Live polls for presentations
Poll Maker Presentations puts your question on the big screen. Your audience answers on their phones with a QR code, no app.
0 / 300
- 0 / 150Right answer
- 0 / 150Right answer
- 0 / 150Right answer
The big screen shows the top answers. All 0 stay in the vote.
Drag to reorder, or press Space to pick up, the arrow keys to move and Space to drop.
Free, no sign-up for anyone.
Used by 650 million+ people since 2003
More than 6 million votes counted in a single day.
Live polling the whole room can see
People join on their phones with a QR code or a 6-digit code, and the chart moves the moment each answer arrives.
The big screen
A phone in the room
- Bars grow, word clouds fill and answers appear as the room votes
- No app, no account and no cap on how many people join
- The join code and a live count of answers stay on screen
Free, with no audience cap and no sign-up.
Read the room without stopping
Phones send emoji that float up the big screen. Slow down and I'm lost reach only you, on your phone remote.
A phone in the room
The big screen
Your phone remote
- Up to six emoji reactions: floating, as counts, or seen only by you
- Slow down and I'm lost never show on the big screen
- A slide change clears every signal, so each slide starts fresh
Audience questions, approved from your phone
People ask and vote questions up. You approve each one on your phone remote before it reaches the screen.
Your phone remote
The big screen
- Run live Q&A at a talk, a town hall or a lecture
- Turn on approval and nothing reaches the screen until you say so
- People vote questions up, so the best one rises to the top
Take the results with you
Every session gets a report: every answer, every question from the room and how people rated it.
- Download CSV, Excel, PDF, PowerPoint or PNG, all free
- Phones rate the session at the end: see the average and the comments
- Share a read-only link to the report with your team
This was very easy to use and worked perfectly for ethics training activities.
Start from a ready-made presentation
32 ready-made polls for talks, classes, workshops and meetings. Open one and change anything.
- Use this template
Panel discussion Q&A
- How much experience do you have with today's theme?Rating scale
- Where should the panel start?Multiple choice
- Ask the panel a questionAudience Q&A
5 questions, about 30 min
Talks and panels
Let the audience steer the questions, and the best ones rise to the top.
- Use this template
Workshop opener
- In one word, what brings you to this workshop?Word cloud
- How much do you already know about this topic?Rating scale
- What do you most want to leave with today?Multiple choice
5 questions, about 8 min
Training and workshops
Learn what people know and want before you teach the first slide.
- Use this template
Lesson exit ticket
- Without your notes: which key ideas do you remember from today?Word cloud
- How well do you understand today's topic?Rating scale
- Which part should we review next time?Multiple choice
4 questions, about 5 min
Classrooms
See what stuck in the last five minutes, and what to teach again.
- Use this template
All-hands pulse and Q&A
- In one word, how do you feel about the quarter ahead?Word cloud
- How clear is our plan for the next quarter?Rating scale
- Which topics should we go deeper on next time?Multiple choice
5 questions, about 15 min
All-hands and team meetings
Take the room's pulse, then answer the questions people really have.
How to run a live poll in a presentation
Four steps, and nobody has to sign in.
Write your questions
Type them in the editor at the top of this page, pick from ten question types, or start from one of 32 ready-made polls.
Press Create poll
Put your screen up. The room joins with the QR code or the 6-digit code, in their phone's browser.
Ask, then reveal
Answers arrive as people send them. Press H to show or hide results, C to close voting and T for a 30 second countdown.
Keep the report
When you end, the report holds every answer. Download it or share its link.
Live polling questions, answered
Is live polling free?
Yes. Building, presenting and sharing a live poll is free, with no audience cap, and nobody needs an account. Pro adds your own brand: it hides the Poll Maker badge everywhere and shares your poll library with your team.
How many people can join a live poll?
There is no audience cap on any plan, free included. If your polls regularly run far beyond 50,000 people in a room, we will get in touch about a plan that fits, and nobody is ever cut off mid-poll.
Do people need an app or an account to join?
No. They scan the QR code, or type the 6-digit code at poll-maker.com/join, in their phone's browser. You do not need an account to present either.
How do I run a live poll during a presentation?
Type your questions in the editor at the top of this page and press Create poll. Put the big screen up: people scan the QR code or type the 6-digit code, and each answer appears the moment it is sent.
