---
url: 'https://seoafterai.com/ai-search/in-ai-mode-shoppers-take-the-shortlist-they-are-given/'
title: 'In AI Mode, shoppers take the shortlist they are given'
author:
  name: Nadia Sorensen
  url: 'https://seoafterai.com/author/n-sorensen/'
date: '2026-11-02T14:04:56+00:00'
modified: '2026-11-02T14:04:59+00:00'
type: post
summary: 'Kevin Indig''s Growth Memo reports that 88% of shoppers in its AI Mode study accepted the product shortlist as shown, and 74% chose the product ranked first. In classic search, 56% built their own list. The article cites these figures from earlier studies and does not repeat their sample sizes or dat'
categories:
  - AI Search
tags:
  - AI Mode
  - AI Overviews
  - ChatGPT Search
image: 'https://seoafterai.com/wp-content/uploads/2026/11/cover-276-14cfb260.png'
published: true
---

# In AI Mode, shoppers take the shortlist they are given

Kevin Indig’s Growth Memo reports that 88% of shoppers in its AI Mode study accepted the product shortlist as shown, and 74% chose the product ranked first. In classic search, 56% built their own list. The article cites these figures from earlier studies and does not repeat their sample sizes or dates.

## Table of contents

- [What this is about](#what-this-is-about)
- [What the source says](#what-the-source-says)
- [What the data shows](#what-the-data-shows)
- [What it does not mean](#what-it-does-not-mean)
- [What to do](#what-to-do)
- [FAQ](#faq)

**Key takeaways**

- In Growth Memo’s AI Mode study, 88% of participants took the AI shortlist as-is, against 5% who built their own. In classic search, 56% built their own.
- 74% picked the top-ranked product. Rank 3 and below shared 10%.
- Brand recognition was named as a trust signal in 34% of AI Mode sessions and 23% of classic search sessions.
- Participant counts, dates and product categories are not given in this article. Until we see those, the shares are Growth Memo’s findings and nothing firmer.

## What this is about

On October 12, 2026, Kevin Indig made the case that an AI shopping answer works less like a results page and more like a friend’s shortlist, and that most people stop there. If he’s right, a retailer is no longer fighting for position ten versus position four. It is fighting to be one of three or four names, and ideally the first.

## What the source says

Most of Indig’s piece is about e-commerce growth in 2027, built around five stages that feed each other: get on the AI shortlist, sell to whoever still clicks through, ship a box people want to film, earn the second order, and become a name people remember. Recall, in his account, puts the brand back on the shortlist.

The argument rests on one claim from Growth Memo’s own research, that “users accept and trust the shortlist AI serves them”. A shortlist here means the handful of products an AI answer recommends by name.

Most of the article’s other numbers come from outside. Adobe, Salesforce, IAB Australia, Ahrefs, Shopify, Klaviyo, Gorgias and others supply figures on AI shopping use, quizzes, packaging and repeat purchase. We cover only the findings about how shoppers treat AI answers. The retention and email advice is outside this section’s beat. The piece also carries a sponsor message and promotes paid Growth Memo templates. We leave both out.

## What the data shows

Growth Memo’s AI Mode study, as summarised in this article:

| Behaviour | AI Mode | Classic search |
| --- | --- | --- |
| Took the shortlist as shown | 88% | 21% (shortlist in AI Overviews) |
| Built their own shortlist | 5% | 56% |
| Named brand recognition as a trust signal | 34% of sessions | 23% of sessions |
| Picked the product ranked first | 74% | not given |
| Picked a product ranked third or lower | 10% | not given |

*Source: Kevin Indig, Growth Memo, October 12, 2026, citing earlier Growth Memo user studies. Sample size, dates and categories are not stated in this article.*

If ranks one, two and three-or-lower cover every pick, the second-ranked product got about 16%. That is our arithmetic. The article does not give it.

The article adds that 64% of AI Mode users clicked nothing and 23% went to an outside site by typing it in, without a link. A second study on shopping in ChatGPT found that 93% of sessions had no meaningful click-out and 3% checked the answer elsewhere on the web. The article does not say who ran that study or on how many sessions.

