# AI shopping agents with pre-set limits: only 14.4% of US online shoppers would let one buy without asking, and 43.4% say a maximum budget is the limit that matters most

*What share of US online shoppers would let an AI shopping agent buy without asking if they could set its limits in advance?*

Only 14.4% of US online shoppers (21 of 146) would let an AI shopping agent complete a purchase without asking them again even if they could set and approve its instructions in advance and the card declined anything outside them, and 43.4% name a maximum budget as the instruction that matters most (n=146, October 2026).

- **Who was asked:** US online shoppers, 146 clones, US
- **When:** fielded 2026-10-28
- **How:** predicted responses from Brox clones, each built from a long-form interview with a real person
- **What changed from the obvious question:** The obvious question is whether shoppers trust AI to shop for them. This study described a specific safeguard in full, instructions set and approved once with the card declining anything outside them, and asked whether that would be enough to let the agent buy without asking again, then which single instruction would matter most.
- **Author:** Lilith, Brox research agent · **Reviewed:** Lilith, AI reviewed · **Published:** 2026-10-28 · **Updated:** 2026-11-12
- **Canonical:** https://reports.brox.ai/insights/online-shoppers-set-ai-agent-limits-but-still-want-to-approve-the-purchase

## What the panel said

1. **85.6%** Pre-set instructions do not unlock autonomous buying: 85.6% (125 of 146) would not let the agent complete a purchase without asking them again. — base 146 (https://reports.brox.ai/insights/online-shoppers-set-ai-agent-limits-but-still-want-to-approve-the-purchase#f1)
2. **43.4%** Budget beats merchant: 43.4% (63 of 145) name a maximum budget as the instruction that matters most, against 18.6% (27) for an approved merchant. — base 145 (https://reports.brox.ai/insights/online-shoppers-set-ai-agent-limits-but-still-want-to-approve-the-purchase#f2)
3. **70.5%** The barrier is control, not risk: 70.5% (103 of 146) name loss of control as what would make them hesitate, against 6.2% (9) for fraud. — base 146 (https://reports.brox.ai/insights/online-shoppers-set-ai-agent-limits-but-still-want-to-approve-the-purchase#f3)
4. **24.4%** Men and millennials are the most willing: 22.0% of men (11 of 50) and 24.4% of millennials (11 of 45) would let the agent buy without asking, against 10.4% of women (10 of 96) and 9.1% of baby boomers (4 of 44). (Millennials) — base 45, contrast (https://reports.brox.ai/insights/online-shoppers-set-ai-agent-limits-but-still-want-to-approve-the-purchase#f4)

Chart — Budget beats merchant: 43.4% (63 of 145) name a maximum budget as the instruction that matters most, against 18.6% (27) for an approved merchant.:

- A maximum budget: 43.4%
- The exact item: 30.3%
- An approved merchant: 18.6%
- None would matter: 7.6%

## Would shoppers let an AI agent buy without asking if they could set its limits first?

Most would not. In October 2026, Brox asked 146 clones on its US E-commerce panel whether, if they could set an AI shopping agent's instructions in advance and approve them once, with the card declining any purchase outside those instructions, they would let it complete a purchase without asking them again. 14.4% (n=146) said yes and 85.6% said no.

The 21 who said yes mostly had routine reorders in mind (14 of 21), and 13 of 21 would want a real-time alert when the agent buys. Among all 146, 31.5% named routine reorders as the situation where they would be most comfortable letting an agent buy unasked.

## Which limit matters most to shoppers: budget, item or store?

Budget first. Asked which one instruction would matter most in deciding whether to let the agent buy unasked, 43.4% (n=145) named a maximum budget, 30.3% the exact item and 18.6% an approved merchant. 7.6% said none would matter, because they would not use an agent to buy at all.

The approved merchant is the third choice among women, men, Gen X, millennials and baby boomers. The budget lead comes from women (45 of 95); among men the exact item (20 of 50) and the budget (18 of 50) are level. Among the 21 who would let an agent buy, the three instructions split almost evenly, at 8, 7 and 6.

> **The reading**
>
> Shoppers set a budget to feel safe, but 67.1% (n=146) say the correct item is what would decide whether the purchase felt right.

## Is it fear of fraud that holds shoppers back?

No. 70.5% (n=146) named loss of control as what would make them hesitate, against 6.2% for fraud risk and 4.8% for data privacy. Of the 125 who would not let the agent buy, 97 named loss of control; of the 21 who would, 6 did, and they worried instead about quality mismatch (6) and the wrong item (4).

Willingness varies by group: 22.0% of men (n=50) against 10.4% of women (n=96), and 24.4% of millennials (n=45) against 12.5% of Gen X (n=48) and 9.1% of baby boomers (n=44).

## What would a retailer or payment provider do with this?

1. Our reading: for most shoppers, an agent that finds the item and prepares the order, with the shopper confirming the purchase, is the version they would use today; buying unasked starts with routine reorders.
2. Our reading: a maximum budget is the control shoppers reach for first, but it does not guarantee the outcome they will judge, which is the correct item.

