LLM Perception by CARMA
LLM Perception is CARMA's analyst-led research into how AI platforms represent your organisation. Our Insights Consultants measure your visibility, positioning, and strategic opportunities in AI, connect them to your earned, owned, and social media, and deliver recommendations you can act on. Paired with AI Visibility, it gives you insights, analysis, and recommendations, market by market.
The opportunity
The influence of earned media on AI answers brings significant opportunities, and some risks, for PR and communications teams.
84%
of AI citations come from earned media. Is yours cutting through, and how do you compare with your peers?
See whether your coverage is being cited in AI answers, and how you compare with your peers.
AI platforms cite different sources, so your LLM media list may look different for each one.
Identify inaccurate responses and misinformation, and see whether historical issues are resurfacing.
Pinpoint the technical and content optimisation needed to lift visibility and message resonance.
Source: Muck Rack, What Is AI Reading?, May 2026.
Works with AI Visibility
AI Visibility tracks how your organisation appears in AI answers inside the CARMA Insight platform. LLM Perception adds CARMA's Insights Consultants, who interpret that picture alongside your earned, owned, and social media and tell you what to do next.
Ongoing tracking in CARMA Insight
Analyst-led research and recommendations
Insights, analysis, and recommendations, market by market
Beyond vanity metrics
CARMA treats AI answers as an important communication channel, and like any channel, it can and should be measured to improve effectiveness and inform strategy.
We look at three things together: the upstream reputation built by your earned, shared, and owned media, the downstream answers AI gives about you, and your search and content readiness. That shows you which channels, topics, and activities influence AI answers, and how to strengthen your position.
Our approach combines the latest technology, deep research expertise, and AMEC's GEO Principles, putting those principles into practice in a way that's easy to act on.
Earned, owned, and social
• Share of voice
• Key themes and narratives
• Tonality, with positive and negative drivers
• Key message pull-through
• Spokesperson visibility
• Leading media, themes, and topics by tonality
LLM metrics
• Share of voice
• Comparison and recommendation positioning
• Peer comparison win/loss ratio
• Key messages and narratives
• Tonality
• Citations, types of citation, and leading citations
• Topics
The right metrics
Beyond visibility, LLM Perception shows how you're positioned in AI answers, how credible you appear, and where positive and negative answers come from.
We apply a consistent coding brief across channels, so AI results can be compared and benchmarked against your earned and owned media. The goal is to understand how your earned and owned activity influences your visibility and positioning in AI.
Personas and markets
We build personas for the audiences and stakeholders that matter to you, and use role-based and persona prompting to see the responses they're likely to receive from AI platforms.
You choose the geographic markets to test, and results are broken down by persona and market.
How it works
CARMA takes time to understand your objectives and tailors the approach to your needs. Many teams play a role in GEO, so the first step is making sure every stakeholder benefits from the research.
We agree with you which geographic markets, audiences, and types of questions to test, such as informational, commercial, and navigational questions.
Our Insights Consultants build prompts for your personas and markets, using search, social, and auto-complete trends. A narrow prompt list can skew results, so we build it carefully and agree it with you upfront.
We run a defined set of prompts against each AI platform at an agreed frequency and capture how each one responds.
For each response, we assess visibility, tonality, citations, content types, narratives, messages, and competitive comparisons, and how each varies across AI models.
We benchmark the AI results against your earned media coverage, social media presence, and owned content, so you can see where the AI picture lines up with reality and where it doesn't.
We present the findings and align on next steps with you in an Insights Roundtable.
Each answer is analysed as a piece of content in the CARMA Insight platform, using your earned media coding brief so results are measured consistently.
Findings are human-validated by CARMA's Insights Consultants.
AI platforms
Different platforms draw on different sources and can describe your organisation in different ways. LLM Perception shows you where those differences lie and what they mean for your strategy.
ChatGPT
Perplexity
Claude
DeepSeek
Google AI Overviews
Gemini
Meta AI
Grok
Why CARMA
Every report focuses on two questions: so what, and now what? Our Insights Consultants present recommendations and strategic opportunities to strengthen your visibility and message resonance in AI.
Recommendations from our Insights Consultants
Earned, social, owned, and AI in one analysis
The answers your stakeholders are likely receiving
Prompts, frequency, and scope set to your needs, as part of your CARMA media analysis or as stand-alone research
What you'll learn
Examples of the insight CARMA delivers, from share of voice to message resonance.
Share of voice on each AI model, benchmarked against peers
How you're portrayed, and what drives positive and negative answers
Which content and articles AI cites, by persona, market, and topic
Which key messages surface in AI, compared with the media
Talk to our team about an LLM Perception engagement, as part of your CARMA media analysis, alongside AI Visibility, or as stand-alone research.
