The customer may ask AI before asking you.
“Which one is right for me?” A customer asks an assistant. A few names appear. Your website has not had a visit. The comparison has already begun.
Businesses want a place in that answer. But what happens once their name appears? I would look for three things: facts a customer can compare, a reason to choose, and a promise the business can keep.
The search can start somewhere else.
On September 30, 2026, Salesforce reported a 200% year-over-year increase in the use of agentic search as the first step in shopping. That is a reported growth rate, not the share of all consumers using AI. The release distinguishes changes in discovery from fully autonomous purchasing, which it describes as early. [1]
Not everyone has handed their shopping over to an agent. The narrower point matters: a customer’s first question can happen elsewhere. You may be improving your homepage while someone is comparing you with a competitor in a chat.
A mention is not a choice.
What does it cost? Will it work with what I own? When will it arrive? Can I return it? These are ordinary customer questions. Missing or conflicting answers make comparison harder. Clear information helps. It does not guarantee a better ranking in an AI answer.
Then comes the harder question: Why you? For whom are you a better choice? For whom are you not? “The best for everyone” sounds confident. It gives the customer little to work with. A reason to choose needs to be more specific.
The recommendation comes first. The experience follows.
Imagine an assistant recommending a product because delivery is fast. The customer places an order, then learns that their address is an exception. This is a hypothetical example, not a documented case.
The customer might blame the assistant or the seller. Neither reaction makes the disappointment disappear. The sentence that won their attention has become the reason for their complaint.
A brand has to do more than explain itself well. It has to deliver what it explained. Does the delivery date in the ad match the date operations can meet? Does the return policy work when someone actually needs it? That is where a customer can test the promise.
“Help me choose” is not “Spend my money.”
A recommendation is not permission to pay. Mastercard’s September 30, 2026 introduction to its agentic commerce framework addresses that distinction through agent identity, approved intent, operating limits, trusted execution and risk intelligence. This is a company-proposed approach, not a universally validated standard. [2]
The questions a business should ask can be straightforward.
- Did the customer ask for advice, or authorize an order?
- Will a change in price or terms trigger another confirmation?
- Is there a clear way to cancel an unwanted order and recover the money?
- Who can explain what happened and put it right?
These are my review prompts, not a list of legal requirements or a description of any particular assistant. Delegating a task should not make correcting a mistake harder.
Give the answer a reason to include you.
Being visible to AI can matter. But chasing visibility alone can become another exercise in writing for a platform. I would begin with comparable facts, a reason to choose, and an experience that confirms the choice was sound.
An assistant may put your name in an answer. It cannot keep your promise for you. The customer can ask AI. Your business still has to give them a reason to say, “That is why I chose this brand.”
Sources and limits.
[1] Salesforce, fourth State of Commerce release, September 30, 2026. Its methodology describes a survey of 3,450 commerce professionals across 20 countries and 13 industries (April 10–June 4, 2026), platform behavior across 37 countries (Q1 2024–Q1 2026), and 4,689 consumers across eight countries (May 12–18, 2026). The release rounds the consumer count elsewhere; this note uses the methodology’s exact count. These findings do not establish whole-market Korean behavior or a particular brand’s sales effect.
[2] Mastercard, Trust in agentic commerce, September 30, 2026. This note uses the public introduction, not an independent validation of the full white paper. Both sources are published by businesses with commercial interests in this market. The brand and operating argument is the author’s interpretation; recommendation visibility, conversion and loyalty gains are not guaranteed.