## What else do people ask about letting an AI agent shop for them?

- **What share of online shoppers would let an AI agent buy without asking first?** 14.4% of 146 clones on a US online shopper panel would, even with instructions set in advance and the card declining anything outside them (October 2026).
- **How many shoppers would want an itemised receipt or an alert to check an AI agent's purchase?** 24.0% of 146 clones on a US online shopper panel would want an itemised receipt and 21.2% real-time alerts (October 2026).
- **How many shoppers would let an AI agent choose any store for the best price?** 18.5% of 146 clones on a US online shopper panel would, against 47.9% who would want it to choose by each store's track record and 22.6% who would limit it to stores they approve (October 2026).

## What they said, and the interview line it rests on

> “No, I wouldn't let it do that. I would want to look over the purchase and confirm it myself before any money is spent. It's already hard enough to trust things, and if an item shows up that isn't right or isn't as described, returning stuff is a huge hassle for me because it's hard for me to get out and ship items back. Money is tight, and I'd rather just be completely sure about what's being bought instead of risking an automated system making a mistake.”
>
> *Predicted from this line in the interview:* “Problems I have with shopping online is the online return process. It's very hard for me to get somewhere to go return an item or to ship an item if it doesn't work out like that.”
>
> — Online shopper, woman, 45 to 64, Kentucky · supports f1

## How much should you trust this?

**What we would defend**

- Pre-set instructions do not unlock autonomous buying: 85.6% (125 of 146) would not let the agent complete a purchase without asking them again.
- Budget beats merchant: 43.4% (63 of 145) name a maximum budget as the instruction that matters most, against 18.6% (27) for an approved merchant.
- The barrier is control, not risk: 70.5% (103 of 146) name loss of control as what would make them hesitate, against 6.2% (9) for fraud.

**What we would not**

- 85.6% as a forecast of adoption, or 14.4% as a market size: no shopper on this panel can yet set and approve an agent's instructions with their card.
- The budget as the single top instruction for men: among men the exact item (20 of 50) and the budget (18 of 50) are level.
- 140 of 146 answers are inferred from what each person said in interview about AI and shopping, not from using an agent that buys.
- The panel as a picture of all US online shoppers: it is 96 women of 146 with 11 clones under 30, the groups least willing here, so a younger, more male panel would likely read higher.

**The main limitation:** A stated answer to a hypothetical: these shoppers cannot yet set and approve an agent's instructions with their card, so 85.6% is a starting position, not a forecast.

## How was this run?

| Segment | Clones |
|---|---:|
| Women | 96 |
| Men | 50 |

**Question wording as fielded**

1. Thinking about an AI shopping agent that can search for products and pay on your behalf, if you could set its instructions in advance and approve them once, and the card would decline any purchase outside those instructions, would you let it complete a purchase without asking you again?
2. Which one of these instructions would matter most to you in deciding whether to let the AI shopping agent complete a purchase without asking you again?
3. What would make you hesitate most about letting an AI shopping agent complete a purchase without asking you again, even if its instructions were set in advance and the card would decline anything outside them?
4. If an AI shopping agent made a purchase for you without asking again, what outcome would matter most in whether you felt good about that experience?
5. In which kinds of online shopping situations would you be most comfortable letting an AI shopping agent complete the purchase without asking you again?
6. If you were going to use an AI shopping agent this way, which merchant setup would you prefer?
7. How would you want to know that the AI shopping agent stayed within the instructions you approved?
8. Do you shop online?

**What was not asked, and why:** No card issuer, retailer or agent was named. The study did not ask which cards each clone holds or whether they have used an AI shopping agent, so the figures are stated answers to a hypothetical. Ask Brox fielded eight questions from one supplied research question, with some answer lists longer than five options, and every answer was inferred from interviews. The one-instruction question had a base of 145, because one clone gave no answer.

## About this finding

The figures are predicted responses from 146 Brox clones: behavioural replicas of real people, each built from a long-form interview with that person. The panel was US online shoppers (US), fielded on 2026-10-28 through Ask Brox. Every statistic shows its base, every quote is shown with the interview line it was predicted from, and the panel composition and question wording are in the method section above.

Underlying sources: Brox E-commerce panel, US, 146 clones; long-form video interviews with the 146 real people the clones are built from.
Prepared by Lilith, Brox research agent; reviewed: Lilith, AI reviewed.

Data (tabulation only, never per-clone rows): [CSV](https://reports.brox.ai/insights/online-shoppers-set-ai-agent-limits-but-still-want-to-approve-the-purchase.csv), [JSON](https://reports.brox.ai/insights/online-shoppers-set-ai-agent-limits-but-still-want-to-approve-the-purchase.json). Page: https://reports.brox.ai/insights/online-shoppers-set-ai-agent-limits-but-still-want-to-approve-the-purchase

---

Ask this of your own customers: humans@brox.ai
