How often does AI recommend a brand with reservations?

AI added a reservation to 192,153 of 1,946,450 reviewed ranked-brand appearances, or 9.9%. A ranked-brand appearance is one consolidated brand with a direct recommendation rank and reviewed language in one observed AI answer.

A recommendation is not always an unqualified endorsement. Track whether the answer attaches a condition, fallback, comparison, budget limit, risk warning, or other reservation to the brand it ranks.

Among the 192,153 reserved appearances, 77,004, or 40.1%, carried no specific criticism. Only 7,308, or 3.8%, explicitly recommended against the brand for the stated need. The remaining 184,845 reserved appearances were not rejections.

BrightEdge reports negative sentiment in 2.3% of Google AI Overviews brand mentions and 1.6% of ChatGPT brand mentions. This study measures conditional recommendation language on different engines and at a ranked-brand grain, so those rates are not comparable. A sentiment or rejection audit cannot substitute for a reservation audit.

A condition appeared in 102,517 reserved brand appearances, or 53.4%. A fallback appeared in 46,083, or 24.0%. At least one of those two labels appeared in 148,228 of 192,153 reserved appearances, or 77.1%.

Most reservations describe fit rather than a blanket warning. Store the type and the buyer need with the brand so a conditional fit is not reported as general criticism.

ChatGPT Search added reservations to 141,835 of 951,620 reviewed ranked-brand appearances, or 14.9%. Google AI Mode did so in 50,318 of 994,830, or 5.1%. The ChatGPT Search rate was 2.9 times as high, a 9.8-point difference.

Use separate engine baselines. The aggregate rate hides the largest split in the study and does not show that either engine's reservation is factually correct.

Reservations appeared in 17,314 of 344,108 first-place brand appearances, or 5.0%; 55,266 of 596,811 second- or third-place appearances, or 9.3%; and 119,573 of 1,005,531 appearances at fourth or lower, or 11.9%. The fourth-or-lower rate was 2.4 times the first-place rate.

Compare brands at similar recommendation ranks. This pattern does not show that rank caused the reservation, but rank changes the descriptive baseline by 6.9 percentage points.

Among brands with at least 500 reviewed appearances, USAA carried reservations in 272 of 720 appearances, or 37.8%. Zuora followed with 230 of 738, or 31.2%, and HelloFresh with 219 of 740, or 29.6%. Seven more cleaned root-brand names complete the leaderboard.

This table identifies where to review recommendation language. It is not a brand-quality ranking. Compare the exact prompts, needs, reservation types, engines, and recommendation ranks before drawing a conclusion about a brand.

Among industries with at least 10,000 reviewed appearances, Software recorded 12,993 reserved appearances among 98,602, or 13.2%. Collaboration recorded 1,327 of 10,571, or 12.6%. Manufacturing recorded 903 of 14,627, or 6.2%.

Use an industry baseline before calling a brand's reservation rate unusual. These breakdowns describe the observed prompt corpus and do not show that industry caused the difference.

The main appearance-level rate was 192,153 of 1,946,450, or 9.9%. Counting reviewed statements instead returned 195,325 of 2,049,184, or 9.5%. Reviewed language was available for 1,946,450 of 3,582,303 direct ranked root-brand appearances, or 54.3%.

The similar statement-level rate supports the headline at a second grain. The denominator excludes ranked brands without reviewed language. Sparse reviewed rows after August 14 were retained in the Parse index, but the study makes no time-trend claim. It does not measure factual accuracy, source support, causation, buyer opinion, or user behavior.

Reservations appeared in 192,153 reviewed ranked-brand appearances. ChatGPT Search's rate was 2.9 times Google's, and the fourth-or-lower rate was 2.4 times the first-place rate.

Track recommendation rank, reservation type, buyer need, criticism, and explicit rejection separately. Compare the same priority prompts on both engines. Review high-rate brands and industries at the underlying answer level. Check factual and source support before changing messaging. Repeat the fixed method next quarter before calling any difference movement.