Opinion polls estimate what a much larger group of people thinks by asking a sample of them. A Westminster voting-intention poll usually asks how people would vote if a general election were held now.
The published figures reflect the answers, who took part and the adjustments the pollster made. Two firms can ask about the same election in the same week and produce different results.
Who takes part?
Most British political polling is conducted online, often through panels of people who have agreed to answer surveys. The pollster selects respondents to get a suitable mix of ages, regions and other characteristics.
Some surveys recruit people through random sampling. An open vote on a newspaper website or social media works differently: anyone who sees it can join in, so its participants may bear little resemblance to the electorate. The British Polling Council explains the main sampling approaches.
Most national Westminster polls cover Great Britain. Northern Ireland has a different party system and is usually excluded, so a GB poll cannot be compared directly with a UK-wide election percentage.
How the headline is calculated
Pollsters use weighting to adjust for groups that are overrepresented or underrepresented in the sample. They also decide how to handle people who have not chosen a party and how to estimate which respondents will vote.
The question itself can affect the result. Someone who prefers one party may plan to back another candidate locally. YouGov’s published method, for example, asks respondents about their own constituency as well as a general election nationally.
The full tables show more of this process than the headline percentages, including the questions, sample sizes and breakdowns by age or previous vote.
Has public opinion changed?
The interview dates, usually called fieldwork, tell you when the answers were collected. Publication can come several days later.
A one-point rise in a single poll may be ordinary variation between samples. Repeated increases in successive polls, especially across several organisations, give stronger evidence of a change.
Our polling database links to the original releases. The Poll of Polls combines recent findings, while the margin-of-error guide explains why even a carefully conducted poll has uncertainty.
Reviewed 19 October 2026.