Sample audit: real data, with every brand and product name replaced and answer quotes removed. Request the audit for your brand.
Brand A on ChatGPT and Google AI Mode
414 answers to 26 US shopping questions about robot vacuums (US), each asked three times a day in fresh, signed-out sessions. Every product shown was recorded with its position and price.
The short version
- Brand A products appeared in 81% of ChatGPT answers and 40% of Google AI Mode answers.
- They were the first product recommended in 47% and 21% of answers.
- Google AI Mode showed discontinued Brand A models in 18% of answers: products shoppers can no longer buy new.
- 24 of 171 prices shown for Brand A products did not match the retailer page they linked to.
- When Brand A was not the first pick on ChatGPT, it was most often Brand B model 1 (24% of those answers).
- When Brand A was not the first pick on Google AI Mode, it was most often Brand C model 1 (25% of those answers).
- 3 answer(s) advised against a Brand A product; see below.
Scorecard
| Surface | Appears in answers | Share of products shown | First product recommended |
|---|---|---|---|
| ChatGPT | 81% ± 10 pts | 41% ± 6 pts | 47% ± 11 pts |
| Google AI Mode | 40% ± 12 pts | 23% ± 5 pts | 21% ± 9 pts |
Ranges (±) are 95% uncertainty ranges that treat repeated runs of the same question as related. Answers change from run to run, so one answer is not a measurement.
Which Brand A products appear
| Product as shown | ChatGPT | Google AI Mode |
|---|---|---|
| Brand A model 1 | 31% (first 23%) | 5% (first 2%) |
| Brand A model 2 (discontinued) | not shown | 15% (first 5%) |
| Brand A model 3 | 15% (first 4%) | 1% (first 0%) |
| Brand A model 4 | 12% (first 5%) | 2% (first 1%) |
| Brand A model 5 | 4% (first 1%) | 10% (first 5%) |
| Brand A model 6 | 9% (first 3%) | 7% |
| Brand A model 7 | 7% (first 1%) | not shown |
| Brand A model 8 | 7% (first 1%) | 1% |
| Brand A model 9 | 3% (first 0%) | 5% (first 2%) |
| Brand A model 10 | 5% (first 1%) | not shown |
| Brand A model 11 | 4% (first 0%) | 3% (first 2%) |
| Brand A model 12 | 4% (first 0%) | not shown |
| Brand A model 13 | 3% (first 0%) | 3% (first 0%) |
| Brand A model 14 | 3% (first 1%) | not shown |
| Brand A model 15 | 3% (first 0%) | not shown |
Share of answers showing the product as a product card; in brackets, how often it was the first one.
7 of 32 current Brand A models we track never appeared, including Brand A model 49, Brand A model 50, Brand A model 51, Brand A model 52, Brand A model 60, Brand A model 62.
Discontinued models still recommended: Brand A model 2 (discontinued) on Google AI Mode (10 answers read); Brand A model 33 (discontinued) on Google AI Mode (3 answers read); Brand A model 47 (discontinued) on Google AI Mode (3 answers read); Brand A model 23 (discontinued) on Google AI Mode (2 answers read); Brand A model 55 (discontinued) on ChatGPT (1 answers read); Brand A model 55 (discontinued) on Google AI Mode (1 answers read).
Who takes your place
The first product recommended in answers where it was not a Brand A product.
ChatGPT
| First product instead | Share of those answers |
|---|---|
| Brand B model 1 | 24% |
| Brand B model 2 | 14% |
| Brand B model 3 | 9% |
| Brand B model 5 | 8% |
| Brand C model 2 | 8% |
Google AI Mode
| First product instead | Share of those answers |
|---|---|
| Brand C model 1 | 25% |
| Brand B model 1 | 10% |
| Brand C model 2 | 9% |
| Brand E model 1 | 9% |
| Brand G model 1 | 7% |
Prices shown
Product cards that link to a retailer were checked against that retailer's page within hours of the answer. Brand A: 147 matched, 17 card higher, 7 card lower, 60 with no readable price on the page.
| Product | Retailer | Card | Retailer page |
|---|---|---|---|
| Brand A model 22 | homedepot.com | $699.99 | $1599.99 |
| Brand A model 37 | bestbuy.com | $699.99 | $1149.99 |
| Brand A model 28 | bestbuy.com | $549.99 | $799.99 |
| Brand A model 28 | bestbuy.com | $549.99 | $799.99 |
| Brand A model 6 | Brand A store | $299.99 | $249.99 |
What assistants say about Brand A
Reasons given when recommending Brand A products, by theme (share of recommendations with a stated reason):
ChatGPT: pet hair and tangles (38%), navigation and obstacle avoidance (34%), mopping (24%), price and value (18%).
Google AI Mode: navigation and obstacle avoidance (39%), pet hair and tangles (36%), suction and cleaning power (27%), dock and self-maintenance (17%).
Advised against
- Brand A model 8 on ChatGPT: not for a messy floor; can miss cords and cat toys
- Brand A model 4 on ChatGPT: obstacle avoidance to be inconsistent; cannot be relied upon to avoid pet waste
- Brand A model 11 on ChatGPT: I wouldn't choose it over another model
From a sample of 81 answers read in full; every finding quotes the answer (quotes are removed in this sample).
What a pilot would add
This audit uses a standard question set for the category. A Recovery Pilot runs your own shopping journeys for your priority products, checks every claim against your catalog, ranks what is costing you recommendations, and measures whether one agreed fix recovers them, against products you did not change.