Can I use it with PowerPoint or Google Slides?
Yes. Import a PowerPoint file (.pptx) and its titles, text, pictures and notes become slides between your polls; from Google Slides, download the deck as a .pptx first. Or download the join slide, a picture with your code and QR code, and drop it into your own deck.
Can I run a live poll on Zoom or Teams?
Yes. Open the big screen in its own window and share that window in Zoom, Teams or any video call. The code and QR code stay on screen, and people join on their phones or in a second browser tab.
Are live poll answers anonymous?
Yes, by default. You only see names if you ask for them in Settings, and you choose whether a name is optional or required.
What kinds of questions can I ask?
Ten kinds: multiple choice, image choice, type the answer, word cloud, open-ended, rating scale, rating grid, ranking, brainstorm and audience Q&A. Add slides between questions, mark a right answer for a quiz, or ask the room to guess the winner.
Can I hide results until everyone has voted?
Yes. In Settings, choose to show results when you reveal them, and change it for any single question. While you present, H shows or hides results, C closes voting and T starts a 30 second countdown.
What if someone sends something rude?
Common swear words are masked and held back from the screen until you choose to show them. You can also approve written answers and questions first, and hide any of them from your phone.
Can I run a quiz with points?
Yes. Mark the right answer and the room scores points, with bonuses for speed and streaks, then show a top 5 or top 10 leaderboard on the big screen.
Can I control the presentation from my phone?
Yes. Press R on the big screen for the link to your private phone remote: next, back, results, voting, the countdown and audience questions. It needs no sign-in.
Can I download the results?
Yes, free. Every session gets a report you can share, and you can download it as CSV, Excel, PDF, PowerPoint or PNG.
What is an audience response system?
An audience response system lets a presenter ask a question and see the whole room answer live, once with handheld clickers and now on people's own phones. Poll Maker Presentations is a free one: no clickers, no app, just a QR code.
Your next talk, with the whole room in it
Type your first question at the top of this page.
Free, with no audience cap and no sign-up.
See what the room really thinks
Ten ways to read your results that go further than a bar chart. Pick one to see it with example answers, then press Show me how and the editor sets it up for you.
Before the answer shows, the room guesses what share picked it. Then everyone sees how close they were.
ExampleOn the big screenWhen you end, each phone gets a short recap of its own: how its answers compare with the room. Only that phone sees it.
ExampleOn each voter's phoneSee how different groups in the room answered, such as managers and team members, side by side.
ExampleOn the big screenAsk a question at the start and again at the end. The results show who changed their mind.
ExampleOn the big screenFind the groups in your room: every person is a dot, and people who answer alike gather in a few named groups.
ExampleOn the big screenShow what every group in the room agrees on. A good way to end.
ExampleOn the big screenAfter your talk, your report shows how the room felt, slide by slide, from the reactions people sent. Only you see it.
ExampleIn your reportOpen answers from the room are grouped into a few themes, each shown with one real answer.
ExampleOn the big screenWatch the vote happen again, answer by answer, and see whether showing the results changed it.
ExampleIn your reportAsk a standard question word for word, and your report shows where your room sits among other rooms that asked it. It appears once 30 rooms have asked the same question.
ExampleIn your report
For stats nerds: the methods and tests
Guess the room
The actual value is a count of this room's own votes, so it is treated as known. Crowd error is |mean guess - actual|; the average miss is the mean of |guess - actual|; by the triangle inequality the crowd's error is never larger than the average miss, and the diversity prediction theorem (Page 2007) holds exactly: crowd squared error = average squared error - variance of the guesses. Is the room's bias real? With 20 or more guesses, a one-sample t test of the guesses against the actual value (two-sided, df = n - 1, 95% t interval for the mean guess); with 5 to 19, an exact two-sided sign test of the median guess (guesses equal to the answer dropped). "Likely real" means p < 0.05 with at least 10 guesses; otherwise "Could be chance". Pickers versus non-pickers: Welch's t test, shown only when both sides have 5 or more guessers and labelled only when both have 10 or more. Guesses are bounded, skewed and heaped on round numbers, so the t test is approximate; people can talk before guessing; one test per question with no correction, so in a deck with many guess questions about 1 in 20 true nulls will read "Likely real".