The other number that matters is a correlation. In ChatGPT, Indig reports, a brand’s share of voice and how often users chose it correlate at 0.57 on average. (Share of voice is just how often a brand turns up across a set of answers.) The category spread is wild. Groceries sit at 1.0 and supplements at 0.9, while fitness apps land at −0.4 and coaching at −1.0.

Against the seven method points we ask of any study, this article supplies almost none for the AI Mode figures. It does not give the number of participants, the tasks they were set, the dates, the product categories or the country. It does not say whether people knew they were observed. The figures may be well founded in the original reports. Here they arrive without their method, so we report them as Growth Memo’s findings and not as fact.

## What it does not mean

**It does not mean 74% of AI shoppers buy the top product.** Participants picked a product in a study. Would they have paid for it, at that price, after a look at the reviews? Nobody asked. What got measured is choice inside the answer, and that’s not revenue.

**It does not mean share of voice drives choice in every category.** A correlation of exactly 1.0 or −1.0 usually means very few data points. With four brands in a category, a perfect score is easy to get by chance. Coaching at −1.0 would mean the more often a brand appears, the less often it is chosen. That is more likely noise than a law of coaching purchases.

**It does not show that rank causes choice.** The top product may sit first because it is the best known, the cheapest or the best reviewed, and shoppers could be picking it for exactly those reasons. You’d only untangle that by shuffling the order of the same products, and we can’t tell from the article whether anyone did.

**It does not turn one prompt into evidence.** For brand recall, Indig leans on one exchange with ChatGPT about minimalist wallets. Ask generically and you get one brand back. Name a brand and the answer stays with it. Fine as an illustration, but it is one run. AI answers vary between runs, and our [guide to how answer engines cite](https://seoafterai.com/ai-search/ai-search-how-answer-engines-cite/) explains why one response proves nothing about the next.

**It does not cover B2B.** The shortlist effect looks similar to what BrightEdge found for software buyers, where [ChatGPT named 27.9% fewer B2B tech brands per answer](https://seoafterai.com/ai-search/chatgpt-names-fewer-b2b-tech-brands-per-answer/) between March and September. But that study counted names in answers. This one watched people. The two point the same way without measuring the same thing.

Here is what would sink the central claim. Someone publishes a study with a stated sample, and AI Mode shoppers in it compare outside the answer about as often as classic searchers do, or pick lower-ranked products at similar rates.

## What to do

1. **Check if you’re on the shortlist at all.** Take 20 buying questions from your category. Ask each one in AI Mode and in ChatGPT on several days across a week, and write down who gets named and in which order. A single pass per question tells you very little. 2. **Give the assistant something to recommend.** That means the boring stuff in plain text, on product pages and in the feed: price, size, materials, who it’s for, how it stacks up. Indig’s rule of thumb is eight to ten real buyer questions answered on each product and category page. 3. **Look at what other people say about you.** Indig quotes Ahrefs figures: AI Overview visibility tracks YouTube mentions at 0.7 and branded web mentions at 0.7, but domain authority at only 0.3. Correlation, so don’t read cause into it. Still, reviews and creator videos are a big part of what these answers are built from. 4. **Measure recall directly.** Put an unaided question in the post-purchase survey, something like “which brands came to mind first?” The answer moves slowly. It’s also the only signal here that holds up whether shortlists exist or not.

This week, try your five most valuable buying questions in AI Mode, three times each, and note your position. Anything below second place is your starting point.

## FAQ

What do you mean by a shortlist?
It is the small set of products an AI answer recommends by name, usually three to five. In Growth Memo’s study, 88% of AI Mode participants accepted that set without building their own.

How many people were in the AI Mode study?
This article does not say. It refers to earlier Growth Memo research for the method. Until the sample size and dates are checked in the original report, the percentages should be quoted as Growth Memo’s findings.

Does this mean classic search no longer matters for shopping?
No. Indig cites an analysis of about 4,000 commercial queries in which 96% showed product grids in Google. Shoppers still see and use those grids. What changes in AI Mode is that fewer of them compare options on their own.