A measurement you can defend — not a vibe.
Most “AI visibility” tools ask an engine once and call the answer a fact. Growgence treats AI like the noisy, non-stationary instrument it is: sampled, gated for significance, and epoch-aware — with the receipts to prove every number.
From your domain to a defensible score.
- 01
Add your brand & ground truth
Enter your domain — we fetch your pricing, services, team and location into an editable profile. You confirm what’s true. That confirmed ground truth is what accuracy is measured against (and nothing is ever invented).
- 02
We probe the engines
A library of real category, comparison, recommendation and direct-knowledge prompts runs across the engines — each prompt sampled several times, because one answer is an anecdote, not a measurement.
- 03
We parse every answer
For each response we detect whether you’re named, where you rank, which URLs are cited, the sentiment, and which competitors appear — turned into structured observations.
- 04
We score five dimensions
Presence, citation, accuracy, share-of-model and sentiment become a rate with a 95% confidence interval, then a single weighted 0–100 AI Visibility Score.
- 05
We gate for real change
A movement only counts if it clears a statistical noise band — and trend lines break at model updates, so a new model version is never mistaken for a change in your brand.
- 06
You get the evidence
The score, the per-engine breakdown, the exact prompts and citations behind it, and a branded PDF. Every number traces back to a stored response.
Six principles behind every number
Sampled, not single-shot
Every (prompt × engine) is run K times → a rate with a Wilson confidence interval. Small-sample noise can’t masquerade as signal.
Model-epoch aware
We record the model version behind every observation and break trends at epoch boundaries. Cross-epoch “changes” are excluded from alerts.
Honest framing
We measure LLM-API visibility in grounded mode — a rigorous, repeatable proxy. We never claim it’s exactly “what every real ChatGPT user sees.”
Accuracy never fabricated
Accuracy is scored only against your confirmed ground truth. No ground truth for a fact? It’s left blank, not guessed.
Reproducible
Prompt, engine, model, version, temperature and the raw response are stored with every observation — so any claim can be defended or audited.
Your data, isolated
Every tenant’s data is isolated at the database layer. Your prompts, competitors and ground truth are yours alone.
Do it yourself, or have us do it.
Growgence is the tool that diagnoses the problem. When you want the problem fixed, our specialist agency implements the work — and you keep the tool to prove it worked.