Your AI agent can generate twenty plausible taglines. The next useful step is to learn what your intended audience thinks each one promises, which feels relevant, and what causes confusion.
Tagline testing uses User Intuition’s standard study workflow. The agent retrieves the study findings, each with prevalence and the participant quotes behind it.
What is the path from draft to participant reactions?
- Provide the candidate taglines, product context, and intended audience.
- Choose panel recruitment or your own participants.
- Ask the agent to create the study and send the brief through
customize_study. - Answer any planning questions, then review the full saved plan from
get_study. - Approve the launch before fieldwork: the agent shows the saved plan and a cost estimate, and you approve in one click inside Claude or ChatGPT before anything is spent.
- Retrieve the report and check the interview evidence behind the recommendation.
For a panel study, estimate_panel returns the cost and timeline for the audience and country you choose, and launch_panel fields the study once you approve it in chat. First interviews arrive within hours, and most panel studies fill within 24 hours. Synthetic respondents are also available for a directional first read on early drafts.
Worked example: three snack-brand taglines
Suppose a plant-based snack brand is comparing:
- “Snacks that taste like a cheat day.”
- “Real crunch. Nothing fake.”
- “Plants, but make it delicious.”
Here is how the agent would design that study. Ask participants what each line means before asking which they prefer. Explore whether the promise feels credible, what makes it distinctive, and what information is missing. Include a way for people to reject all three.
An overall preference can hide a segment difference. A parent packing lunchboxes may respond to “nothing fake,” while a gym-goer reaches for “cheat day.” Ask the agent to explain those differences and show the interviews supporting them.
Should participants see one line or all of them?
The exposure design shapes what the study can tell you, so settle it in the brief before the agent drafts the guide.
Monadic exposure shows each participant one tagline. Reactions are uncontaminated by the other options, which makes this the closest match to how a shopper meets a line on a shelf or in a feed. The trade-off is that each line needs its own group of participants, so a three-line test triples the sample needed to hear each line’s reactions repeat.
Comparative exposure shows every participant all the lines. Each person can explain why one beats another, which is where the sharpest language about distinctiveness tends to come from. To keep the first line from anchoring everything after it, ask the agent to rotate the order across interviews and to collect a first reaction to each line before any comparison begins.
A common pattern combines the two: open with one line alone, capture the unprompted read, then introduce the others for a head-to-head. Tell the agent which design you want, and the planning step builds the rotation and the screeners around it.
What does the moderator probe on each line?
For every tagline, the AI moderator works through four questions and ladders 5–7 levels deep on the answers.
- Comprehension. What does this line promise, in the participant’s own words? A line that half the room paraphrases differently has a clarity problem, however much people like it.
- Believability. Would this brand deliver that promise? “Nothing fake” invites a look at the ingredient list, and the moderator asks what the participant would need to see before trusting it.
- Fit with the brand. Does the line sound like the product they know or saw in the stimulus? A playful line on a premium pack can read as a mismatch.
- Distinctiveness. Could a competitor say the same thing? If participants attach the line to another snack brand, it is doing that brand’s work.
Because the moderator follows up on each answer, a reply like “it’s catchy” turns into what made it stick, where the participant imagines seeing it, and whether it would change what they put in the basket.
What does the report show?
For a tagline test, the report shows how many participants read each line the same way and the words they used. generate_report refreshes the analysis as interviews complete, and get_interview opens the transcript and recording behind any quote. Each finding arrives with its prevalence (n of N) and verbatim quotes, the same shape as example output from a User Intuition study of 43 Walmart shoppers (explore the Walmart study), where 20 of 41 preferred no staff contact.
For a preference tally, include the same preference question in every interview. The report then counts the explicit picks, states the denominator, and lists who chose none.
How should you read tagline results?
Read the four probes before the preference count. A line can win the head-to-head and still fail comprehension, which usually means people liked the sound of it while taking away different promises. Ask the agent for a short scorecard per line: how participants paraphrased it, what made it credible or not, whether it fit the brand, and which competitor it reminded people of. Each cell should carry the quote behind it.
Then look at who preferred what. If the lunchbox parents and the gym-goers split, the decision is about which audience the campaign serves, and the research has done its job by making that trade-off visible. Ask the agent to open two or three full interviews from each side so the team hears the reasoning in context.
How do you go from participant reactions to a final line?
A focused qualitative study finds confusing language, explains reactions, and sharpens the candidates. When the shortlist comes down to a close call, the agent can field a larger round with the refined lines, or hand them to a quantitative test that measures the gap.
When participants disagree, carry that disagreement into the brief for the next round. Minority objections can reveal an audience mismatch that a single headline metric would conceal.
A second round works best when it changes one thing at a time. Keep the strongest line as a control, then add variants that fix the specific problem round one found: a clearer verb where comprehension split, a proof point where believability was thin, a sharper word where people confused the line with a rival’s. Ask the agent to write the round-two brief from the round-one findings, quoting the objections each variant is meant to answer, and to keep the four probes identical so the two rounds compare directly. The agent can then answer the question that matters: did the fix move the reaction, or did a new objection replace the old one?
Keep the learning available
Each round stays in your account, so the next tagline brief can start from what earlier audiences said about the same promise, with the agent citing the interviews behind it.
See the research API and MCP server for AI agents for the supported workflow and concept testing for research design guidance.