The Difference Between Being Mentioned, Cited, Recommended, and Chosen by AI
The market treats these four behaviors as one thing. They are not. Each is a different level of AI confidence, requires different evidence, and responds to a different fix.
Research from the Axis Suite initiative. Part of the Four Levels of AI Visibility framework.
The AI visibility conversation treats “showing up in AI answers” as a single outcome. It is not.
When ChatGPT, Claude, Gemini, or Perplexity responds to a buyer query about your category, your brand can appear in four fundamentally different ways. Each represents a different level of AI confidence. Each requires different evidence. Each responds to different interventions.
Most brands measure whether they appear at all. Very few diagnose how they appear. That distinction determines whether your next investment should be in crawlability, content, independent evidence, or category alignment.
The Four Levels of AI Visibility framework defines these four states: Mentioned, Cited, Recommended, and Chosen. What each looks like in practice, what evidence drives the transitions, and what determines which level you occupy — that’s the whole game.
The four levels of AI visibility
Each level is a higher degree of AI confidence than the last — and each is earned with a different kind of evidence, not simply more of the same.
Level 1 — Mentioned: AI knows you exist
What it looks like
AI knows your brand exists and occasionally includes your name in a response, but the mention is surface-level — your name appears alongside several alternatives with no emphasis, differentiation, or confidence signal. Ask ChatGPT “what are some platforms for [your category]?” and you may appear in a list of six or eight options: no elaboration, no positioning, no reason given. On Claude you might appear once and then disappear on the follow-up. On Perplexity your name shows in the answer text, but the cited sources do not include your own content.
What evidence drives this level
Mentioned status usually means AI has encountered your name across enough sources to include it in broad category responses, but has not formed specific beliefs about what you do or how you compare. The evidence is indirect — industry lists, directory entries, third-party articles that mention multiple vendors. Enough signal to include your name; not enough to describe you with confidence.
What keeps brands stuck here
Two structural issues. Weak retrieval: AI cannot access your website directly, so it relies entirely on third-party references that mention your name without context — the fix is crawlability. Category ambiguity: sources describe you inconsistently (a marketing tool here, an analytics platform there), so AI cannot form a stable association and defaults to a bare mention.
Level 2 — Cited: AI can describe you
What it looks like
AI references your brand with specific context — not just your name, but what you do, who you serve, or what makes you relevant. You are no longer just on a list; you are described. When ChatGPT compares options, you appear with a brief description of your positioning and at least one distinguishing characteristic. When Claude builds a comparison framework, you are included with differentiating attributes. When Perplexity responds, it cites your own content — or content specifically about you — as a source.
What evidence drives this level
Cited status means AI has accessed enough information to form initial beliefs about your positioning, category, and capabilities. It can describe you, not just name you. The evidence typically includes direct access to your website content (retrieval is working), a clear enough category signal that AI can classify you, and at least some consistency across sources about what you do.
The gap between Cited and Recommended
Many brands reach Cited and assume their AI visibility is working — they appear in answers, they are described with some accuracy, the dashboard looks reasonable. But Cited is not Recommended. Being described by AI is different from being selected by AI, and this is where most brands experience the most frustration: their metrics look healthy while recommendation behavior stays weak. Moving from Mentioned to Cited requires AI to access your information. Moving from Cited to Recommended requires AI to trust it.
Level 3 — Recommended: AI selects you with confidence
What it looks like
AI actively selects your brand when buyers ask for suggestions, comparisons, or evaluations — not just listed, not just described, but selected as a recommended option with confidence. Ask ChatGPT “which platform should I choose for [your use case]?” and you appear among a small number of recommended options with specific reasons. On Claude the recommendation persists through follow-ups — when the buyer narrows the criteria, you survive the filter. When Gemini structures a vendor scorecard, you appear with positioning that reflects genuine differentiation.
What evidence drives this level
Recommended status requires independent corroboration — AI has moved beyond your own content and found evidence from sources it trusts: G2 reviews where customers describe you in language consistent with your positioning, analyst reports that independently name you, comparison articles that articulate your differentiation, case studies on third-party platforms. This is where the distinction between content and evidence becomes critical. Content is what your brand publishes about itself; evidence is what independent sources say about you that AI can verify. Recommended status requires evidence, not just content.
What keeps brands from reaching this level
Three factors. The evidence gap (most common): your content is strong and accessible, but independent sources do not corroborate your claims. Category misalignment: your content says “AI recommendation intelligence” but AI files you under “SEO tools,” placing you in the wrong competitive set. Narrative inconsistency: your website, G2 profile, and LinkedIn each emphasize a different positioning, so AI encounters conflicting narratives and lowers its confidence rather than recommending a brand it cannot describe consistently.
