K4 · Stage 4 — Prove: the delta at every model generation — the flagship of the product family
Release Audit· KEOre-measurement timed to model releases
The verified, canonical, readable, citable record AI models learn from — so each new model is trained to know and recall your business correctly.
Read the definitionHow do you know the work done on the AI's memory actually paid off? We repeat the measurement at every significant new model version. The visible change against the previous result is the verifiable evidence.
The Machine equityMachine equity is the business asset that arises from AI systems being able to recognize, understand, recall, recommend — and, on a user's behalf, transact with — a brand. Composed of three layers: rented reach (retrieval visibility), owned memory (trained model recall), and machine buyability (agent-side transactability). Measured honestly only as a portfolio of ranged, per-layer metrics — never as a single blended score.
Read the definition loop
Machine equity is the business asset that arises from AI systems being able to recognize, understand, recall, recommend — and, on a user's behalf, transact with — a brand. Composed of three layers: rented reach (retrieval visibility), owned memory (trained model recall), and machine buyability (agent-side transactability). Measured honestly only as a portfolio of ranged, per-layer metrics — never as a single blended score.
Read the definition- 1 · MEASURE — where you stand and which questions matter (Citation Tracker, Selector)
- 2 · FIX — why the AI skips you, fixed in code (Gap Analysis, Gap Closure)
- 3 · ANCHOR — an unambiguous entity, present where models learn (Entity Anchor, Corpus Campaign)
- 4 · PROVE ↺ — closed-book baseline and the delta at every new model generation (Model Memory Audit, Release Audit)
KG Doctor and the Claim Coherence Engine — the studio's own instruments — serve the entire loop.
AVE Studio doesn't run short-term campaigns: we build our partners' machine equity, across model generations.
The results of the Entity Anchor and the Corpus Campaign can only be judged credibly if we can show: did anything change? That can be meaningfully examined when a new model version ships — a new GPT, Claude or Gemini generation — because a model's own trained knowledge only refreshes with a new generation. The Release Audit is built for exactly that moment.
Watching for new model versions
The system automatically and free of charge signals when a significant new model appears, and offers a re-measurement. A paid measurement never starts automatically.
Measurement only after approval
After your approval, we run the web-off measurement on the new model — with exactly the same fixed question set as the K1 baseline.
Change, and comparison against competitors
We compare the result with the previous model generation. Optionally we measure competitors too: their movement helps show what would have happened without any intervention.
Every service runs on an application we built in-house for exactly this job — not manual spreadsheets, not general-purpose tools.
A release watcher and re-measurement system: automatic alerts on new model generations, same-question delta measurement with a competitor panel.
- A shareable Release report: shows the changes per model and per language.
- An itemized change list of false information: what the model newly learned, what it forgot, and what new errors appeared.
- Optional competitor comparison: helps separate the market's general drift from your own result.
- A long-term trend on the dashboard: generation by generation, you can follow how the AI's knowledge of you evolves.
Reports are delivered in Hungarian and English.
By our market analysis, a unique method
Re-measurement structured uniformly and tied to model releases is, by our market analysis, unique. The outcome of work on AI memory can only be judged from repeated measurement like this.
Comparison on identical footing
The baseline and the re-measurement use the same question set, so the results are genuinely comparable.
The system claims no more than the data shows
If the measurement ranges overlap, the result reads: "no demonstrable change." If competitor data is insufficient, we say: "not enough data for a reliable claim."
You pay only when there's something to measure
Paid measurement is tied to the release of a new model version. We charge no ongoing monthly fee in periods with no new, meaningfully measurable model generation.
Without a control measurement we do not claim that an improvement was certainly caused by our work. Examining your change together with competitors' change gives stronger evidence, but a competitor panel alone still does not prove causation. The system automatically signals only the new model version; paid measurement always requires separate approval. Instead of a combined score, we report results per measurement dimension, with confidence ranges. ---
- How much does the release audit cost?
- From €225 (net) per model release. The annual pass starts from €700 with at least three release audits and a trend report — cheaper from three major releases up, and it makes an unpredictable release calendar plannable.
- What do I receive?
- A before–after comparison on the new model generation: what the old model knew about you from memory versus the new one — with a competitor control group, so your own movement separates from the market's. Delivered within days of the release.
- Why re-measure at a model release?
- Because a new model generation's memory is built from the text corpus that existed before the cutoff: what was embedded by then, it knows — what was not stays out for months or years. The recall delta with a control group is the only honest proof that corpus work actually worked.
Curious what AI says about you?
Start with the free entry check — no commitment.
Request your free check