The quality of your question shapes the quality of the answer. This guide covers how to write questions that get useful results, whether you are starting a free-form chat or building a preset prompt.
Tips for Better Questions
Be specific about what you want
The more precise your question, the more useful the answer. Instead of "How is the hotel doing?", try "What are the top five complaints about the spa in the last 90 days?" A focused question gives the AI a clear task and makes the result easier to act on.
Name the category or topic
Ask TrustYou understands TrustYou's sentiment categories: Cleanliness, Service, Food & Beverage, Room, and others. Naming one focuses the analysis. For example: "What do guests say about Room cleanliness compared to last quarter?" works better than "What's the issue with the rooms?"
Say what format and level of detail you want
If you have a preference for how the result is presented, include it in the question. "Give me a ranked list of the top five complaints" or "Summarize this in three bullet points" or "Group the results by entity" all steer the output toward something more immediately usable. You can also set the depth: "Give me a brief summary" returns a high-level view, while "Give me a detailed breakdown with examples per entity" goes deeper. You can also ask for a table or a chart. Without either, the AI picks a format and length on its own.
Use positive phrasing for guest segments
Describing a group you want to exclude tends to confuse the AI. Instead of "guests who are NOT loyalty members," ask about the group you do want: "What do loyalty members say about check-in?" or "What do guests without loyalty status say about check-in?" The same applies to other segments. Always name the group positively.
Use exact source names
When asking about a specific review platform, use the platform name exactly as it appears in CXP. For example: "What do Booking.com guests say about breakfast?" The AI matches names precisely and may not resolve near-matches automatically.
Ask about a specific topic to find relevant insights
You can use Ask TrustYou like a smart topic search. Instead of hunting through individual reviews, ask: "What do guests say about pool maintenance?" or "Are there any mentions of noise complaints near the bar?" The AI reads across your entire dataset and surfaces the patterns, so you get insights rather than a list of reviews.
Follow up to go deeper
Your next question can build on the previous answer. If the AI tells you the top complaint is slow room service, follow up with "Which entities have the most room service complaints?" You don't need to repeat context. The AI carries it forward within the same chat.
Start a new chat for a new topic
Each chat is tuned to a specific data scope and question thread. If you want to switch to a completely different topic or a different set of entities, start a new chat rather than continuing in the same one. This keeps answers accurate and avoids context drift.
Writing Preset Prompts
A preset prompt is different from a one-off question. It needs to work every time someone runs it, with different values filled in. Keep these in mind when writing one.
Write for repeatability
Your prompt should produce a useful result regardless of who runs it or when. Avoid hardcoding values that change, like specific dates, hotel names, or topics. If something changes between runs, use a variable instead.
Use variables for anything that changes
Variables let the person running the preset fill in the specific values for that run. A prompt like "Summarize the top complaints about [Category] for the period [Start date] to [End date]" works for any category and any time range. A prompt like "Summarize the top complaints about Food & Beverage in October" only works once. See Adding variables to a preset for the available variable types.
Give the AI clear output guidance
Tell the AI what you expect back. "List the top five complaints, grouped by entity" is easier to act on than "tell me about complaints." If you want a comparison, say so: "Compare overall scores across surveys and highlight which has the highest NPS." The clearer the expected output, the more consistent the results across runs.
You can also embed a structural example directly in the prompt: "Structure the output like this: entity name | top complaint | number of mentions." The AI will follow that pattern on every run, which makes results easier to copy into a report or share with your team.
Check for conflicting instructions
Read the prompt as if you were the AI and ask whether any two rules contradict each other. A common pattern: "use the exact verbatim text" combined with "translate everything to English" will produce inconsistent output for non-English reviews. Use conditional logic instead: "For reviews in English, copy the exact text. For other languages, translate and note [translated] at the end of the line."
Handle edge cases in the prompt
Tell the AI what to do when expected data isn't there. If a section has no reviews, should it write "No items this week" or skip the section? If the date range returns nothing, should it say so clearly rather than produce a partial output? Without instructions, the AI picks on its own and results vary between runs. Explicit rules keep output consistent.
One analysis per preset
Keep each preset focused on a single question. A preset that asks for complaints, NPS trend, and an entity comparison in one prompt is hard to interpret and harder to reuse. Split complex analyses into separate presets so each one has a clear purpose.
Test before sharing
Run the preset yourself with representative values before sharing it with your organization. Check that the result matches what the description promises. If you used date variables, confirm the dates fall within your plan's data scope, as dates outside that range will prevent the preset from running.
What Ask TrustYou Can Answer
The AI works with the review data in your selected scope. It can answer questions about:
- NPS scores and how they change over time
- Sentiment scores by category (Cleanliness, Service, Room, Food & Beverage, and others)
- Guest type comparisons (business vs. leisure, couples vs. families)
- Loyalty status comparisons (loyalty members vs. guests without loyalty status)
- Cross-survey analysis (comparing results across multiple sources)
- Top praise and top complaints for any entity or group
- Period-over-period comparisons (this month vs. last month, this quarter vs. last quarter), as long as both periods fall within your data scope
- Reviews in any language. The AI analyses multilingual data natively, so you can ask in any language and it will cover your full dataset regardless of what language each review was written in.
Note: NPS, guest type, loyalty status, and cross-survey analysis are supported for public surveys only. Private surveys and custom survey questions are not available for analysis.
What Ask TrustYou Cannot Answer
Some questions are outside what the AI can reliably address right now:
- Guest nationality: Country-level guest segmentation is not yet supported.
- Guest gender or age: These dimensions are not available for analysis.
- Competitor data: Ask TrustYou works only with your own review data collected through TrustYou. Competitor benchmarks are not available.
- Private surveys: Surveys configured as private in TrustYou stay internal to your organization and are not part of the data Ask TrustYou can query.
- Custom survey questions: For public surveys, only standard fields such as NPS, guest type, and loyalty status are supported. Custom questions specific to your organization are not available for analysis.
- Reviews before January 1, 2021: Older reviews aren't available. Depending on your plan's review limit, your available history may start later. See Dataset for details.
- Negation-based segments: Phrasing like "guests who are not in segment X" tends to produce inaccurate results. Rephrase to name the group you want to include.
If the AI cannot answer with the data available, it will say so. You can then try rephrasing or narrowing the question.