> ## Documentation Index > Fetch the complete documentation index at: https://docs.tryxlr8.ai/llms.txt > Use this file to discover all available pages before exploring further. # Experiments > Track your brand's AI presence and sentiment by running experiments across multiple LLMs. Learn how to create, manage, and act on experiments in XLR8.ai. An **experiment** is the core unit of measurement in XLR8.ai. Each experiment defines what you're tracking — your brand, the type of analysis, the queries to test, and the AI models to run against. You run experiments repeatedly over time to see how your AI visibility and brand sentiment change. Measure how often your brand appears in AI-generated responses across leading models. Understand how AI models describe and perceive your brand in their responses. Interpret mention rates, sentiment scores, citations, and AI-generated analysis. Compare executions over time with trends, baselines, and source concentration. Automate recurring experiment runs so your data stays current without manual effort. ## Visibility vs. sentiment — which should you use? Choose a **Visibility Analysis** experiment when you want to measure whether your brand is being mentioned in AI responses at all. Visibility answers questions like: * "Does Perplexity mention us when users search for our category?" * "How often do we appear vs. our competitors?" * "Is our website being cited as a source?" Visibility results show mention rates per model, citation presence, and competitor comparison — all broken down by query category. **Best for:** Brands establishing an AI presence baseline, competitive benchmarking, and tracking the impact of content changes over time. Choose a **Sentiment Analysis** experiment when you want to understand the tone and framing of how AI models talk about your brand. Sentiment answers questions like: * "Does ChatGPT describe us positively or negatively?" * "How do AI models position us relative to a competitor?" * "What pros and cons do AI models associate with our brand?" Sentiment results include sentiment scores, listed pros and cons per query, and competitor positioning data. **Best for:** Brand reputation monitoring, competitive positioning analysis, and identifying specific messaging improvements to address. ## Creating a new experiment Creating a new experiment has four steps: **Company**, **Categories**, **Queries**, and **Create**. From the sidebar under **Strategy**, click **Experiments**. Then click **New Experiment**. Choose your experiment type and subtype, then enter your company details. **Experiment types:** * **Visibility Analysis** * **Sentiment Analysis** **Visibility subtypes:** Discovery, Informational, Competitor Alternatives **Sentiment subtypes:** Competitor Comparison, Brand Sentiment Fill in: * **Company Name** (required) * **Company Website** (required) * **Company Aliases** (optional) — alternative names used when detecting brand mentions * **Sales transcript** (optional) — `.txt` file to improve query suggestions * **Categories & queries CSV** (optional) — import your own category/query map Click **Start**. XLR8.ai analyzes your company and prepares category suggestions. Review the generated **Description** and **Categories**. Edit them as needed, or expand **Add additional information** to steer the suggestions. If you imported a CSV, categories come from that file and you can edit queries on the next step. Click **Continue**. On **Review and Edit Search Queries**, edit, add, or remove queries in each category. Click **Next** when the query map looks right. Confirm the **Experiment Summary**, then configure: * **Model Selection** — choose which AI models to test * **Location Information** (optional) — country, region/state, and city for geo-targeted answers * **Enforce Web Search** (optional) — require GPT models to use web search before responding Click **Create Experiment**. XLR8.ai saves the configuration, opens the experiment detail page, and starts the first execution automatically. ## Managing your experiments ### Filtering and sorting On the **Experiments** page, search experiments and filter by type, subtype, or archival status. Sort by creation date or name. ### Pinning Pin important experiments to keep them at the top of your list. Click the pin icon on any experiment card to toggle the pinned state. ### Tags Add tags to categorize and filter your work. Tags are editable from the experiment detail view. ### Archiving Archive experiments you no longer actively track. Archived experiments are hidden from the default view but remain accessible via the archived filter. Archiving does not delete any data. ### Rerunning and scheduling From an experiment detail page you can: * **Rerun Experiment** — start a new execution with the current configuration * **Schedule** — set a recurring cadence (see [Scheduling](/experiments/scheduling)) ### Experiment settings Open **Settings** on an experiment to manage: * Web Search preferences * Company aliases * Competitors * Tracked URLs * Queries and categories * Outreach excluded domains (visibility experiments) ### Deleting Delete an experiment permanently from the experiment detail view. This removes the experiment and all of its executions and results. Deleting an experiment is irreversible. All execution history and results are permanently removed. Archive the experiment instead if you may need the data later. ## What's next? Configure tracked URLs, company aliases, geo-targeting, and model selection for visibility. Understand sentiment subtypes, action item sync, and how to interpret sentiment scores. Track how visibility and sentiment change across executions. Set up automatic recurring executions to track changes over time.