Level 4 — Chosen: AI reaches for you by default
What it looks like
AI treats your brand as a default in your specific context — not one of several options, but the first one mentioned, the one described with the most confidence, the one that survives every refinement of the query. Add constraints on ChatGPT (“for a team under 50 people,” “for a mid-market company”) and you keep appearing while others drop off. When Claude builds a decision framework, your brand anchors the evaluation — the benchmark others are compared against. When Perplexity curates a multi-source response, your content is among the primary sources shaping it. This is Recommendation Persistence: the brand AI reaches for consistently across query variations and levels of specificity — not because it was programmed to prefer you, but because the accumulated evidence makes you the most confident recommendation.
What evidence drives this level
Chosen status requires the deepest evidence foundation — everything from Recommended plus two factors. Evidence depth across contexts: AI has evidence connecting your brand to specific use cases, company sizes, and industries, so when a query gets specific it has specific evidence to keep recommending you. Evidence durability over time: AI updates its beliefs on ongoing evidence, so Chosen status requires that yours stays consistent and current. Outdated information, stale case studies, and abandoned profiles create evidence decay that gradually weakens recommendation confidence.
What separates Chosen from Recommended
Not volume — specificity and durability. Recommended brands appear when queries are broad; Chosen brands appear when queries are specific. Recommended brands show up in “who are the options” queries; Chosen brands survive “which one should I pick for my exact situation” queries. That survival requires evidence connecting your brand to narrow buyer contexts — case studies from specific industries, customer proof from specific company sizes, use-case documentation for specific problems. Each piece extends the range of queries where AI maintains its recommendation.
The engineering implications
Each level transition responds to a different structural intervention.
- Invisible → Mentioned requires fixing retrieval — crawlability, structured data, basic web presence. A purely engineering fix.
- Mentioned → Cited requires fixing clarity — consistent category language, clear positioning on accessible pages, structured information AI can extract. A collaboration between marketing and engineering.
- Cited → Recommended requires building evidence — G2 reviews, analyst coverage, comparison-article inclusion, independent validation from sources AI trusts. A strategic investment in third-party evidence.
- Recommended → Chosen requires deepening evidence — context-specific proof, durable and current information, sustained consistency across every source AI draws from. An ongoing engineering discipline.
Notice the pattern. The first transition is purely structural. The middle transitions require collaboration between marketing and engineering. The final transition is continuous maintenance.
AI visibility is not one problem. It is four different problems, each requiring a different diagnosis and a different fix. Understanding which level your brand currently occupies determines what you should invest in next.
How to diagnose your current level
A practical test across ChatGPT, Claude, Gemini, and Perplexity. Each diagnostic question maps to a level, and each level maps to a specific structural fix.
Ask each platform to list vendors in your category. If your brand does not appear, you are below Mentioned. Fix retrieval first.
Ask each platform to describe your brand specifically. If it can name you but not describe you, you are Mentioned. Fix clarity and category consistency.
Ask each platform to recommend a solution for your use case. If your brand is described but not recommended, you are Cited. Build independent evidence.
Ask each platform the same question with added constraints (specific industry, company size, use case). If your brand drops out when queries get specific, you are Recommended but not Chosen. Deepen context-specific evidence.
Frequently asked questions
What is the difference between being mentioned and being recommended by AI?
Being mentioned means AI includes your brand name in a response without elaboration or confidence. Being recommended means AI actively selects your brand with specific reasons and confidence signals. The difference is driven by independent evidence from sources AI trusts, not just awareness of your brand.
Why does my brand get cited but never recommended by ChatGPT or Claude?
The most common cause is an evidence gap. AI can access and describe your information (cited), but lacks independent corroboration from G2 reviews, analyst reports, or comparison articles to recommend with confidence. The fix is building evidence in third-party sources, not publishing more content on your own site.
What does Recommendation Persistence mean?
Recommendation Persistence means AI continues recommending your brand even as the buyer query becomes more specific or the conversation evolves. Brands with strong persistence survive when constraints are added. Brands with weak persistence appear in broad queries but drop out when buyers narrow their criteria.
Can I move directly from Mentioned to Chosen?
Typically no. Each level requires different evidence foundations that build on the previous level. Skipping from Mentioned to Chosen would require simultaneously fixing retrieval, clarity, independent evidence, and context-specific depth. A sequential approach, diagnosing and fixing one level at a time, produces more reliable results.
How do different AI platforms evaluate these levels differently?
ChatGPT tends to favor entities with strong definitional clarity and corroborating sources. Claude shows higher sensitivity to nuanced positioning and differentiation. Perplexity relies more on indexed source material and real-time web access. Gemini weights authority signals and structured data. Each platform may place your brand at a different level, which is why testing across all four matters.
How long does it take to move between levels?
Moving from invisible to Mentioned can happen in days after fixing crawlability. Moving from Mentioned to Cited typically takes weeks as AI processes your content. Moving from Cited to Recommended can take weeks to months depending on how quickly independent evidence accumulates. Moving from Recommended to Chosen is an ongoing process that deepens over time with sustained evidence investment.
Find out which level you occupy
Axis Suite diagnoses where your brand sits across Mentioned, Cited, Recommended, and Chosen — which sources AI cites for competitors, who persistently holds your spot, and the single next thing to fix to move up a level.
Related reading: