Trust and accuracy

A polling platform is only worth anything if you can believe the numbers on it. Anything measuring public opinion at scale will eventually be worth gaming, and the honest response is not to claim it cannot happen, but to be specific about what we do and do not do with the data.

What follows are commitments, not marketing. They describe how results are presented and what is never done with your answers.

What Peopoll commits to

  • Individual answers are never exposed

    Results are only ever shown in aggregate. No screen, export or feature reveals how a specific person answered a specific question. Demographics exist to make a result readable, not to identify anyone inside it.

  • The number of answers is shown, not hidden

    Every result carries how many people answered, and every filtered view carries the count for that slice. A percentage without a sample size is not a finding, it is a decoration, and the difference between fifty answers and fifty thousand is the difference between a hunch and a signal.

  • Results stay aggregated, however you filter them

    Filters stack, so you can narrow a result to a specific slice: an age bracket inside a single country, for example. What comes back is still a count, never a list of people. Narrowing changes which answers are counted; it never reveals who gave them, and no combination of filters resolves to an individual.

  • Nothing is weighted, and nothing is promoted

    Every answer counts once, and a result is a plain count of them. Numbers are never weighted, adjusted or estimated, and no one can pay to change a result or to place a question in front of you. The feed does order polls, surfacing ones people are actively answering, but that ordering never touches the numbers themselves.

  • Demographic sharing is yours to control

    You decide what context you contribute and who can see what. You can withdraw an answer at any time, which removes your contribution from the dataset entirely.

On manipulation

If a poll becomes influential, someone will try to skew it. That is true of every measurement of public opinion ever built, and pretending otherwise would be the least trustworthy thing this page could do.

What is in place today is deliberately unglamorous: rate limits on how fast anyone can act, an automated filter on the words a question or comment may contain, reporting tools, and a person who reads the reports. Every result also carries the number of answers behind it, which makes a small coordinated push much harder to pass off as a headline number. We would rather list that plainly than claim detection systems we have not built.

The structural protection is the participation rule: to see a result you must answer first, so inflating a number means bringing real people who are willing to answer honestly. That does not make manipulation impossible. It makes it expensive.

How to read a result

A percentage on its own is the least informative part of any poll. Three things around it decide whether it means anything.

How many people answered. A result drawn from a few hundred people is a hint. One drawn from tens of thousands is evidence. The number is on every result for exactly this reason, and it is worth looking at before the percentage rather than after it.

How old the question is. Opinion moves. A result gathered during a news cycle is a snapshot of that moment, not a permanent fact about a population. Each poll shows its age, and where a question has collected enough answers over time you can see the shape of that movement in its vote history rather than only the total it ended on.

Which slice you are reading. A global number and a single country's number are different findings, and neither is more true than the other. The interesting question is usually not what the average was, but which groups departed from it.

This matters most when you stack filters. Every filter you add narrows the group the percentage describes, so a global result and “Gen Z in France” on the same question are findings of very different weight. The answer count travels with whatever filter you have applied, for exactly that reason. Read it before you quote the number.

Where this sits alongside the legal pages

This page is about credibility: how results are produced and presented. It is not the legal document.

The Privacy Policy is the binding statement of what data is collected, how it is used and what rights you have over it. If the two ever appear to disagree, the Privacy Policy governs. See also deleting your account and our Child Safety Standards.

For how answering, results and comparisons actually work, see how Peopoll works.