Brand monitoring

The AI Image Score: how Deep Narrative Intelligence measures what models say about you

October 24, 2026 6 min read
Kamel Malek Director of SetRecord.ai

Every Deep Narrative Intelligence engagement produces one number that matters more than any other: the AI Image Score. It is the metric that turns "what does AI say about us?" from an open-ended anxiety into a tracked, trendable, board-presentable signal. Here is how it works.

Beyond sentiment

Traditional brand-monitoring tools produce sentiment scores: positive, neutral, negative. That framework was built for social media, where the question is "how do people feel about us?" In the generative-AI layer, the question is different: "are the facts the model states about us correct?" Sentiment can be positive and factually wrong — a model that says your company is "a thriving leader with €2B revenue" sounds great, until you realise your actual revenue is €400M and a prospect just made an investment decision based on that inflated figure.

The AI Image Score ignores sentiment. It measures factual accuracy: what share of the verifiable claims a model makes about your organisation can be confirmed against primary sources? That is the only metric that matters when a model's answer becomes someone's reality.

How the score is computed

Each interrogation cycle produces a set of factual claims from each model. We classify every claim into one of four categories:

  • Verified — the claim matches a primary source exactly.
  • Approximate — the claim is directionally correct but contains a factual imprecision (wrong date, wrong figure within 20%, outdated title).
  • False — the claim contradicts a primary source.
  • Unverifiable — the claim cannot be checked against any source (e.g., a subjective characterisation or a prediction).

The AI Image Score is the weighted ratio: verified claims score full, approximate claims score partial, false claims score zero and pull the average down, unverifiable claims are excluded from the denominator. The result is a number between 0 and 100 — one that moves only when the factual record the model draws from actually changes.

Don't know what AI says about you today?

A SetRecord audit queries supported AI models about your organisation and measures the gap with verifiable facts — within five business days.

Request an audit

One score per model, per language

There is no single "AI Image Score" for your organisation. There is one per model and per language. ChatGPT may score 78 in English and 52 in French. Gemini may score 85 in English and 40 in German. These gaps are the most diagnostic output of the entire method, because they tell you exactly where to focus: a low score in German on Gemini means something specific is wrong in the German-language corpus that model reads.

This granularity is what makes the score actionable. An aggregate "your AI image is 67/100" is a headline; a breakdown that says "Gemini in German is at 40 because of three recurring false claims about your funding history" is an action plan.

What the curve tells you

A single score is a snapshot. The value of Deep Narrative Intelligence is the curve — the score tracked over consecutive interrogation cycles, across months. The curve answers questions a snapshot cannot:

  • Is the correction we published last month working? (If the score rises in the next cycle, yes.)
  • Is a new error emerging that we haven't detected yet? (If the score drops without a known cause, investigate.)
  • Are we improving in English but stagnating in Spanish? (Different languages, different sources, different correction strategies.)
  • Has the model's training cycle absorbed our corrections? (A step-change in score after months of stability suggests a new training run.)

From score to action

The AI Image Score is not a vanity metric. It is a trigger. When a score drops below a threshold the client has defined, an alert fires. When a score in a specific language-model cell stays below target for two consecutive cycles, a correction cycle is initiated. When a score jumps after a training cycle, the team confirms which corrections propagated and which didn't.

This is what it means for Deep Narrative Intelligence to be "deep": not just interrogating models, but turning the results into a measured signal that drives decisions — cycle after cycle, for six months, until the curve tells you the record has been set straight.

See what AI says about you

A SetRecord audit checks what supported AI models say against your verifiable facts — within five business days.