GEO report: Perception Affect Field | Veritas Links | GEO, AEO, AI Focus Groups - analyzes
VeritasLinks

Help center

Documentation

Guides for GEO reports, AI Focus Groups, scores, sharing, and the Partner API.

GEO report: Perception Affect Field

On a GEO report, Perception Affect Field answers: when assistants answer your structured brand prompts, what emotional field do those answers form? It is not mention rate and not VeritasScore. The map is built from the GEO answer grid—prompt × model response texts for this run.

GEO corpus — prompt × model answers

Affect Field on GEO scores assistant answers from your structured prompt grid (visibility, competitors, citations, rankings, …)—not buyer chat turns. Polarization here means models/prompts split emotionally about the brand.

  • ChatGPT · visibility

    “…rarely recommended for mid-market SOC teams…”

  • Perplexity · competitors

    “…stronger alternatives include Contoso and Acme…”

  • Claude · citations

    “…proof points thin versus category leaders…”

  • Grok · rankings

    “…occasionally listed second for workflow AI…”

What gets scored on GEO: assistant answers from the prompt grid.

What the analyst should ask

  • Do model answers about us cluster as one mood, or split into praise vs skepticism?
  • Is the skeptical mode tied to specific prompt categories (competitors, citations, rankings) or specific assistants?
  • Would a buyer who lands on the negative mode hear a different brand story than one who lands on the positive mode?

Where it appears

  • GEO dossier — Perception Affect Field tab (short label Affect), after Overview when affect_field exists.
  • Combined report — GEO · Perception Affect Field.
  • Funnel GEO view — may embed the same block inline.

Needs enough scorable answers (pipeline skips under ~5 texts). Trial/shared views may show a plan gate (split_perception_locked) until the full model set is unlocked.

Metrics on the tab

Affect field metrics

Valence × arousal map

Computed PNG in live reports

Valence
0.22
Arousal
0.61
Dispersion
0.48
CI radius
0.19
Dominant zone
tense / activated
Responses scored
186
Polarization index
0.72
Valence × arousal map plus metric cards.

Split perception alert

Split perception detected

Responses cluster in two modes — not a single average. Treat this as a reputational risk signal, not neutral consensus.

Amber alert when the GEO field is polarized.
  • Valence / Arousal — centroid of scored GEO answers on the affect plane.
  • Dispersion / CI radius — spread and uncertainty of that centroid.
  • Dominant zone — named region where most answer mass sits.
  • Responses scored — count of GEO answers included (often large: prompts × models).
  • Polarization index + Split perception detected — bimodal emotion in assistant answers, not buyer panel disagreement.
  • by_analyzer (under the hood) — counts per assistant; useful when diagnosing which model drives a mode.

How to act (GEO-specific)

  1. Open Mentions Explorer and filter prompts/assistants that sit in the skeptical cluster—copy the exact language assistants use.
  2. Cross-check Competitors, Source Tracker, and Rankings: polarization often tracks “not recommended / thin proof / rival preferred” answers.
  3. Fix proof and citation footprint so the next GEO run’s answers converge; monitoring may flag split_perception_changed.
  4. Do not treat this as Focus Group panel mood—re-run FG Affect Field separately if you need buyer-turn emotion.

Related