A Brand Can Win One of These and Lose the Others
A brand can be recommended by ChatGPT, Claude, Gemini, or Perplexity and still lose the sale. It can be chosen by a customer without anyone tracing that choice back to a specific AI recommendation. And now, for the first time, it can be acted upon directly inside an AI system, a conversation started, a task completed, without a person doing the choosing in between at all.
There’s a four-stage framework worth starting from: Mentioned, Cited, Recommended, Chosen. Recent product news made it clear there’s a fifth stage most AI visibility reporting isn’t tracking separately: Acted Upon.
Why a Fifth Stage Exists Now
OpenAI’s Sponsored Agents, launched September 16, 2026, let a person go from a labeled ad directly into a conversation with a business, then click through, without ever asking ChatGPT an independent question first. Salesforce’s AIforce, introduced at Dreamforce ’26, lets an AI agent execute inside a company’s own workflows from an interface outside Salesforce’s own product. Neither of these is a recommendation event in the traditional sense. Both are closer to action, and both can happen without a recommendation or a choice occurring first in any way a brand could trace or measure.
The Assumption Most AI Visibility Reporting Makes
Most reporting on AI visibility treats these stages as one continuous line: enough mentions lead to citations, enough citations lead to recommendations, enough recommendations lead to a choice. It’s a reasonable-sounding assumption. It’s also not something that’s actually been demonstrated to hold all the way through, and the last several weeks of product news across the industry are a reason to stop assuming it does by default.
A brand can be heavily cited and rarely recommended, because citation and recommendation measure different behavior. Being referenced as a source is not the same as being presented as the answer. A brand can be recommended constantly and rarely chosen, because a recommendation only becomes a choice once a person acts on it, and people don’t always do the obvious thing. And now a brand can be acted upon inside an AI system without a clean recommendation event ever happening first, if the action started from a sponsored entry point instead of an organic one.
What This Means for Measurement
None of this means the four-stage framework was wrong. It means it was built for a market that didn’t yet have a fifth stage. Recommendation is not a verdict, it’s accumulated confidence, and that framing hasn’t changed. Acted Upon sits downstream of it as a separate kind of event, sometimes built on a recommendation, sometimes not.
For anyone reporting an AI visibility number to their own leadership, the practical question is which of these five events the number actually describes. A high mention count across ChatGPT, Claude, Gemini, and Perplexity and a high action count are not the same achievement, and reporting them as if they were is how a technically true number ends up implying something it can’t actually support.
Frequently Asked Questions
What are the five stages of AI recommendation visibility?
Mentioned, Cited, Recommended, Chosen, and Acted Upon. Each represents a different distance between a brand and a customer decision, and a brand can score high on one stage while scoring low on another.
What is the difference between being recommended and being chosen?
Being recommended means an AI system, such as ChatGPT, Claude, Gemini, or Perplexity, presented a brand as an answer to a user’s question. Being chosen means a person actually acted on that recommendation, which doesn’t happen automatically just because a recommendation occurred.
What does “Acted Upon” mean as a separate measurement stage?
It refers to a brand being directly engaged with inside an AI system, through a conversation, click-through, or completed task, sometimes without a distinct recommendation event happening first. Features like OpenAI’s Sponsored Agents and Salesforce’s AIforce make this stage newly measurable and newly important.
Can a brand be acted upon without ever being recommended?
Yes. If a person enters a sponsored conversation through a labeled ad and completes an action there, that can happen without the AI system ever presenting the brand as an organic recommendation first.
Why shouldn’t a brand treat a high mention count as proof of strong AI visibility?
Because a mention count measures only one of five distinct stages. A brand can be mentioned frequently by ChatGPT, Claude, Gemini, or Perplexity without being cited as a credible source, recommended as an answer, chosen by a customer, or acted upon in any measurable way.
Does this five-stage framework apply the same way across ChatGPT, Claude, Gemini, and Perplexity?
The stages apply conceptually across all four systems, though each platform is building different mechanisms, like sponsored conversations or agent-executed actions, that affect how and when a brand moves between stages. Measuring accurately requires accounting for those platform-specific differences rather than assuming one number works everywhere.
ABOUT AXIS SUITE
Axis Suite is an AI Recommendation Intelligence platform built under TrendAxis LLC. It measures and explains how brands are mentioned, cited, recommended, and chosen across AI systems including ChatGPT, Claude, Gemini, and Perplexity, then helps brands understand what to do about it with evidence, not guesswork. Axis Suite is the independent intelligence layer that explains what AI believes about your brand, why it believes it, and what decision that belief ultimately drives. Learn more at axissuite.ai or visit the Proof Center for real, disclosed examples of the platform’s own evidence work.