Gemini Declines to Pick a Winner in 15.3% of Product Questions vs 4.2% for Google AI Overviews

Ask an AI engine for the best credit card and there is a good chance it will not tell you. It lists several cards, explains that the right one depends on how you spend and whether you carry a balance, and leaves the choice to you. Ask for the best blender and it is far more likely to name one.

For brands, that difference changes the game. When an AI engine crowns a winner, that brand gets the recommendation and everyone else in the list is a footnote. When it says "it depends", every brand it names gets similar weight, and the reader keeps researching, which gives challengers a way in.

Gemini is the most cautious engine. It answers "it depends" without naming any pick in 15.3% of its answers to "best" questions, more than three times the 4.2% rate of Google AI Overviews. Across all five engines, 9.4% of answers decline to pick, and the rate climbs in pets, finance and software.

On September 23, 2026 we asked five AI engines, ChatGPT, Gemini, Perplexity, Google AI Mode and Google AI Overviews, 2,000 "best" questions such as "best dog food for puppies" or "best budget laptop", across 20 categories, from a US location. ChatGPT, Gemini and Google AI Mode answered each question twice, Perplexity once, and AI Overviews appeared on 3,524 of the searches, for 17,524 answers in all.

An answer declines to pick when it says the best choice depends on the reader, with phrases like "depends on", "no single best" or "one-size-fits-all", and never names an overall pick with phrases like "best overall", "top pick", "I'd recommend" or "winner". An answer names an explicit pick when it uses one of those pick phrases. An "it depends" qualifier is any answer that includes a depends phrase, whether or not it also names a pick.

The engines fall into two groups. Gemini (15.3%), ChatGPT (10.3%) and Google AI Mode (9.6%) decline to pick in about one answer in ten or more. Perplexity (4.8%) and Google AI Overviews (4.2%) do so in fewer than one answer in 20, and almost always offer the reader a direction.

Hedging is more common than outright refusal. Across all engines, 24.5% of answers include an "it depends" qualifier, about one in four, even when many of them go on to name a pick. The engines are reluctant to sound certain, but most of the time they still point somewhere.

Declining to pick is one side of the picture; naming a winner is the other. Perplexity names an explicit pick, such as "best overall" or "top pick", in 84.6% of its answers to "best" questions, more than two and a half times ChatGPT's 32.1%. Google AI Overviews names a pick in 68.1% of its answers and does so with no "it depends" qualifier in 61.4%, the highest rate of clean, unqualified picks of any engine.

Across all five engines, 53.1% of answers name a pick, and "best overall" is by far the most common way they do it: it is the first pick phrase in 6,645 answers, against 943 for "top pick" and 738 for "top choice". On Perplexity and AI Overviews the prize is that single label. On ChatGPT, which names a winner in about one answer in three, the goal is a place on the shortlist it lays out.

The category matters as much as the engine. AI answers about pet products decline to pick 16.6% of the time, finance 15.2%, software 14.2%, health 13.6% and insurance 13.5%. These are questions where the right answer depends on personal circumstances: a pet's breed and health, a person's credit and spending, a company's size and existing tools.

Finance stands out. AI engines name an explicit pick in just 21.1% of answers about the best financial products, the lowest of any category, and add an "it depends" qualifier in 27.3%. Insurance carries the most qualifiers of all, 35.3%. In regulated money categories, AI engines lay out options and conditions rather than recommending one provider.

Physical products with clear specifications get the most decisive answers. AI answers about home products decline to pick 4.3% of the time and name an explicit pick in 75.2%; kitchen products name a pick in 77.5%. Gaming (4.4%) and office products (4.8%) also rarely get an "it depends" answer.

In these categories, the engine picks and the reader follows. Being the named pick for "best air fryer" or "best office chair" is worth far more than being one of the alternatives listed after it.

If you sell financial products, pet products, software or insurance, AI engines are less likely to name a single winner in your category, and that favors challengers. Instead of competing to be the best overall, compete to be the best for a specific situation: the best card for frequent travelers, the best food for senior dogs, the best CRM for a five-person team. When an engine says "it depends", it is telling the reader the answer changes by situation, and each situation is a recommendation you can win with a page built for it.

In categories where AI picks decisively, like home, kitchen and gaming products, being the named pick matters far more. Publish clear comparisons, specifications and evidence from real customers that make the case for your product as the default choice. Suff Digital, an ai seo company, helps brands find the situations where they can be the recommended option and build the pages that win them.

Embed this research

Cite this study