The research behind the app
The AI that reads your pitch can only point to these studies; it never writes a citation itself. Each entry says whether the app measures the behaviour the study measured, measures a stand-in for it, or only uses the study to explain a design choice.
Delivery changes how investors judge a venture. It is not evidence that the venture is good: in Hu & Ma (2025), positive delivery raised the chance of funding but did not predict how funded startups performed. Pitch training in Clingingsmith & Shane (2018) helped investors tell strong ideas from weak ones. Use the app to make your idea easy to judge, not to hide its weaknesses.
Speakers trying to persuade talk louder and vary their volume more. Listeners read this as confidence, which made the speakers more persuasive without making them seem less sincere.
Limits: Lab experiments with recorded messages, not investor pitches.
An energetic, high-arousal voice made founders seem more passionate, whether the emotion was positive or negative.
Limits: We measure pitch range and loudness variation, not their arousal model. Crowdfunding audiences, not investors.
Earlier studies link a wider pitch range, more vocal intensity and fewer disfluencies to sounding charismatic. Business students who trained on these features sounded more charismatic afterwards; women gained more.
Limits: No investor outcomes. We never coach a higher or lower pitch level, only variation.
In 1,460 crowdfunding pitch videos, a stronger peak of displayed joy went with more funding, especially at the start and at the end. Joy that lasted the whole time did not help more.
Limits: Observational. Crowdfunding, not angel or VC pitches.
In 489 crowdfunding pitches, happiness had an inverted-U relation with funding: some helped, a lot did not. More changes in expression went with more funding.
Limits: Observational. Why the share of time smiling is shown but not scored.
In 1,139 one-minute accelerator application videos, more positive delivery in face, voice and words went with a higher chance of funding. Among funded startups, positive delivery did not predict better performance.
Limits: Delivery shapes how investors judge a venture, not how good the venture is.
Broad smiles raised perceived warmth but lowered perceived competence compared with slight smiles. The competence cost was larger in high-risk situations.
Limits: Photos, not video. Why more smiling is not scored as better.
When the same voices were played faster, listeners rated the speakers as more competent. Ratings of benevolence were highest at each speaker's normal rate and fell when the voice was sped up or slowed down.
Limits: Old study, few voices. Why speaking rate is compared with the founder's own past attempts, not a fixed ideal.
Speakers sounded more confident with falling intonation and a faster rate, and vocal confidence changed how persuasive their message was.
Limits: The study also found a lower pitch sounded more confident. We do not use that part: pitch level differs by sex and other studies disagree.
Listeners preferred a speaker who dodged a question fluently over one who answered it with stumbles, and often did not notice the dodge. Showing the question on screen made dodges easier to spot.
Limits: We count filler words as a stand-in for fluency. The app flags dodges; it never teaches them.
Recruiters who heard a candidate's pitch judged the candidate more intelligent and more hireable than recruiters who read the same words.
Limits: Job pitches, not investor pitches.
Venture capitalists' interest in funding followed how prepared the founder seemed, not how much passion the founder displayed.
Limits: Preparedness was rated by investors; we use coverage of the screening factors as a stand-in.
In 113 funded TV pitches, founders who seemed more prepared gained more legitimacy and more funding.
Limits: Only funded pitches were sampled, and TV footage is edited.
Angel investors on Dragons' Den first rejected any pitch that showed one of eight fatal flaws: product not close to market, no protection from competition, unclear value to customers, no known customers, no route to market, too small a market, a team without relevant experience, or no sound financial plan. Only pitches with none of these got a closer look.
Limits: We check whether the pitch addresses each factor, not whether the venture really meets it.
In a field experiment with 271 aspiring entrepreneurs, pitch training added pitch elements. Experienced investors then rated strong ideas higher and weak ideas lower: training made the pitch a clearer signal of the idea.
Limits: Mostly student entrepreneurs; investor interest, not funding.
Managers and MBA students trained to use charismatic leadership tactics (metaphors, stories, contrasts, rhetorical questions, three-part lists, moral conviction, shared sentiment, high goals, confidence the goals can be met) were rated more charismatic and more leader-like.
Limits: Leadership speeches, not investor pitches.
TED talks with more verbal charisma tactics drew more views and were rated more inspiring.
Limits: Observational; popularity, not funding.
The same interviewees were rated higher when looking at the camera than when looking down at the screen. Looking at the screen was rated even below audio only.
