Stimulus

Stimulus is any media you show participants during a session — a prototype screenshot, a video ad, a product mockup. Attach it to a question and the AI moderator shows it automatically when that topic comes up.

​

Supported formats

• Images — PNG, JPG, GIF, or WebP
• Videos — MP4 or WebM, up to 720p, 10 minutes, and 250 MB. An MP4 needs H.264 video, and any sound track must be AAC. Then every participant’s browser can play an MP4
• PDFs — you pick the page, and it uploads as an image

The builder checks a video when you pick it. If it can’t use the file, the question shows the file’s name and the reason, with buttons to pick another file or dismiss it. A video it accepts opens in the stimulus editor and starts uploading right away. After you save, the question shows a frame of it, its file name, and its length. To change a video, open it and select Replace. To change an image, remove it and add another. If the upload fails before you save, the editor shows the error. Save keeps the file, and the question then offers Retry. A language version whose upload failed has its own Retry, in the editor and on the question. While the builder stays open, the study can’t launch until a saved upload finishes. If you leave or reload the page before you save, the builder asks first, and the new file is lost. If you reload after you save but before the upload finishes, the question says the video didn’t finish uploading. The study then can’t launch until you remove the video and add it again. If the video has no sound track, the stimulus editor says so. Nothing you change in the stimulus editor applies until you select Save. If you select Cancel after picking a file or making an edit, the builder asks Discard your changes? first. Select Keep editing to go back. The stimulus editor’s Loop sets how many times a video plays in a row: Off (once), 2 times, or 3 times. The question shows the count beside the length when it plays more than once. A video you add also needs an Introduction, the sentence the AI moderator says before the video plays. When you save a video, the builder fills it in with a short draft in the study’s language. Rewrite it in the editor or ask the builder chat to. A video in another language version needs its own file of the same kind, picked in the editor. Once a video finishes uploading, the builder writes its description and a Breakdown: the video’s moments, each with its start time and what happens. The AI moderator reads both during the session. The analysis reads both too, along with the video’s length and how many times it played. Edit them in the stimulus editor: select a moment to change its time or text, or select Add a moment. Generate all text writes both again. If the breakdown can’t be generated, the editor says so, and you can try again or add the moments yourself. A video you add needs its description and breakdown before the study can launch. When you add an image, the builder writes its description. In a concept test, it also writes the concept’s title, unless the concept already has one. If that text can’t be generated, the builder says so. To try again, select Generate for a question’s image, or Generate all text in the stimulus editor. You can also write it yourself. Images over 5 MB upload as JPGs, so the builder can still write their description. If the JPG is still too large, it’s scaled down, but never below 2560px on its long edge. GIFs upload unchanged. An image that’s still over 5 MB at that size uploads without an automatic description. A question’s image skips it, and a concept image shows that it couldn’t be generated. Write one yourself. An image over 10 MB at that point doesn’t upload.

​

Always add a description

When you attach stimulus, add a clear, human-readable description of what it is. This isn’t just for you — the AI moderator reads the description so it knows what participants are looking at and can ask sharper follow-up questions. It doesn’t read the description aloud. The analysis also reads it when it judges how participants reacted to the stimulus. A description like “Version A of the pricing page — three tiers, free/pro/enterprise, annual pricing shown by default” is much better than “pricing page.” The more context the AI moderator has, the better it can guide the discussion.

​

How it works in a session

When the AI moderator reaches a question with stimulus, it displays on every participant’s screen automatically. The moderator asks for first reactions, then uses your description to guide the discussion around it. A video plays differently. The moderator says the question’s introduction first. The video then plays for everyone at once, with sound, as many times as its Loop says, while everyone’s microphone is off and the moderator stays silent. When it ends, it leaves every screen, and the moderator asks the question.

