🤖 Your Next Customer Isn’t Human

Think about the last thing you bought for your business and count how many websites you actually read before you decided. 

For most of this year, my number has been close to zero. 

I describe what I need, I say what I will spend, and an agent does the reading, then it hands me a few options with a reason attached to each one.

The companies that missed that list have no idea a sale happened.

There was no click to track, no abandoned cart, and no form fill to follow up on, so they were never in the room at all.

So I ran a test. 

I gave an agent a budget, a buyer, and one job: make the shortlist. 

Then I made it show me its work.

The results were… surprising.

🔎 WHAT I FOUND

AI has moved past answering questions about products. It now does the research, cuts the list, and delivers a verdict. 

Here’s what I found:

• 🛒 AI shoppers buy more - Visits to US retail sites arriving from AI sources convert 60% higher and produce 53% more revenue per visit than non-AI traffic (source) 🔗
• 📈 The channel is compounding - That traffic grew 62% year over year in July 2026, and it is up 1,219% since October 2024 (Adobe Analytics) 🔗
• 🙈 Most pages can't be read - Adobe now ships a tool that scores how much of a page a language model can read, because pricing and availability are often invisible to it (Adobe) 🔗

🧪 My Test

I gave an agent a budget of $500 and one job: pick the best AI education expert for a small business owner. 

I never named a person or hinted at a favorite, so it had to build the field itself.

It came back with a ranked top three, then I asked why those three made it and why everyone else got cut.

This is the interesting part:

→ 📄 Pricing gave it something to compare. The options with a clear price and a specific offer were easier to evaluate against my $500 constraint. Programs outside the budget disappeared even when the person behind them had strong credentials.

→ 🔎 Specific offers beat broad authority. The agent wanted to know exactly what someone taught, who it was for, and what the buyer would walk away able to do. The easier that connection was to make, the stronger the recommendation became.

→ 🧾 Evidence changed the recommendation. It pulled from course pages, membership pages and outside sources to verify background, pricing, and what was actually included before deciding who belonged on the list.

The part that threw me most was that the strongest résumé did not automatically win.

The agent found people with bigger credentials, broader name recognition, and impressive experience, then ranked someone else higher because it could make a clearer case that the offer matched the buyer I described.

That distinction is important.

The agent was not asking, Who is the biggest name in AI?

It was asking, Who can I prove is the best match for this buyer, this budget, and this job?

My read is that this is what companies are about to compete on. 

Not whether an agent can find your name, but whether it can find enough specific information to make the case for you.

📌 WHAT IT MEANS FOR YOU

Most people hear AI shopping and picture sneakers and groceries, so they file this under somebody else's problem. 

The near-term shift is AI-mediated research and evaluation 🔗, and it lands in every category that has a buyer, including B2B, where the agent builds the shortlist weeks before anyone talks to sales.

Bottom line: if you sell anything from consulting to HVAC repair to a $40,000 custom build, someone is asking an AI who to hire this week. Here's what we know: 

1) The shortlist got short

Search used to hand you ten links and a second page of results, so a scrappy company could sit at position fourteen and still get found by someone patient. 

An agent returns a few options at a time (sometimes only one), and there is no second page to be discovered on, so you are either on the list or you are don't exist.

2) Your strongest proof might be unreadable

Look at where your best evidence lives right now. 

A case study video?

A carousel on Instagram?

A testimonial baked into a graphic?

These all work beautifully on a human, but an agent understands almost none of it, so the assets you spent the most money on can be the ones the machine skips.

A plain text page with clear scope and a real price gets cited instead, which I find annoying.

3) Other people carry your story now

Since the agent trusted outside sources over first-party copy, the podcast you guested on and the comparison post someone else wrote are pulling more weight than another homepage rewrite. 

That changes where the next dollar of marketing budget should sit, and it means getting mentioned somewhere credible beats polishing your own about page.

Example. My name is Jim Carter III. For those of you in America keeping track, someone with the name Jimmy Carter beat me to name recognition who at one point was our president. Also, Jim Carter happens to be the name of the butler from Downton Abbey (incredible show).

So, search for Jim Carter and you get a mix (of a lot). But you know what helps my ‘first page’ results? The fact I've been on big shows talking about what you're looking for. 

That matters. 

Buyers arrive warm from all of this, because 39% of consumers say they have used AI for shopping and 85% of that group say it improved the experience (Adobe) 🔗 

They show up already holding three options, so you are either one of them or you are arguing against a decision that closed before you knew it opened.

Buying inside a chat window is still clunky. We covered a checkout feature getting cut earlier this year for that reason. 

But…

Deciding inside a chat window works fine today, which is the part worth planning around.

👉 Three things to sit with, whatever industry you are in:

• Could a stranger find a real price for what you sell in under 30 seconds without talking to anyone?
• What share of your best proof is trapped inside video, images or PDFs that a machine cannot read?
• Who outside your company has published anything about you in the last twelve months?

✌️ YOUR TURN

One test, fifteen minutes. 

Open an AI tool with web access, do not mention your own company, and run this:

"You are helping me choose a [YOUR CATEGORY] for a [YOUR CUSTOMER TYPE] with a budget of [$X]. Research real options available today. Give me a ranked top 3 with one paragraph on why for each, and cite the sources you used. Then list the companies you considered and rejected, and the exact reason you rejected each one."

Then push, because the follow-up is where the money sits.

"For each of your top 3, list the specific information that made you confident. Now look at [MY COMPANY] and tell me what information is missing, unclear, or unreadable that would keep it off this list."

If you made the shortlist, find which source got you there and go feed that source more material.

If you missed it, you just got a free audit from your next customer's research assistant, and the fix is usually a pricing page or a services page written in plain words about what you do.

Share your thoughts and tell me what it said about you, even if it stung, because the ugly results teach me more than the flattering ones (and I have collected a few unflattering ones about my own site this year).

P.S. Your next customer isn’t human, and your next employee shouldn't be human either. My client, and good friend Callan Faulkner runs a company with 36 humans and 150+ AI employees handling the repeatable work across marketing, sales, operations, customer success and finance. Her AI Employee Bootcamp help you build two of your own. It’s incredible. Build your first AI employees here.