Playground

The Playground is a sandbox for running one-off LLM conversations that do not affect your visibility or recommendation metrics.

It exists so you can explore, test, and demo without polluting the measurements that drive the rest of the platform.

Why the Playground Exists

Every scenario Genezio runs as part of its daily schedule feeds into your AI Recommendations, AI Visibility, and the other KPIs the platform reports.

That's the right behavior for measurement — but it's the wrong behavior when you want to:

• test how a new persona will behave before committing it to a scenario
• run an ad-hoc investigation ("let me just ask ChatGPT this one thing as a 35-year-old marketer in Germany")
• try a wording variation before turning it into a real scenario
• run a live demo without changing the customer's numbers

The Playground separates exploration from measurement. Anything you run there stays out of your reported metrics.

What You Can Configure

When you start a Playground run, you choose:

• the answer engine (any supported answer engine)
• the persona to run as (including their language and location)
• the Agent type — Prompter, Recommender, Comparer, or Introspector — which determines how the conversation is structured
• the language for the run

Genezio then executes a full multi-step Agent conversation, the same way it would for a real scenario — just outside of metric collection.

Where to Find It

The Playground is accessible from the Settings screen.

You can also load a specific scenario into the Playground from its scenario detail view. This is useful when you want to re-run an existing scenario with a different persona, engine, or wording — without disturbing the metrics that the original scenario contributes to.

History and Auto-Cleaning

Every Playground run is saved to a history view so you can:

• compare attempts side by side
• return to a past run
• iterate on persona, wording, or engine choices

Older Playground messages are auto-cleaned to keep the history tidy. Treat the Playground as a working sandbox, not a long-term archive — if a run produces something you want to preserve, promote it into a real scenario, capture the findings in a brief, or send it to the Geo Assistant for analysis.

Typical Workflows

Test a new persona before committing. Draft a persona, run a few Playground conversations against it, see how answer engines respond, then promote it to a real persona once you're happy with the shape.

Investigate a hypothesis. "What does ChatGPT recommend in our category to a buyer with constraint X?" Fire one Playground run; read the answer; refine if needed.

Pre-flight a scenario change. Load an existing scenario into the Playground, tweak the wording, and verify the AI response looks the way you expect before saving the change to the real scenario.

Run a demo. During a sales or stakeholder demo, show live LLM responses for the customer's brand without altering their measured numbers.

Playground vs. Scheduled Scenarios

Related Pages

• Creating Scenarios
• Selecting Answer Engines
• Running Conversations
• Core Concepts -> Personas