As AI evolves from lab-curiosity to primary mode of discovery, Rand Fishkin’s new SparkToro research lays down a striking reality for marketers and ecommerce leaders: AI tools are wildly inconsistent when they recommend brands or products and you should not treat these outputs like SEO rankings.
That sounds unsettling at first. But when you view this chaos through the lens of Answer Machine strategy, it becomes a clarifying signal: brand visibility in AI isn’t about winning a fixed rank anymore, it’s about building a system that consistently answers real customer intents. Let’s unpack what Fishkin found, how Answer Machines account for this instability, and what brands must do next.
The Insight: Why AI Recommendations Behave Differently From Search Rankings
Fishkin and collaborators ran the same prompts many times through leading AI models. The result?
- AI tools rarely return the same list of brands twice (<1 in 100 runs).
- They almost never repeat the same ordering (<1 in 1,000 tries).
- The number of items returned can vary wildly from one run to the next.
In other words: AI recommendations are not stable, deterministic rankings like you see on Google SERPs. Instead, they are probabilistic outputs where every response is a unique sample drawn from a vast distribution of plausible text.
That’s core to how generative models work, but it also means ranking your brand #1 in AI tools isn’t a meaningful KPI the way an organic #1 rank on Google used to be.
Why AI’s Instability Matters for Answer Machine SEO Strategy
At Ecommerce Growth I’ve been talking about the Answer Machine as a strategic system, not a collection of blog posts. An Answer Machine is:
- a content ecosystem that maps customer intent;
- built to answer clusters of questions, not chase single keywords;
- designed to align content with how people actually search, ask, and decide; and (critically)
- structured so that AI systems cite and synthesise from your content when they generate answers.
Fishkin’s research sounds like bad news for visibility measurement but it’s actually proof that this strategy is the right one:
AI Answers Are Unstable, But Customer Intent Remains Stable
If AI answers vary each time, the only stable signal is whether your content consistently provides the right answers for real customer intents.
For example: if a shopper is asking “best workout shoes for wide feet”, it doesn’t help to aim for a fixed rank you need content that answers that exact intent clearly, repeatedly, and with authority.
Answer Machines do this by mapping intent clusters (check out KeywordInsights.ai for an easy way to do this) and building content around them which AI models are far more likely to synthesize and cite. Fishkin’s work proves why keyword-centric tactics don’t cut it in the AI age.
Stop Tracking AI Rankings and Start Measuring How Often and Where Your Brand Appears
One key takeaway from SparkToro’s research is that metrics promising precise AI visibility rankings are unreliable. Because the same prompt answered two hours later can produce a different result, a “position” in AI answers is basically noise.
That has huge implications:
- You can’t treat AI recommendations like search ranking positions.
- You can measure frequency and context of your brand appearing across many intents and prompts.
- You can analyse how often AI systems cite your content, especially in high-intent queries relevant to your ecommerce niche.
This aligns perfectly with how an Answer Machine strategy works: not chasing ranks but building topical authority at scale.
The Real KPI: How Well Your Content Answers High-Intent Customer Questions
A traditional SEO mindset would optimise for “ranking #1”.
The Answer Machine mindset optimises for being the best possible answer, because AI engines are trained to synthesise best from authoritative, relevant content.
That means:
✔️ Structuring content to directly answer specific, commercial or informational questions
✔️ Using clear, customer-centred language that matches how intent is expressed
✔️ Creating clusters of related content so that AI models recognise topical depth
✔️ Earning citations and backlinks that reinforce authority in your niche
This is called Answer Engine Optimisation (AEO) it’s not traditional SEO where you’re focusing in on independent keywords it’s a different optimisation logic focused on answers, not ranks. It’s still fundamentally SEO in terms of delivering the best, most relevant, authoritative content to your reader.
Three Essential Answer Machine Strategy Shifts for Ecommerce Leaders
Build Content Around Customer Intent Clusters, Not Search Positions
Your content strategy should start with what customers want to know, not what keywords you want to rank for. Cluster those intents into a web of answerable content… that’s the core of Answer Machine SEO.
Measure AI Visibility Patterns (Frequency, Context, Citations) Instead of Single Positions
Rather than tracking “position” in AI outputs, measure:
- how often your content surfaces for relevant intents;
- how many times AI systems cite your content as a source;
- and the quality of context in which your brand appears.
These patterns reflect true visibility in a world where outputs vary.
Build Topical Authority So Answer Engines Prefer Your Content as a Source
AI systems tend to synthesise from content that is authoritative, clear, and well-structured. Brands that invest in content with depth, clarity and real human problem solving will over-time become the default answer source that generative systems rely on, even if specific outputs vary from one run to the next.
Takeaway: Why Answer Machine Strategy Must Underpin Ecommerce Growth in the AI Era
Fishkin’s research doesn’t say “AI will replace search,” or “brand visibility is impossible.” What it does say is:
AI recommendation outputs are not a fixed leaderboard… they don’t behave like traditional search rankings.
For ecommerce brands, that’s exactly why Answer Machine strategies matter now more than ever.
Traditional SEO built traffic. Answer Machine strategies build visibility as defined by the ways customers actually ask and AI systems actually deliver answers.
If you’re still chasing ranks in AI outputs, you’re asking the wrong question.
But if you’re building an Answer Machine that consistently answers intent, you’re future-proofing your brand in the age of generative discovery.
Written By:
