Audit whether AI can find, understand, choose, and prove a brand.

VectorGap gives agencies a repeatable method for diagnosing whether a brand is present, understood, preferred, and retrievable in the exact market where it competes. The methodology should not dead-end in education: it should lead to a report the buyer can inspect, pricing proof that reduces risk, and Agency OS capacity to run the loop.

See the methodology in a sample report Review proof before payment

Perception shows what AI answers currently say: how systems describe, trust, price, and recommend the brand.
Preference shows which brand AI chooses when the client is compared with named competitors, and which proof gap made the choice easier.
The Generative Brand Index summarizes Presence, Perception, Preference, AI Readiness, missing evidence, and confidence in one executive 0-100 view.
Presence and AI Readiness explain whether the brand appears for buyer problems and whether current market proof can be retrieved, cited, and defended; missions, retests, and exports turn the gap into agency work.

standard providers

LLM prompts

Preference prompts

0-100

Generative Brand Index

Method standard

Four diagnostic questions

The methodology separates Presence, Perception, Preference, and AI Readiness so agencies can diagnose the real failure layer.

Evidence model

Prompt + source + context

Each audit keeps provider, prompt category, answer evidence, source quality, market, language, persona, competitor, mission, and retest context attached.

Agency output

Baseline → fix → retest

The method is designed to produce client-ready reports, scoped remediation work, and repeated proof rather than a one-off visibility score.

VectorGap is built to find where AI memory diverges from market proof

A negative gap can come from four places: AI may omit the brand, misunderstand it, prefer a competitor, or fail to retrieve proof that supports the right answer. VectorGap separates those causes so the agency fixes the failure instead of rewriting pages blindly.

Perception

Tests how AI systems describe, classify, trust, price, and recommend the brand across buyer-like prompts.

Preference

Tests whether AI chooses the brand or a named competitor when the buyer asks for a recommendation.

Generative Brand Index

Combines target-aware Presence, Perception, Preference, AI Readiness, evidence coverage, and confidence into one executive 0-100 score.

Presence and AI Readiness

Checks whether the brand appears for buyer-problem prompts and whether pages, schema, facts, source blocks, freshness, and entity clarity are ready for Google AI and LLM retrieval.

The executive index is not a detached vanity score

The Generative Brand Index exists to summarize the current state without flattening the audit. Exact target evidence comes from Presence, Perception, and Preference. AI Readiness, Source Registry, Web Evidence, and Share of Model add retrieval context. Confidence explains how complete and reliable the current evidence set is.

Target-aware core

The index respects the market, language, persona, provider panel, prompt category, and competitor set instead of mixing unrelated global checks.

Context layers

Source quality, extractability, entity readiness, and public proof alignment explain why the score is strong or weak.

Agency workflow

The score points to the report narrative, Mission Control priorities, same-target retests, and the next remediation sprint the agency can sell.

Seven standard providers, interpreted as separate evidence streams

VectorGap compares provider behaviour instead of pretending one answer represents the market. ChatGPT, Claude, Gemini, Perplexity, Grok, Mistral, and DeepSeek each have different answer, retrieval, and citation patterns.

Provider variance

Differences between providers reveal where the brand has fragile proof, unclear entities, or source gaps.

Market and language context

Audits preserve country, language, industry, persona, and competitor context so results can support targeted agency work.

Prompt-level evidence

Scores stay tied to prompts, answers, source context, and category labels so teams can inspect the reason behind the number.

Treat the client website as the answer AI should trust first

For AI Readiness work, the website must behave like the authoritative record: clear facts, comparison-safe claims, extractable proof, cited sources, and dated updates that make it easier for AI systems to repeat the right story.

Authoritative facts

Brand Knowledge, company facts, pricing, markets, products, and proof claims should agree across the public site, reports, exports, and AI-readable files.

Claim-level proof

Every commercial claim needs a compact proof block: what the brand does, who it serves, what outcome it creates, and which public evidence supports it.

Contradiction control

VectorGap surfaces stale facts, hallucinated features, weak sources, and competitor confusion so agencies can replace ambiguity with a stronger public record.

Presence and AI Readiness explain why answers drift

When AI misunderstands a brand, the cause is often outside the prompt: thin third-party proof, unclear company facts, weak schema, poor source blocks, missing entity references, or pages that Google and answer engines cannot ingest cleanly.

Presence

Tests whether AI systems surface the brand for specific Bottom of Funnel buyer problems when the prompt does not name the brand.

AI Readiness

Checks whether pages, schema, FAQ blocks, source blocks, canonical facts, and llms.txt make the brand easy to extract, cite, and retest.

Google AI/entity readiness

Looks for clean organization facts, sameAs links, public profiles, knowledge-graph signals, sitemap health, and crawlable pages that help ingestion systems reconcile the brand.

Perception and Preference answer different questions

Perception checks how AI describes the client. Preference checks whether AI would choose the client against competitors. Agencies need both to understand visibility and commercial recommendation risk.

Perception

Measures coverage, sentiment, accuracy, hallucination risk, source quality, and how answers change by provider and target context.

Preference

Uses direct competitive matchups across categories such as price/value, capabilities, innovation, ease of use, trust, market fit, support, and overall recommendation.

Score construction

Preference scoring weighs category strength, direct wins, rank strength, evidence support, and provider consistency so a score reflects decision quality, not raw mentions.

Progress only matters when the same target is repeated

Retests compare the same brand, market, language, persona, provider set, and competitor context after remediation work. That makes reports defensible and helps agencies show whether shipped work changed AI answers.

Before state

Capture the weak answer, provider, source gap, competitor implication, and mission before work starts.

Remediation evidence

Link actions such as proof blocks, schema, entity facts, and source updates to the specific audit gaps they target.

After state

Rerun the same audit target and compare recommendation rate, rank, evidence strength, hallucination reduction, and provider consistency.

Method → report → purchase path

A methodology page should answer how the work is done, then route the buyer into the proof they can inspect and the operating capacity they can buy. The method becomes commercially useful when it shows the sample report, reduces checkout risk, and connects to Agency OS.

Inspect the report format

Show how the five-layer method becomes executive summary, prompt evidence, Mission Control actions, same-target retests, and a client-ready report.

Open page

Reduce purchase risk

Pricing proof-before-payment lets the buyer check the output and evidence chain before committing to Agency OS.

Open page

Buy the operating capacity

Agency OS turns the method into portfolio audits, remediation missions, retests, exports, and client-ready reporting across clients.

Open page

Which evidence does AI need before it can reuse the page?

This table turns the page into a structured extraction target: the buyer question, the evidence an AI system can read, and the action an agency can sell or execute next.

Buyer question	What AI can extract	Agency action
Which diagnostic question should an agency answer first?	The methodology defines when to test Presence, Perception, Preference, or AI Readiness instead of collapsing every problem into one score.	Select the diagnostic layer that matches the buyer risk: wrong story, competitor choice, weak source layer, extractability blocker, or entity/readiness gap.
What evidence makes the result explainable?	Provider, prompt, answer, source quality, market, language, persona, competitor, mission, and retest context stay connected.	Use that attached context to brief the client, scope the fix, and avoid generic recommendations that cannot be retested.
How does the method prove progress?	The same target context can be repeated after remediation, preserving provider and competitor scope for before/after comparison.	Ship public proof, schema, entity, source, and page updates, then rerun the same audit target for a defensible client-ready report.

Questions agencies ask before turning AI visibility into client work

Does VectorGap claim perfect ground truth across AI systems?

No. AI answers are probabilistic and provider behaviour changes. VectorGap makes audits inspectable, repeatable, and evidence-led so agencies can reason about movement instead of relying on a single opaque score.

Why include Preference?

Mention visibility is not the same as being chosen. Preference shows whether a provider recommends the client against named alternatives and which proof gaps affect that recommendation.