Personal recap
Only questions with results visible to voters and 10 or more answers count, and a recap needs 3 of them. "Most popular" is the answer with the most votes (it need not be a majority); a question where the top answer is tied is left out. Your score is how many counted questions you answered with the most popular answer (for multiple choice, if any of your picks was it; for rankings, your first place). Shares are the same rounded numbers the results screen shows, never shown as 0% or 100% unless exact. Any answer you share with fewer than 5 other people shows no number. Your opinion group comes from the Opinion Groups view and is shown only when that view found groups. Everything describes this session's answers: no test, no p-value, and nothing here generalises beyond the room.
Split by an answer
Each group is compared with everyone else who answered (one against the rest, so two independent samples; with two groups it is one comparison). Single answer: Pearson chi-square test of the group-by-answer table without continuity correction when every expected count is at least 1 and at most 20% are under 5; otherwise Fisher's exact test for a 2 x 2 table, or a Monte Carlo permutation test of the chi-square statistic (2,000 seeded shuffles, p = (1 + hits) / 2,001). Multiple choice and words: a two-proportion z test per answer or word, Fisher's exact test when an expected count is under 5. Ratings: Welch's t test (Welch-Satterthwaite df) with a 95% interval and Hedges' g; rankings: the same on Borda points. All p-values are two-sided. Within one split every comparison is corrected together with Benjamini-Hochberg. "Likely real" means q < 0.05 and at least 10 people on each side; groups of 5 to 9 are always "Could be chance". "Worth a look" marks a split the tool found by searching: it passed a Benjamini-Hochberg correction over every comparison of every split tried (q < 0.05), has 10 or more people on each side and a gap of at least 10 points (or g of 0.3), and is a lead to check, not a finding. Intervals are 95% (Wilson, Newcombe for differences) and are not adjusted for the correction. The room is what is described: a test asks whether the gap is larger than a random reshuffle of these people would usually give, not whether it holds for people in general, and people in a room influence each other.
Mind Shift
Paired design: only people who answered both times count (n pairs); people who answered once are shown but left out of every figure. Two answers: McNemar's exact test on the people who switched (two-sided binomial, p = 1 when nobody switched). Three or more: the Stuart-Maxwell test of marginal homogeneity (chi-square, df = number of answers anyone moved into or out of, minus 1), with Bowker's test of symmetry alongside. Each answer's net shift in points has an Agresti-Min 95% interval. Ratings: the mean change with a paired t interval and Cohen's dz, and the Wilcoxon signed-rank test (zeros dropped, average ranks for ties, exact p) which sets the label. "Likely real" means p < 0.05 with 10 or more pairs; one test chosen in advance, so no correction. When a move of fewer than 5 people exists, exact figures are withheld and the label is shown only if it is the same at the tables with the fewest and the most changes that fit the chart. Limits: one device is one person; people who answered only once may differ from those who stayed; seeing the first result can anchor the second answer; the room is not a sample of any wider group.
Opinion groups
Each answer becomes numbers scaled so that no question can move two people more than a distance of 1 (missing answers take the question's average, and people who skipped questions are scaled back out). A principal component analysis gives the 2D map; k-means++ (fixed seed, 10 restarts) groups the map into 2 to 4 groups of at least 5 people, and the number of groups is chosen by the mean silhouette measured on all the answers, not on the map. Groups are shown only if they are clearer than chance: a permutation test reshuffles each question's answers between people 19 times (keeping every question's totals, breaking the links between questions), reruns the whole method, and groups appear only when the real silhouette beats all 19 (p = 0.05). Signature answers are those at least 15 points more common in the group than outside it, held by at least 40% of the group, with a one-sided Fisher exact p under 0.05; because the groups were built from these same answers, that p is a filter, not evidence. Groups describe this room; k-means can cut one continuous spectrum into pieces, two dimensions keep only part of the variation, and boundaries move in rooms of 15 to 40 people.
