What are the required fields before I can publish?
There are 11 required fields across three sections in the design tab. In Instructions: interviewer name, company name, research objective (max 700 characters), target audience (max 100 characters), primary language, language selection toggle (yes or no), analysis language, privacy policy URL (must be valid https:// format), and tone of voice. In Discussion Guide: an opening message and at least one interview question. In Participants: maximum number of participants.
How many questions can I include?
Each discussion guide supports up to 20 scripted questions. Each scripted question can have between 0 and 3 AI probing follow-ups.
To estimate interview length, allow roughly 45 seconds per question — counting both scripted questions and AI probes. Most clients aim for interviews of 20 to 25 minutes, which typically works out to around 12–15 scripted questions with approximately 10 AI probes in total.
As a general rule, we recommend setting 1 AI probe per scripted question. This gives you enough depth to explore each topic meaningfully while keeping the conversation from feeling repetitive.
What are question groups and how does randomisation work?
Questions can be organised into named groups. Each group has a randomise toggle. When enabled, the items inside that group (questions or nested subgroups) are presented in random order to each participant. This lets you control randomisation independently at different levels of your discussion guide. Use groups to prevent bias in concept testing or to structure a long guide into logical sections.
How do AI probing questions work?
After a participant answers a scripted question, the AI analyses the response and generates a follow-up probe if probing is enabled. The probe is guided by the participant's answer, your research objective, the tone of voice, and an optional Probing Guide field where you can specify exactly what you want the AI to explore. We recommend 1 probe per question to maintain flow.
What is a Probing Guide and when should I use one?
A Probing Guide is an optional text field available for each question when probing is set to 1, 2, or 3. Use it to focus the AI on specific aspects ("focus follow-ups on the dashboard and reporting features specifically"), avoid sensitive areas ("do not probe on interpersonal conflicts"), or direct the AI toward specific insights ("dig into which competitor they almost chose and why they didn't"). When not needed, leave it blank and let the AI probe naturally.
What response modes can participants use?
Text (type answers),
Speech or audio (record voice),
Video (record camera)
Image uploads.
At least one mode must be enabled. You can combine modes in the same interview. If multiple modes are enabled it is up to the participant to choose. You can also override available modes per link using the URL parameter inputMethods (e.g. ?inputMethods=text,audio).
What tones of voice are available?
Friendly (warm, conversational, uses contractions, good for consumer and general research)
Formal (polished, professional, no contractions, suited for B2B, executive, and academic audiences)
Neutral (balanced and objective, good for general-purpose and mixed audiences)
Empathetic (supportive and sensitive, validates participant experiences, essential for healthcare, wellbeing, and sensitive topics)
Challenging (direct and probing, pushes respondents to think deeper, suited for expert interviews and stress-testing ideas).
The default is Friendly.
What stimuli can I add to questions?
Images (JPEG, PNG, WebP, GIF), videos (MP4, max 200 MB per file), audio (MP3, WAV), and PDF documents. Up to 10 stimuli items per question. Mixed types are allowed in a single question. You can add custom labels up to 50 characters per item (automatic fallback labels A, B, C are assigned if you don't). Recommended image dimensions are 1200 x 800 px minimum in landscape or square format. Shorter video clips (under 2 minutes) work best in interview flows.
What is the maximum file size for stimuli?
200 MB per file.
Can I ask quantitative questions?
Yes. Participants speak their answers ("I'd rate this 7 out of 10" or "I strongly prefer option A") and Tellet automatically quantifies and extracts these into structured data visualised as bar charts and distributions alongside the qualitative insights.
Can I edit questions after publishing?
The discussion guide is locked after publishing to protect data integrity. Research objective, interviewer name, company name, primary language, allow language selection, and tone of voice are also locked. Fields that remain editable after publishing: response types, analysis language, privacy policy URL, project branding, maximum participants (increase only), and panel integration settings. Contact [email protected] if you need a minor change.
Can I re-use a discussion guide from a previous project?
You cannot import an old discussion guide, but you can duplicate an existing project (three dots on the project > Copy to > select workspace). The copy includes questions, settings, and media but not response data or analysis, and starts in Draft status.
Can I test my interview before going live?
Yes. Use the Test Chat button in the project editor (available before publishing) to preview the interview as you build it, or use the Sandbox link in the Share tab after publishing. Sandbox conversations are marked as test data, do not count toward your participant limit, and can be deleted separately. Always test on multiple devices and browsers before sharing the real link.
What guardrails keep the AI on track?
Scripted questions anchor the flow (the AI cannot skip or replace them). Probe limits prevent topic drift. Years of refined system-level prompts constrain the AI to your research topic and prevent inappropriate content. Safety filters are always active. Tone and content safeguards ensure a professional, research-appropriate experience.
What is the Context Training (Context Window) field?
An optional text field where you can provide the AI with background knowledge about your product, brand terminology, industry jargon, or unusual spellings (e.g. stylised product names). Each text snippet can be up to 1,400 characters. You can add unlimited snippets. Use it when the topic requires specific knowledge or when participants may reference terminology the AI would not otherwise understand.
Can I set mandatory questions?
All scripted questions in the discussion guide are effectively mandatory. Participants cannot skip them. If a question does not apply to a participant they can say so ("this doesn't apply to me") and the AI will adapt, but a response must be given before the interview progresses.
Can I do A/B concept testing?
Yes. Use the question randomisation feature to control the order in which concepts appear. Add stimuli to questions and use a Probing Guide that references the stimuli explicitly to ensure the AI probes on the right elements.