1. Where the data comes from
We read the public search predictions that appear when you begin typing a question. Each request is pinned to US English, so the game does not change depending on where our server or you happen to be. A scheduled job captures a fresh set every day and freezes it; that daily archive is what powers the time periods:
- Last week — a consensus of the last seven daily captures (a ranked vote), so one odd morning cannot decide the answer.
- Last month and Last year — the capture closest to 30 and 365 days ago.
- All time — the oldest capture we have.
Where the daily archive does not reach far enough back yet, the game uses our reference corpus: predictions captured in bulk and checked by hand. The correct answer is always the top-ranked prediction for that sentence, and the ones shown after you answer are the real runners-up, in order.
2. What we never include
Search predictions can be cruel. Before anything reaches the game it passes several layers of filtering, and captures for the Stereotypes theme are read in full by a person before they ship. We exclude predictions about:
- race, ethnicity or skin colour, and religion;
- party politics, wars and disputed borders;
- sex and sexuality;
- health, disability, self-harm and eating disorders;
- hygiene and appearance put-downs about groups of people;
- slurs of any kind, and predictions naming private individuals.
We also remove name collisions — predictions that are really about something else with the same name, such as a TV character or a flower — and platform noise like “… reddit”. Some subjects are phrased the way people actually search (“why is nursing so” rather than “why are nurses so”) because that is the version with real predictions behind it.
3. How a question is built
- The question is the start of a real search, and the correct option is its top prediction.
- Wrong options are also real predictions, from the same sentence shape: on Easy they come from other subjects, on Average from a mix, on Hard from the same subject’s #2 to #4.
- An option that is genuinely true of the subject is never used as a “wrong” answer below Hard.
- Everyone playing the same date, theme and period gets the same five questions.
4. Tone
We treat predictions as a record of what people ask, never as a description of what anyone is. Game copy and articles laugh at the questions, the phrasing and the coincidences — not at nationalities, generations or any group of people. When a prediction is funny only because it demeans someone, we leave it out even if it passes every filter.
5. Articles
Every post on the blog is researched and edited by the GuessTheSearch team. Search predictions quoted in an article come from our own captures and are shown in the order they were suggested, with the capture period noted. Research findings are attributed to the people who published them; we do not invent statistics, surveys or quotes. Where the evidence is weak or disputed, we say so.
6. Corrections and removals
If you think a prediction should not be in the game, or an article gets something wrong, email hello@guessthesearch.com with the date, theme and sentence or the article link. We review every report. Offensive content is removed from future puzzles as soon as we confirm it, and factual corrections to articles are noted at the end of the post with the date.
7. Independence and advertising
GuessTheSearch is independent and is not affiliated with Google or any search engine. The site is paid for by advertising. Advertisers have no say over puzzles or articles, and ads are never placed inside a quiz round.