Common ground
Each statement is turned into agree or not per person: a Yes or agree-side answer, the top two points of a rating, or picking an answer; "Not sure" counts as not agreeing; people who did not answer are left out. A statement is common ground when at least 70% of every opinion group agrees (70% is a convention). Statements are ranked by group-aware consensus, the product over groups of (agreeing + 1) / (answering + 2), which is high only when every group is high; ties go to the bridging rank, the smallest Wilson 95% lower bound over the groups. Divisive statements: a group-by-agree table per statement, tested with Pearson chi-square (Fisher or a 2,000-shuffle permutation test when counts are small), Benjamini-Hochberg over every statement tested, shown as "Worth a look" only when q < 0.05, the smallest group has 10 or more people and the groups differ by 30 points or more. The groups were built from these same answers, so differences are expected by construction and the p-values act as a filter, not as evidence. Q&A upvotes count as agreement but no upvote is not disagreement, so Q&A is ranked separately and never called common ground.
Room Pulse
Time is cut into 10-second steps (longer for long talks). For each step the view counts phones, not taps: phones that sent an emoji, and phones that raised a signal (each phone once per signal per slide), so one enthusiastic phone cannot make a spike. A burst is a window inside one slide (one step for emoji, 40 seconds for signals) whose count X is unlikely under a steady rate: the expected count is the window length times the largest of the rate over the previous 2 minutes, the whole-talk rate and 0.1 per step, and p = P(Poisson >= X). A window is flagged when p is below 0.05 divided by the number of windows tested in that series (Bonferroni) and X is at least 5 phones and 10% of the room for emoji, 5% for signals. This is a screening rule for this talk, not inference about a population: people react to each other, the baseline is only locally steady, and the room count includes people who already left. The drawn curve is smoothed for display; detection uses the raw counts.
Themes
Every counted answer is turned into a vector: an AI text embedding (256 dimensions, unit length) or, without AI, TF-IDF word weights over word stems used in 2 or more answers. Spherical k-means (cosine similarity, k-means++ seeding from a seed fixed by the answers themselves, 3 starts) groups them into 3 to 6 themes; the number of themes is the smallest whose mean silhouette is within 0.03 of the best, and below a silhouette of 0.03 the view says there are no clear themes. Answers far from every centre and themes under 5 answers go to "Other answers". Shares are shares of answers, not of people. The quote is a real answer near the theme's centre, never edited except shortened. No test or p-value is computed: one person can give several answers and the room is self-selected, so the view describes this set of answers only. Embeddings group by topic more than by stance, short answers carry little signal, and theme names are AI summaries.
Vote Replay
People are split by the time of their first vote: before and after results first became visible to people who had not yet voted (or, when results were visible from the start, the first and second half of voters). The answer tracked is the one leading on screen at that moment (or the final leader). Its share among later voters is compared with its share among earlier voters with a two-sided permutation test that reshuffles who voted early and late 2,000 times and picks the leader again in every shuffle, so choosing the leader after looking cannot fake a drop; 95% Wilson intervals per group and a Newcombe interval for the difference describe the size. "Likely real" means p < 0.05 with at least 10 people on each side; one comparison per question. This is observational: late voters differ from early ones in many ways, people talk, and votes that were changed are kept at their final answer from their first vote time. A pattern, not proof that showing results caused it.
Rooms like yours
Each room counts once: its share of voters choosing an answer (or of right answers, or its average rating) on a template question asked word for word, from rooms in the last 365 days with 10 or more voters, at most 5 rooms per account, shown only with 30 or more rooms from 10 or more accounts. Rooms are stored in 40 bins; your room's percentile is the share of rooms below it, assuming rooms are spread evenly within a bin (error at most the share of rooms in your bin), with rooms exactly tied at the ends of the scale split half above and half below. Your room's likely range is a Wilson 95% interval for a share (a t interval for an average), mapped through the same curve. There is no test or p-value: the other rooms are a reference, not a population your room was drawn from, and with this much spread between rooms a test would mostly measure your room's size. Other rooms' values carry their own sampling noise, which widens the tails, so an extreme percentile for a large room is, if anything, understated. The rooms are those whose presenters chose this question on Poll Maker, not a sample of teams or classes.
How these were checked: every test above was recomputed on fixed cases, including zeros, ties and the smallest allowed groups, with an independent SciPy and NumPy version. 191 of 193 values agree; the 2 that differ are tail values below 0.000000000001, far under any threshold used.