This article is part 4 of a 11 in a Series of article that will help your organisation rank with AI. Read the overview article “SEO Got You Found. AIO Gets You Recommended" for a broad understanding.
Ask an AI tool the same question repeatedly and the list of brands it recommends will keep changing. This isn't a bug in the model, it's how these systems fundamentally work and it changes what "success" should mean for your AI visibility strategy.
Why AI answers aren't fixed, like a Google ranking
A Google or any traditional search ranking is a relatively stable position, if you're #3 today, you're likely still near #3 tomorrow unless something changes. An AI-generated answer is different: it's assembled fresh, probabilistically, from whatever the model weighs as relevant at that moment. Two identical prompts, seconds apart, can produce two different answers. That means a single good result, that one time you showed up in ChatGPT, isn't evidence of anything durable on its own.
Different engines don't even read the same web
It goes further than randomness within one engine. Different AI platforms favour different kinds of sources. Studies of citation patterns show ChatGPT, Perplexity and Google's AI features each leaning on different sites. A brand that is cited often on one platform can be much less visible on another, so tracking only one engine leaves part of the picture unmeasured. Treating "AI search" as a single category to optimize for is the mistake underneath most disappointing AIO results.
What this means practically
Stop chasing a single good result. One appearance in one answer, on one platform, tells you almost nothing. Track patterns across multiple runs and multiple platforms instead. This is exactly what the three-question test from earlier in this series is designed to surface.
Don't assume one platform represents them all. A brand that shows up on one platform may not show up on another. If you only ever check ChatGPT, you may be missing where your actual buyers spend their time.
Expect volatility, and don't overreact to it. A bad result this week doesn't mean your last change failed. Look at trend lines over months, not single snapshots.
The upside of probabilistic answers
Though this feels like an overwhelming new landscape, here's the silver lining: because these systems reward consistency and pattern over a single lucky mention, brands that show up credibly and repeatedly; across their own site, reviews, and third-party coverage, compound their advantage over time in a way a single SEO ranking never could. The volatility that makes AI visibility hard to game in the short term is the same thing that rewards genuine, sustained clarity in the long term.
Sources
Search Engine Journal, report on SparkToro's research into how often AI recommendations change: searchenginejournal.com/ai-recommendations-change-with-nearly-every-query-sparktoro/
Profound, AI platform citation patterns: tryprofound.com/blog/ai-platform-citation-patterns
Semrush, most-cited domains in AI search: semrush.com/blog/most-cited-domains-ai/
