Zicy · the truth layer

The AI Reality Score

The AI Reality Score is Zicy’s per-engine accuracy score: it measures the gap between what is true about a brand and what AI actually says about it, across the tracked prompt set.

01 Why it exists

Being named is not the same as being described accurately.

Most tools measure whether a brand was named in an AI answer. Zicy measures whether what AI said was true. Those are different questions, and the second one is the one that decides deals, because a buyer acts on the claim, not on the mention.

A hallucination or an outdated claim sitting in an AI answer is a buyer-decision risk, not a visibility metric. It reaches the buyer with the same confidence as a true fact, before anyone on your side sees it. The AI Reality Score exists to surface that gap early, so an inaccurate description can be corrected before it costs you the decision.

02 What it measures

Four signals in every answer.

Strength

Accurate descriptions

Claims AI makes about the brand that are true and current, the share of the answer that holds up against the facts. This is what a high score is made of.

Risk

Outdated claims

Statements that were true once but no longer are: an old price, a former product name, a discontinued feature. Accurate yesterday, wrong today.

Risk

Hallucinations

Claims with no basis in fact: an invented feature, a fabricated statistic, a relationship that does not exist. The engine states them with the same confidence as a fact.

Risk

Entity gaps

Places where AI cannot confidently say who or what the brand is, so it hedges, omits, or fills the gap with a guess. Low confidence reads as low presence in the answer.

03 How it is calculated

Scored on what AI says, per engine, over time.

  • The score is calculated per engine, across the set of prompts you track, because each engine describes a brand in its own way.
  • Scoring is based on the content of the AI response, what was actually said, not just whether the brand was present or absent.
  • A score of 100 means AI described the brand accurately across all tracked prompts on that engine. Lower scores reflect the proportion and the severity of the inaccuracies found, so one confident hallucination weighs more than a small omission.

Read it as a direction, not a verdict. AI-search visibility is probabilistic: results vary with query phrasing and the time window in which a prompt is run. The AI Reality Score is best read as a directional indicator tracked over time, not a single definitive snapshot. A reading that holds or improves across several cycles tells you more than any one week.

04 What pulls the score down

Five ways an answer goes wrong.

Hallucinations

Invented facts stated as confidently as true ones, the heaviest single drag on a score.

Outdated claims

Facts that were once right and have since changed, repeated by AI as if still current.

Wrong attributions

A fact, quote or source credited to the brand that belongs to someone else, or the reverse.

Missing key facts

The answer omits something material, so the picture is accurate as far as it goes but incomplete.

Competitor conflation

AI blends the brand with a rival, attributing one’s features, pricing or positioning to the other.

05 How to improve it

Give AI clear, current facts to read.

The score moves when AI can find clear, current, machine-readable facts about the brand. In practice that means a clean entity definition in schema, an llms.txt that states the facts plainly, and content that is accurate and up to date where AI reads it.

Zicy’s Action Center generates the schema and llms.txt and flags the content to correct, so the work of moving the score happens inside the same platform that measures it. See the full platform →

06 How it connects to demand

Accuracy removes hesitation at the moment of decision.

Accuracy is not an end in itself. A buyer who reads a confident, wrong claim about a brand mid-decision hesitates, or moves on. Reducing inaccurate descriptions removes friction at the moment the decision is forming.

Zicy’s Prove layer connects the visibility work to AI-referred sessions and branded-search behaviour through the Site Traffic module, so an improving AI Reality Score can be read alongside movement in demand, rather than as accuracy for its own sake.

FAQ

Questions about the AI Reality Score.

What is the AI Reality Score?

The AI Reality Score is Zicy’s per-engine accuracy score. It measures the gap between what is true about your brand and what AI actually says about it, across the prompts you track on each engine.

How is it different from a brand mention score?

A mention score counts whether AI named your brand. The AI Reality Score judges whether what AI said was accurate. You can be mentioned often and still score poorly if the claims attached to those mentions are wrong or outdated.

Why does my score differ between engines?

Each engine, ChatGPT, Gemini, Perplexity, Google AI Overviews and Google AI Mode, draws on different sources and weights them differently, so they can describe the same brand with different levels of accuracy. The score is calculated per engine for exactly this reason.

Can I improve my AI Reality Score?

Yes. The score improves when AI can find clear, current, machine-readable facts about your brand. Zicy’s Action Center generates the schema and llms.txt and flags the content to correct, which are the levers that move it.

How often is it recalculated?

It is recalculated on your tracking frequency, weekly by default, because answer engines evolve gradually rather than by the hour. Reading it over several cycles is more meaningful than reacting to a single week.

Is a score of 92 out of 100 good?

Generally yes, but read it as a direction rather than a grade. AI-search results are probabilistic and vary with phrasing and timing, so a score in the nineties that holds or climbs over several weeks matters more than any single reading. (92 out of 100 is used here as an illustrative figure, not a real result.)

What counts as a hallucination in Zicy's scoring?

A hallucination is a claim AI states as fact that has no basis in reality: an invented feature, a fabricated statistic, or a relationship that does not exist. Because it is stated as confidently as a true fact, it weighs heavily against the score.

See your AI Reality Score.

Find out whether AI describes your brand accurately, across all five engines.