Signal / United States and Europe survey context; local test applies only to the chosen shop, market and item. / Online retail, product listings and shopping assistants

An AI shopping answer can name your shop and still show the wrong offer.

Bain found that specialist retailers often appeared in AI shopping answers more than their market share would suggest. But a mention is not a sale, and it can still send a shopper to the wrong size or price. If you run a shop, test one product before your next promotion. Count how often an assistant names your shop, then separately count how often it gets the exact offer right.

Use this if

You run or support an online shop and can verify one public offer in one delivery area.

No action if

Stop if you cannot identify the exact version or the offer varies in ways you cannot match. If everything agrees, do not change the listing or spend on this test alone.

Article time zone:

Asia/Singapore

Who is outside the scope:

No accounts, orders, customer data, paid advertising, ranking promises or sales forecasts. Only public test questions and public shop information.

First published

General

OTR Emerging Data

Retail | The difference between being found and being right

Being named is not the same as showing the right offer

A shop can be visible in an AI answer while the price or stock for the exact item is wrong.

What the sources show

Bain published its findings at 7:15 pm Singapore time on 20 November. Its study covered more than 4,400 shopping questions and more than 300,000 runs across four retail fields. It measured which retailers appeared in AI answers. Specialists often appeared more than their market share would suggest.

A mention is not an exact offer
  • Retailer namedThe shop appeared
  • Exact offer matchesThe version, price, stock and link agree
Second row is a proposed check, not a finding from Bain.

The new connection

A shop can appear in an answer and still have the wrong version, price or stock shown. Google’s Merchant Center guidance says a shop’s public page and product listing can disagree when updates happen at different times, when discounts change or when one page covers several versions. That can explain a verified mismatch in a Google listing, but does not establish which information an assistant actually used.

Where to look when the offer differs
Where to look when the offer differs
What differsWhat to check first
Out of stockWhether the shop and listing changed at different times
Wrong priceWhether the version or sale price differs
Shop records agreeThe cause is still unknown
Google’s guidance describes its own tools, not how every assistant works.
  1. An assistant names the shop
  2. Check the item link, price and stock
  3. Fix only a mismatch you can verify

A realistic next move

Choose one product and one delivery area. Ask the same three questions in two public shopping assistants. For each answer, check the product link, version, price and stock against your store page and public listing at the same time. Repeat when the next promotion starts. Only fix a shop-page or listing mismatch you can reproduce.

One product, two dates
  1. Before promotionExact offer, location and three questions
  2. After promotion beginsRepeat the same questions in two services
  3. ReviewCompare mentions with exact matches; do not infer sales
This is a proposed test, with no observed results.

Bain’s results cover its sample and do not show how often mentions led to sales. Google’s instructions describe its own service. This article does not establish that a shop’s listing caused an assistant to give a wrong answer, or that a fix will increase sales.

The connection

A mention and a correct offer are different results

Bain disclosed the comparison between retailers’ market share and how often they appeared in shopping answers on 20 November at 06:15 EST, or 19:15 SGT.

Observed

Bain reports a study of more than 4,400 shopping questions and more than 300,000 runs across four retail fields using Profound data. We did not run those tests or inspect merchant transactions.

The connection

Imagine a customer asks for a blue jacket in medium. The assistant can give a shop the credit for being named while linking to the small jacket or quoting a price that has already changed. Keep two tallies: shops named and exact offers right.

How the effect travels

An online shop and its product listing may update at different times. A page can also show several versions at different prices. When your own sources disagree on the exact version, reconcile them first. If they agree, an assistant’s answer alone cannot tell you why it differs.

In February 2024, Google was already telling shops how to show price and stock for each version of the same product in Search. Bain’s November 2026 study instead asked which shops appear in answers. It did not measure whether every product a shop was named for was the right version, price and stock.

The decision path

Compare one offer before and after a promotion

1. Set one fair test

Choose one item version, one delivery area and three ordinary shopping questions.

2. Count two results

On two public services, record mentions and exact matches separately. Read the public shop and listing at the same time.

3. Repeat after the offer changes

Run the same check when the next promotion starts. Fix only what the shop can verify and leave uncertain causes open.

Bain examined four retail fields and also reported survey findings in the United States and Europe. This small shop test does not extend its results to any one country or assistant.

Optional research / Checks 1–10

What the sources support

Money flow

No purchase or advertising payment was observed.

Orders

No order or deposit is established by a mention.

Bottleneck

The exact product version can be hard to match across records.

Beneficiaries

A shopper could waste less time if a wrong link is fixed; no saving measured.

Company finances

No retailer margin or financial result established.

Costs payback

Record staff time; no improvement cost or payback measured.

Ownership

No claim about ownership or trades.

Hiring

No hiring claim.

Government

No procurement or public subsidy claim.

Language

Original English sources; no results transplanted to other markets.

Optional research / Checks 11–20

What would change the reading

Behaviour

Answers and mentions are not purchases.

Contradiction

An assistant can name the shop and show the wrong offer.

Missing proof

No matched customer checkout or repeat shop test verified.

Approval

An existing retail process, not a new product rollout.

Alternatives

A current public product page may already answer the question.

Geography

Run the check in one actual delivery area rather than comparing different markets.

History

Merchant documentation explains prior known problems; the research release is a distinct visibility comparison.

Awareness

Match the exact version and promotion before comparing displayed prices.

Countercase

If both shop records agree, an assistant mismatch does not prove the listing failed.

Action clock

Repeat at your next promotion start; stop if the offer cannot be matched.

Evidence & next review

Read the originals

Bain’s timed retail study release

Published 2026-11-20T11:15:00.000Z

Published 20 November at 06:15 EST (19:15 SGT). Reports sampled visibility and recommends a prompt audit. Does not independently verify sales caused by visibility.

Source record

Publisher: Bain & Company

Original language: English

Source publication date 2026-11-20T11:15:00.000Z

Describes mismatches when updates to stock on the shop page and product listing are not aligned. Google’s service only.

Source record

Publisher: Google

Original language: English

Describes differences involving sale prices and versions on one product page. Does not show what any assistant used.

Source record

Publisher: Google

Original language: English

Google documented price and stock for each product version in February 2024. This predates the Bain study and does not explain how other assistants get their information.

Source record

Publisher: Google

Original language: English

published date precision: day

Source publication date 2024-02-20

One timed research release. Google Merchant Center and Google Search guidance describe separate processes from one other organisation. They are not independent replications of Bain’s study or evidence of individual sales.

Consequences and economics

First-order consequence

Bain counted where retailers appeared across a defined sample of shopping questions.

Second-order consequence

A retailer can compare being mentioned with being described correctly for one offer.

Third-order consequence

If its own two public records disagree, the shop can identify which record needs checking before changing either.

Fourth-order consequence

If the same mismatch recurs at promotion starts, staff can review when they update the shop page and listing.

Fifth-order consequence

If the fix holds across later tests, continue the small check rather than assuming it created sales.

Observed costs and benefits

No fee, margin or sale was observed. Record staff minutes and the number of repeat errors before paying for outside help.

Challenge and uncertainty

Disconfirming evidence

The same product may pass every comparison, or the assistant may keep giving the wrong answer even when both merchant records agree.

What would confirm it

A time-stamped comparison showing the exact version, price, stock and link confirms that one answer was accurate in one test.

What would disprove it

If repeating the test shows no consistent record mismatch, do not assume the listing was the cause or pay for a fix on that basis.

AI powered research