​

Know the viewport

Stimulus displays in a fixed area in the center of the session screen, with participant video feeds on either side. The whole image scales to fit that area, so a full-length page screenshot shows at a size nobody can read. If you’re testing a full landing page, participants won’t be able to read it all at once. You have two options:

• Screenshot just the section you care about. If you want feedback on the hero, crop to the hero. If you want feedback on the pricing table, crop to that. This is usually the better approach — it focuses the conversation.
• Break the page into sections across multiple questions. Hero in question one, features in question two, pricing in question three. Each gets its own focused discussion.

Don’t upload a full-length page screenshot and expect participants to read it. Design your stimulus with the viewport in mind.

---

​

What you can test with stimulus

Stimulus works best when you’re showing people something concrete and asking for their reaction. One stimulus per question — if you need reactions to multiple things, use separate questions or a concept test.

​

Ad testing

Upload the ad creative (image or video) as stimulus. Create one question per ad variant. The AI moderator shows each one and asks participants what they notice, how it makes them feel, and whether it would make them take action. A video ad plays in full, with sound, as How it works in a session describes. Write the question’s introduction so it leads into the ad. If you’re testing multiple versions, use separate questions with clear descriptions like “Ad variant A — lifestyle imagery, emotional headline” and “Ad variant B — product screenshot, feature-focused headline.” The AI moderator can then ask participants to compare them.

​

Concept testing

Show early-stage ideas — sketches, wireframes, mockup screens. Concepts don’t need to look polished. Rough is actually better because participants give more honest feedback when something doesn’t look final. To compare several concepts in one question, make it a concept test. On a question with no image or video attached, open the question’s menu and choose Add concept test. Choose how many concepts each session sees in the test’s header. In your question goal, be specific about what you want to learn: “Understand whether users can tell what this product does from the landing page alone” gives the AI moderator a clear direction.

​

Landing page testing

Break the page into sections (hero, features, pricing, footer) across multiple questions. This lets you get focused feedback on each part instead of a vague “what do you think of this page?” For comparison testing, show the current page in one question and the new version in the next. Describe both clearly so the AI moderator can reference the differences.

​

Creative and brand testing

Logos, packaging, brand imagery, social media posts — anything visual. Upload the asset and describe what stage it’s at and what you’re deciding between. Works well for A/B testing: “We’re choosing between two packaging designs for our new product line. I’d like to understand which one communicates premium quality better.”

​

Product and UX testing

Walk participants through a flow by showing one screen per question. For example, a checkout flow might be four questions: cart → shipping → payment → confirmation. Each question gets its own screenshot with a goal like “Understand if users can find the discount code field.”

---

​

What makes good stimulus

Show work-in-progress, not polished finals. Mockups, wireframes, and rough prototypes get better feedback. When something looks finished, participants feel like their feedback won’t matter. One stimulus per question. Don’t attach three screenshots to one question. Participants can’t meaningfully react to all of them. Use separate questions for each thing you want a reaction to. Make it big enough to read. If there’s text in your image, make sure it’s legible on a video call. Participants won’t squint — they’ll just say “looks fine.” Use real examples over descriptions. Showing a competitor’s actual pricing page gets better reactions than saying “imagine a pricing page with three tiers.” People react to what they can see. Write a detailed description. The AI moderator uses your description to understand what’s on screen and guide the conversation. “Homepage redesign v2 — simplified nav, hero with product demo video, social proof section moved above the fold” gives it a lot more to work with than “new homepage.”

​

What to avoid

• Stock photos. Participants can tell. They’ll react to the photo feeling generic, not to the concept you’re testing.
• Overly complex screens. If you’re testing a flow, show one step at a time across multiple questions.
• Stimulus without a description. The AI moderator won’t know what it’s looking at and can’t guide the conversation well.
• Stimulus that needs context you haven’t given. If you show a dashboard, make sure your description and question intro explain what participants are looking at.
• Videos longer than 60 seconds. Participants lose focus. Keep video stimulus short and let the conversation be the main event.