Limits: Small sample; job interviews. Our gaze measure is an estimate from the webcam.
Viewers remembered more of a video presentation when the presenter looked into the camera part of the time than when the presenter never did.
Limits: Recall, not persuasion.
In 2,184 TED talks and follow-up experiments, gestures that illustrate what is being said made speakers clearer, more competent and more persuasive. The number of gestures did not.
Limits: The webcam can see hands but cannot tell what kind of gesture they make, so hands are shown and not scored.
Hand gestures tied to speech made a speaker seem more persuasive and competent than no gestures or self-touching.
Limits: Single speaker.
At pitch competitions, founders who showed stereotypically feminine behaviour (warmth, sensitivity, expressiveness) were less likely to be picked as finalists, whether they were men or women.
Limits: Conflicts with Hu & Ma (2025), where warmth helped. The evidence on warmth displays is mixed.
Warmth and competence are the two main dimensions of how people judge others. The audience-rating items come from this study's trait lists.
Limits: Items were built for groups, not individuals; reliability for one founder is checked as ratings come in.
Investors and lay judges preferred ventures pitched by men, and the same pitch narrated by a male voice was chosen more often.
Limits: Why the app never compares a voice with a gender norm.
Women founders used more concrete language than men; abstract language made investors see more growth potential.
Limits: Why the app does not give generic 'be more concrete' advice.
At TechCrunch Disrupt, investors asked men more questions about gains and women more questions about avoiding losses, and founders answered in the same terms. Each extra loss-avoidance question cut funding. Answering loss-avoidance questions in terms of gains raised funding.
Limits: The study coded questions with a word dictionary and human coders; the app uses an AI model for the same distinction.
In 789 Shark Tank pitches, flattering the investors and putting oneself down went with less funding; agreeing with investors' opinions and self-promotion went with more.
Limits: TV setting; observational.
Angels and VCs found a founder's passion more appealing when the founder also seemed open to feedback.
Limits: We judge openness from the answer's wording.
Investors treat a founder's coachability as a signal in pitches, though how much it counts depends on the investor's own coaching experience.
Limits: Investor judgments only.
Deflecting a direct question can cost less trust than refusing to answer or being caught lying.
Limits: Questions about past behaviour, not investor Q&A. Why the app flags dodges rather than coaching them.
Across 607 effects, feedback improved performance on average, but more than a third of feedback interventions made performance worse, mainly when feedback drew attention to the person rather than the task.
Limits: Why feedback names behaviours, never traits.
Feedback about the task and how to do it works better than feedback about the person.
Limits: Education research.
Across 33 studies, video feedback improved interaction skills by 0.40 standard deviations, more when a structured observation form pointed to specific skills.
Limits: Professional interaction skills, not pitching.
Speakers given live feedback on volume and pace while talking preferred sparse cues, but outside observers saw no improvement in their talks.
Limits: Why nothing but a timer is on screen while recording.
Practising in front of an interactive virtual audience improved expert-rated public speaking.
Limits: Supports practising with the AI investor; not a test of it.
Adding up predictors with equal weights predicts nearly as well as statistically optimal weights, and better than expert judgment.
Limits: Why the pitch score weights its parts equally.
The pitch score is the plain average of the parts that could be measured. No study gives weights for combining these behaviours, so they count equally (Dawes, 1979). The score says how many research-backed behaviours you showed. It is not a chance of being funded.
Screening factors addressed
Share of Maxwell et al.'s eight factors the pitch addresses (addressed = 1, partly = 0.5).
Charisma tactics used
Share of the nine verbal tactics used at least once.
Vocal energyprovisional cutoffs
Average of pitch variation (semitones) and loudness variation (dB), each mapped between the cutoffs below.
Fluencyprovisional cutoffs
Filler words per minute, mapped between the cutoffs below (fewer is higher).
Confident endingsprovisional cutoffs
Share of statements whose pitch falls at the end.
Looking at the cameraprovisional cutoffs
Share of time the head and eyes point at the camera, compared with the 3-second calibration.
Smile at the opening and the closeprovisional cutoffs
50 points for a clear smile in the first 15% of the pitch, 50 for one in the last 15%.
Provisional cutoffs will be checked against the warmth and competence ratings people give through rating links, once enough pitches have them.
Investors have been shown to favour pitches by men and to ask women more questions about risk. The app does not ask your gender. It measures your voice against your own usual pitch, never against a male or female norm, and it gives every founder the same mix of questions.