AI Recommends 93% of Brands Without Mentioning a Single Downside

Read any good product review and you will find the catch: the vacuum that struggles on thick carpet, the laptop with weak battery life, the insurance plan with a high deductible. Ask an AI engine for the best of the same products and the catch is usually gone. The answer lists five brands and says something good about each.

That matters to shoppers and to brands. A downside is often the most useful thing a recommendation can tell a buyer, because it shows who a product is wrong for. When AI answers leave it out, every product sounds like a safe choice, and the shopper has to find the trade-offs somewhere else.

AI engines describe 93.3% of the brands they recommend without a single downside. Google AI Mode is the most critical engine, and it still flags a weakness for fewer than one brand in ten. The few warnings that appear gather in streaming and software answers and around the brands lower in the list.

On September 23, 2026 we asked 2,000 "best product or service" questions in 20 categories, from kitchen appliances to insurance, to five AI engines: ChatGPT, Gemini, Perplexity, Google AI Mode and Google AI Overviews (the AI summary above Google's search results). In the answers we identified 69,214 brand recommendations, each the passage of text an answer writes about one brand.

A brand counts as getting a downside when its passage names any con, drawback, trade-off or weakness ("expensive", "lacks", "noisy"), or qualifies its praise with words such as "however" or "but". That is a generous test: a single "but" counts. We compared the rate by AI engine, by category and by the brand's position in the answer, where position 1 is the first brand named.

Fewer than one brand in fourteen gets any stated drawback, even though the test counts words as mild as "but". An AI shortlist reads as a list of good options, each with its strengths described and its weaknesses left out.

An AI answer gives each brand a few lines, and those lines go to its strengths. For a shopper, that means an AI answer tells you what each product is good at, but rarely what it is bad at or who should avoid it.

No engine is very critical. Google AI Mode (9.7%) and Gemini (8%) are the most likely to mention a weakness. ChatGPT (5.3%) and Google AI Overviews (4.9%) flag a downside for about one brand in twenty, and Perplexity, whose answers are short, for 2.6%, about one in 40.

The gap between engines means the same product can come with a warning on Google AI Mode and none on Perplexity, whose short answers leave room for a verdict and little else. Even on the most critical engine, nine brands in ten are recommended without a downside.

Streaming services (13.5%) and software (13.1%) draw the most warnings, about twice the overall rate. These are products where the trade-offs are easy to state: one streaming service has fewer films, another has ads; one app lacks a feature, another costs more per user. Services, kids products, insurance and kitchen products follow at 8% to 9%.

Beauty and health products are almost never criticized, at 2.5% each, with home and fitness products at 3%. In those categories an AI answer reads like a list of recommendations with no warnings, even though fit, skin type and personal needs decide whether a product works for a given buyer.

The brands AI names first are described with the fewest reservations. The first brand gets a downside 5.6% of the time and the second 5.5%. From there the rate climbs, to 6% at the third position, 6.9% at the fourth and 9.4% for brands named fifth or later.

The warnings gather around the alternatives, where the answer explains why a brand is a second choice: pricier, narrower, or suited to a specific need. For the brands at the top of the list, an AI recommendation is praise with almost no qualification.

For brands, the risk in AI answers is being left out, not being criticized. If you are named, you are very likely to be described favorably. Focus on getting named, and on the positive details the engine repeats: clear product pages and review coverage that state who your product is for and what it does best.

If you sell streaming, software, services or insurance, check what AI engines say about your trade-offs. Those are the categories where warnings appear, and an outdated complaint in a cited review can follow you into the answer. Ask the publisher to update it, and publish your own clear answer to the objection.

Shoppers should treat an AI shortlist as a starting point and open the cited sources to find the trade-offs. Working with an ai seo company that tracks how AI engines describe your brand, including any downsides they repeat, tells you which sources to fix first.

Embed this research

Cite this study