The Five Layers of AI Narrative Intelligence
AI visibility is often presented as one software category. It is not. The market contains several different kinds of evidence, generated through different mechanisms and capable of answering different questions.
One platform may estimate which narratives people bring to ChatGPT. Another may record which publications AI systems cite. A third may repeatedly run the same Decision Narrative across ChatGPT, Gemini, Claude, or Google AI and measure how often a brand appears. A fourth may analyze a deliberately selected corpus of authoritative evidence, including sources that the tested AI answers never cited. A fifth may monitor the much wider public conversation across news, social media, reviews, forums, blogs, video, and other external surfaces.
Those functions are related, but they are not interchangeable. Narr Theory organizes them into the P–A–R–S–E framework:
| Letter | Layer | Primary question |
| P | Prompt Narrative Intelligence | What narratives are people asking—or should the brand systematically test? |
| A | AI-Selected SoA Citation Patterns | Which visible sources, pages, and publishers do AI systems select around the narrative? |
| R | Response Patterns (Target Narrative Position) | What do AI systems say, and how consistently does the brand appear and receive the desired association? |
| S | SoA Corpus Semantic Association Mapping | What does a deliberately selected authoritative corpus support, contradict, or associate with the brand? |
| E | External Web Narrative Intelligence | What narratives are forming and spreading across the wider web, media, and public conversation? |
The first three layers—P, A, and R—primarily describe what can be observed around AI systems today. They reveal the questions, citations, and answers visible from the outside. The final two—S and E—analyze the upstream information environment that communications teams can investigate and influence.
This distinction matters because an AI response is an output, not an x-ray of a model’s internal state. A visible citation is observable evidence, not a complete explanation of why the model generated an answer. A semantic association found in an external corpus may help explain or anticipate AI behavior, but it does not prove that ChatGPT or Gemini holds an identical internal association.
The strongest AI product-recommendation program therefore connects all five layers: identify consequential Decision Narratives, observe the sources and answers produced around them, examine the authoritative evidence environment, monitor the wider web narrative, intervene through truthful communications and product improvements, and then measure the AI response again.
Figure 1. AI Product Recommendation Enablement: the PARSE technology stack.
How the framework classifies vendors
The classifications below are based on publicly documented product claims and methodologies available in September 2026. Core means the vendor publicly documents a material version of the capability. Partial means the vendor owns relevant data or performs a narrower version of the function, but its public product does not document the entire mechanism defined here. Capable means the technical machinery exists, although the customer may need to define the Source-of-Authority set and analytical method.
| Vendor | P | A | R | S | E |
| Profound | P1 | Core | Core | — | — |
| Semrush | P1+P2 (+P3) | Core | Core | Partial | Partial |
| Goodie AI | Claimed P1 + P3 | Core | Core | Limited | — |
| Ahrefs | P2+P3 | Core | Core | Partial | Partial |
| Evertune | P1+P3 | Core | Core | — | — |
| Peec AI | P3 | Core | Core | — | — |
| Centium | P3 | Core | Core | — | — |
| Gumshoe | P3 | Core | Core | Limited | — |
| Rankscale | P2+P3 | Core | Core | — | — |
| Handraise | P3 | Core | Core | Partial | Partial |
| Quid | — | — | — | Capable | Core |
| Brandwatch + Trajaan | P2+P3 | Core | Core | Capable | Core |
| Talkwalker | P3 | — | Core | Capable | Core |
| Signal AI | — | Core | Core | Partial | Core |
| Pulsar | — | — | Partial | Capable | Core |
P1, P2, and P3: three kinds of prompt intelligence
The old PARB framework divided Prompt Demand into P1 and P2. The updated label, Prompt Narrative Intelligence, is intentionally broader and warrants a third subdivision:
| Code | Prompt evidence | Meaning |
| P1 | Observed prompt narratives | Aggregated prompt activity observed from consenting users, panels, or another disclosed behavioral dataset. |
| P2 | Modeled prompt narratives | Demand inferred from search behavior, query fan-out, semantic reconstruction, referral data, or another proxy. |
| P3 | Constructed prompt narratives | Prompts selected by the customer, curated by the vendor, or synthetically generated for systematic testing. |
Clickstream and panel data should not be treated as synonyms. Clickstream describes the data—an ordered record of digital actions. Panel describes one way participants are recruited and sampled. A panel does not normally mean people sitting together in a controlled room. Participants may consent to measurement software and use their own devices normally for weeks or months. Clickstream can come from a panel, but it can also come from applications, extensions, websites, or data partners.
P1 also does not automatically mean that OpenAI or Google licensed private prompts directly to the vendor. Narr Theory has not found public documentation of such a feed for the vendors examined here. Where a company discloses a panel, clickstream source, or another collection method, that provenance is identified. Where it merely says “real prompts,” the uncertainty remains.
Explore the Five Layers of PARSE
Each layer of PARSE tech stack answers a different question about how brands and products appear in AI product recommendations.
The following guides examine each capability separately, including the vendors offering it, the mechanisms they use, and the limits of what their evidence can establish.
P — Prompt Narrative Intelligence
Key Question: what Decision Narrative patterns can we observe in prompts submitted by people to AI systems?
Objective: Identify the observed, modeled, or strategically constructed narratives that people ask—or may ask—AI systems.
Explore Prompt Narrative Intelligence tools →
A — AI-Selected SoA Citation Patterns
Key Question: What patterns emerge in the Sources of Authority (SoA) cited by AI systems when making product recommendations?
Objective: Track the SoAs, publications, pages, and domains that AI systems visibly select when answering Decision Narratives.
Explore AI-Selected SoA Citation Pattern tools →
R — Response Patterns (Target Narrative Position)
Key Question: What products do AI systems recommend, how do they justify those recommendations, and what other contextual patterns emerge?
Objective: Measure what AI systems say, how frequently a brand appears, how it is described, and how consistently it receives the intended narrative association.
Explore AI Response Pattern tools →
S — SoA Corpus Semantic Association Mapping
Key Question: How do the most relevant Sources of Authority (SoA) in our category describe and evaluate our brand and products?
Objective: Analyzes a deliberately selected corpus of authoritative sources to determine what evidence supports, contradicts, or associates a brand with a Target Decision Narrative.
Explore SoA Corpus Semantic Association Mapping tools →
E — External Web Narrative Intelligence
Key Question: What broader narratives surround our brand and products across the public web, beyond the core SoAs?
Objective: Monitors the narratives forming and spreading across news, social media, reviews, forums, video, and the wider public web.
Explore External Web Narrative Intelligence tools →
Together, P, A, and R reveal what can currently be observed around AI systems. S and E examine the wider evidence and narrative environment that organizations can investigate, strengthen, and influence.
Leave a Reply