decides whether you and the human ever meet.

Sometime in the last year, a person had a problem your company can solve. They went looking for a firm like yours, pulled together a short list of who to call, and started the conversation.

You were not on the list. You did not lose the pitch, because there was never a pitch. You were sorted out before anyone picked up the phone, and you never heard a thing.

Some version of that has always happened. What has changed is who does the sorting, and how early. It used to be a person with a browser and a free afternoon. More and more often, it's a piece of software working on that person's behalf, and it is faster, more literal, and far less forgiving than that person ever was.

A proxy buyer enters the ring

When someone needs what you sell, they don't start by visiting websites, or even search engines anymore. Instead, they start their search in an AI chat box. Their AI agent finds the candidates, reads their sites and works out what each one actually does. It decides which ones it can vouch for. Then it hands its human a short list of names to take forward.

I call that agent the proxy buyer, because that is what it is. It stands in for the human buyer in the first round, the round that decides who even gets considered.

In manufacturing it might be an agent assembling a quote request from your specs. In professional services it might be an assistant asked to name the firms worth a call. The surface differs. The job is the same: go ahead of the human, size everyone up, and come back with a shorter list.

The shortlist is written before you show up

And that first round has always been where deals are quietly won and lost, long before AI showed up.

85% of B2B buyers purchase from a shortlist of vendors they already had in mind before they started searching. (Bain & Company, 2025)

95% of the time, the vendor that wins was already on that shortlist before the buyer ever spoke to a salesperson. (6sense, 2025)

This isn't new information. This is the wall you have always been up against. For years a shortlist was built from memory: buyers picked from the firms they already knew. That is the incumbents' moat.

If you are a mid-market firm without a household name, it is the wall you have always been on the wrong side of. You were never in the buyer's head, so you were never on the list, and no one ever told you. A good chunk of all your sales efforts, your tradeshows, and networking luncheons have been spent looking for a way to get on those lists.

Things are beginning to change very rapidly. In another report, Bain also said that B2B buyers "have begun constructing vendor short lists within" AI tools, and warned that if a firm doesn't appear in that first AI-generated list, "it may never advance."

That is not a small change. It changes what the list is made of. A list built from memory is made of who the buyer already knew, which is why incumbents win it. A list built inside AI is made of who the machine can find, understand, and verify.

For the first time, a firm nobody has heard of can make the list, if the machine can read it. That is the crack in the incumbents' wall. Being legible to the machine is how you climb through it.

But that's not where the changes, or the job of a proxy buyer ends.

The efficiency jab

You have almost certainly adopted AI by now, and you have adopted it the sensible way: as an efficiency tool. It drafts the proposal, cleans up the copy, summarizes the call, and speeds through the boring parts of your day. That is real efficiency and there is nothing wrong with it.

But here's the catch. While you're concentrating on efficiency gains and being pleased that AI is working so well in your business that you finally have time to work on the business, you're not looking at AI from the most important perspective of all.

Your buyer is using AI as an efficiency tool too. In one case, they're using it to save the most expensive time they have: the time spent talking to vendors who turn out not to fit. The machine's entire value to them is that it clears those vendors out of the way so they don't have to waste time thinking about them.

The way it clears them out is simple and unforgiving. It keeps the firms that can clearly show they meet the need, and it drops the ones that can't. Not the ones that can't do the work. The ones that can't show they can do the work, in a form the machine can actually read and check.

Read that again because that distinction is everything. An excellent firm that can't demonstrate its fit to the proxy buyer looks exactly like a firm that can't do the job.

The next round is on your website

None of this is a forecast. It's already reality. Consider these stats:

• Half of U.S. adults now use AI chatbots, up from a third two years ago (Pew, 2026)
• ChatGPT alone reached 900 million weekly users this year (OpenAI, 2026)
• Nearly half of B2B buyers already reach for generative AI when they research vendors and products (Gartner, 2026)

And their agents are already at your door:

• automated traffic passed human traffic on the web for the first time in 2024 (Imperva, 2025)
• one crawler, GPTBot, climbed from 5% to 30% of AI crawler traffic in a single year (Cloudflare, 2025)

The proxy buyer is not coming to your website. For a lot of your buyers, it has already been.

Watch it happen

You don't have to take my word for any of this. It's the rare business problem you can watch happen in about two minutes, and watching it is more convincing than anything I can write.

Open ChatGPT, or Claude, or Perplexity, and ask the question your buyer would ask: "Who are the best firms for [what you do], for [the kind of client you serve]?" Read the list it gives back. Then ask the harder follow-up: "Why didn't you include [a firm you know is genuinely excellent]?"

What you'll usually find is that the good firm wasn't left off because the machine judged it and found it wanting. It was left off because the agent couldn't confirm it. That's a different problem, and a fixable one, and it's the whole game.

Why good firms don't make it into the ring

The proof that wins you business in a room is often invisible to the reader that matters now. Your track record, your certifications, your named experts, the results you've delivered, the things a human buyer finds reassuring, tend to live where an AI agent can't use them.

We scanned eleven industrial and consulting sites recently, all of them firms that win work based on their expertise. Not one had published a single credential in a form a machine could read. The proof existed on every one of them. It simply wasn't written down where the new buyer looks.

When the proxy buyer comes through, a handful of ordinary website features are enough to get you weeded out:

• Your content loads with JavaScript the crawler can't run. Roughly 69% of AI crawlers don't execute JavaScript (searchVIU / Merj, 2025), so a site that builds its text in the browser looks like a blank page to them
• Your specs and case studies are trapped in PDFs, often scanned ones, which agents parse poorly or skip, so the fact that would have matched the buyer's requirement never gets read
• Your crawlers are blocked, sometimes on purpose and often by an inherited plugin setting or a default CDN rule, so the agent's fetcher is turned away at the door
• Your credentials are claims with no proof a machine can check. "ISO 9001 certified," "award-winning," stated as plain text with nothing behind it. A cautious agent won't repeat a claim it can't verify, so your strongest differentiator is the first thing dropped
• Your only path to a quote is a human-only form, so even an agent that likes you can't take the next step

None of these are exotic. They are the normal state of a site built for human readers. A proxy buyer is an efficiency tool doing its job, which is to weed out anyone who can't clearly demonstrate a fit to its buyer's needs. If it can't find the information it needs, in a form that it can ingest, it can't prove you're a good fit so it quietly moves on to a company and a website that can.

This is winnable, and the order matters

Here's the good news, and it's genuinely good news. This is winnable, and with work you already know how to think about. It's not a rebuild. It's making what you already are legible to a new kind of reader, so that you're the firm the machine can clearly confirm instead of the one it has to skip.

There's an order to this work that matters:

• First the agent has to find you: reach your pages, follow your structure, see what you've published
• Then it has to understand you: what you do, who you serve, how your services and your proof connect
• Then it has to trust you: confirm your claims against evidence it can check, and find the same story everywhere it looks
• Then, and only then, can a proxy buyer act for its human, submitting RFQs or responding to RFPs

Of course, you can't begin this work until you understand what a proxy buyer sees when it looks at your site and identify your own website "gotchas."

Find out what the proxy buyer sees

So find out. I built a free tool, the Signal Check, that shows you what AI assistants actually see and say about your firm when a buyer sends one to look you over. It takes under a minute, and it's the cheapest possible way to stop guessing and start from the truth.