Ask an AI for 'the best X' a few different ways — does it keep naming the same brands, or does the list scramble each time?

This explores how stable AI-generated brand recommendations are: if you ask an AI for 'the best X' several times, or phrase it a few different ways, do the same brands keep showing up? The most direct answer in the collection is: the exact list almost never repeats. SparkToro ran 2,961 responses and found that the same set of brands came back less than 1 time in 100, and the same brands in the same order about 1 time in 1,000 How consistent are AI brand recommendation lists across repeated prompts?. Two things drive this. Models generate text probabilistically, so each answer is a fresh draw. And real people phrase the same need in many different ways, which adds more variation on top. If you're used to search rankings, where you can hold position three for months, there's no equivalent 'rank' here to hold.

The list being unstable doesn't mean every brand has the same odds, though. Ranqo looked at 100+ brands across five AI engines and found a steep ladder: global brands were cited 73% of the time, mid-market brands 44%, and niche brands 11%. That pattern held even when the query didn't name any brand Do AI search engines favor well-known brands over others?. Put the two findings together: the exact list changes almost every time, but the big names keep coming back. The randomness mostly shuffles who fills the remaining slots and in what order. So the better question for a brand isn't 'what rank am I?' It's 'what share of answers am I in?'

Order matters more than it looks, because position in a list affects choices independently of quality. In a randomized audit of twelve LLM travel assistants, moving a hotel up or down the candidate list changed how often it was picked, by an amount worth about $12 a night, with every quality attribute held fixed Does list order alone shift which hotel an LLM recommends?. Since order rarely repeats, part of each brand's visibility comes down to luck from one answer to the next, and the brand can't see or contest it.

An older thread from classic recommender systems helps explain why the big names stick. Recommenders that squeeze users and items into too few dimensions tend to overfit toward popular items, and niche items get less exposure over time, which makes the gap self-reinforcing Does embedding dimensionality secretly drive popularity bias in recommenders?. The same pressure plausibly shows up in LLMs, which absorb how often each brand appears on the web.

One caution if you want to measure whether these mentions matter. A clickstream study found that simply adding up every time an AI mentioned a brand roughly doubles the apparent effect on searches and site visits. Much of that is people who were already customers, or mentions that were only incidental Does pooling AI brand mentions mask the true recommendation effect?. The gap in the collection: none of these sources tracks one brand across many truly different queries, such as 'best running shoes' vs. 'shoes for flat feet', so how far a brand's presence carries across topics is still an open question here.