Starting a new report
Updated
A report is the core unit of work in LLM Insights. It defines which brand you're measuring, which competitors you're benchmarking against, which AI platforms you're tracking, and which questions are being asked. Every metric you'll later analyze rolls up from how you set up this report so it's worth getting the setup right.
This article walks through the full setup flow, from choosing a use case to launching your first report and managing it after it goes live.
Starting a new report
From the Record Manager, click + Create Report in the top right corner. This opens the report creation flow, beginning with use case selection.
Step 1: Choose a use case
LLM Insights supports two use cases. Pick the one that matches your reporting goal you can always create additional reports for the other use case later.
Brand Visibility Track how your brand is represented, mentioned, and perceived across major AI tools. Choose this if your priority is measuring your own brand's presence and sentiment without explicitly benchmarking against named competitors.
Competitor Visibility Compare your brand's AI visibility and sentiment against key competitors across major AI platforms. Choose this if your priority is share-of-voice benchmarking understanding where competitors are winning visibility you're missing.
Select your use case and click Submit to proceed to configuration. From here, the setup is a three-step guided form.
Step 2: Report Configuration
The configuration step defines the scope of the report. Fields marked with an asterisk are required.
Report Name: A descriptive title (e.g., Acme_Competitor Benchmarking_Q2). Pick something that will make sense to other team members and to your future self.
Brand: The primary brand this report will track.
Competitor Brand: Up to five competitor brands to benchmark against. (Shown for the Competitor Visibility use case.)
Additional Context: Optional. A free-text field that guides prompt generation toward specific themes. For example: "How is my brand Acme perceived in terms of price and product quality?" The more specific your context, the more focused the resulting prompts will be.
AI Tools: Select the AI platforms you want to monitor. Prompts will be generated and run across only the tools you select.
Report Refresh Frequency: How often the report should re-run (Daily, Weekly, etc.). Reports re-run automatically at this cadence until the End Date.
End Date: The date when the report stops auto-refreshing. After this date, the report becomes Expired and can only be viewed, not modified.
Note: A note on consumption: every refresh cycle consumes prompts against your contracted volume. A 50-prompt daily report consumes seven times what the same report would consume on a weekly cadence. Pick the cadence that matches how often you actually act on the data, not the most frequent option available.
Click Next to continue.
Step 3: Review your Prompt Pillars
Prompt Pillars are the themes or categories your report is organized around. They drive both the granularity of your reporting and how prompts are distributed across topics.
Sprinklr AI generates a starting set of pillars based on the brands you selected, the additional context you provided, and real-world Sprinklr Social Listening signals from the last 60 days. This grounding in actual customer conversations is what makes the pillars meaningful rather than generic.
By default, you'll see five pillars suggested, and each pillar can be up to 80 characters long.
Review each pillar and ask: Does this represent a theme my customers actually care about, and one where AI visibility would matter for my business? From there you can keep, edit, remove, or manually add pillars as needed. You need at least one pillar selected to continue.
Click Next when you're satisfied with the pillar list.
Step 4: Review your prompts
Once your pillars are finalized, Sprinklr AI generates the actual prompts within each pillar. These are the real questions that will be sent to AI tools to test whether your brand is mentioned, how it's described, and what sources AI tools cite when answering.
By default, LLM Insights generates 50 prompts, distributed evenly across your pillars. If you've configured more than five pillars, distribution may vary.
Prompts are derived from real social listening signals, recent news coverage, and how AI tools tend to interpret and respond to your selected inputs.
Review each prompt critically. Keep the prompts that reflect how real customers ask about your category. Edit any that are close but not quite right. Add custom prompts to cover specific themes, product lines, or scenarios the AI didn't generate. Remove prompts that don't fit your business. You need at least one prompt selected to continue.
When you're happy with the list, click Generate Report.
After clicking Generate
Your report is now queued for generation. First-run reports typically take one to two hours to complete, depending on the number of prompts and AI tools selected.
You'll receive an in-app notification when the report is ready. Once generated, the View Report button appears next to the report name in the Record Manager.
You don't need to wait on the page the report runs in the background, and you can continue working elsewhere in Sprinklr.