Why this page exists
A word list is only as good as the judgment behind it. Anyone can paste a thousand nouns onto a page; very few of those lists survive contact with a real table of impatient kids, skeptical teens, or tired adults after dinner. This policy explains how our 6,964 distinct prompts are chosen, tested, graded, and corrected, so you can trust the library when it matters: mid-game, timer running, marker in hand.
This page is maintained by the same person who maintains the word lists themselves. If the process ever changes, this document changes first.
Sourcing rule: real game nights first
Every word in the library has one of two origins, and both require that the word was actually played before it was published.
- Hosted sessions. Most words enter the list from game nights hosted by our maintainer: family tables, office parties, classroom warm-ups, and holiday gatherings. When a word produces a fast, loud, correct guess, or a spectacular failure worth remembering, it gets noted down the same evening.
- Reader suggestions. Visitors submit words through the contact page. Suggestions are never published as-is; they join the candidate pool and must survive at least one play-test with a group that did not suggest them.
What we do not do: we do not copy lists from other websites, we do not harvest words from the official boxed game, and we do not publish machine-generated word dumps. Automated tools may help format or organize the data, but a word only enters the library after a human watched real people try to draw it.
The four filters every word passes
Before a candidate word is added, it is checked against four filters. Failing any one of them sends the word to the rejected pile, which is considerably larger than the published library.
- 1. The shape filter. Can a person with average sketching skill give this word a recognizable silhouette in under sixty seconds? Words that exist only as typography or logos are rejected, because drawing letters breaks the game's core rule.
- 2. The familiarity filter. Would the intended audience recognize this word instantly once drawn? A great word for marine biologists is a dead round at a birthday party. Every word is filed under the audience that actually knows it.
- 3. The fairness filter. Does guessing this word require niche trivia, or can a reasonable person get there from the drawing alone? Fair words reward observation; unfair words reward whoever happens to know the drawer's obscure hobby.
- 4. The table-safety filter. Is this word comfortable to play in front of children, grandparents, coworkers, or strangers? That bar applies to every library we publish, and the adults list is no exception: the 2026-09-27 pass re-scoped it so the humor comes from social awkwardness, work or everyday life rather than a substance, a wager or a diagnosis, and it stayed live at its 300 prompts instead of being fenced off behind an age gate.
How difficulty grades are assigned
Difficulty is measured, not guessed. Each word is scored on three criteria:
- Silhouette test. One-stroke recognizability. Cat passes instantly; Chandelier needs a plan.
- Idea count. How many separate visual ideas the word requires. Fondue is a pot plus a dipping fork: two ideas, medium grade.
- Knowledge load. How much world knowledge the guess needs. Folklore creatures carry heavy knowledge load and sit in the hard tier even though they are easy to sketch once identified.
The score sets a provisional grade, and mixed-age play-testing confirms it. If a "medium" word gets guessed in seconds by children, it moves down. If an "easy" word stalls three adult groups in a row, it moves up. Grades are about the audience, never about the word itself.
Review cycle and corrections
Themed lists are revisited before their season starts: the Halloween set in September, the Christmas set in November, and so on. Core difficulty lists are reviewed quarterly. Every dated pass is recorded in the change log below, with the counts it moved; the About page states the same measured totals and points back to this log.
Corrections follow a simple standard: when a player reports a word as misfiled, confusing, or uncomfortable, the report is replayed against the four filters. If the report holds, the word is re-graded or removed at the next update, no debate. Reports reach us through the contact page or [email protected] â the same two channels the privacy policy and terms pages publish. The library gets better by subtraction as often as by addition.
Change log: audits we actually ran
This section exists because a policy page that describes a process nobody can check is decoration. Below are the audits we actually ran and the counts each pass moved. The numbers come from a script that reads the library files and our dated backups of those files, not from memory; the current snapshot was measured on October 4, 2026.
How a word gets removed
Four checks run over every library file. A prompt has to fail one of them to leave:
- Trademark and brand scan. Whole-word match against a watchlist of product names, characters and platforms, with word-boundary guards so ordinary nouns survive — the library keeps apple and removed Airbnb.
- Scope check. Does the prompt need a substance, a wager or someone’s diagnosis to land the joke? If yes, it does not belong in a list we send people to for a party.
- Duplicate check. Repeated rows inside one tier are merged, which is why removal and addition counts do not always net to zero.
- Collision check before replacement. A new prompt is accepted only if it is not already in that library and cannot be confused with an existing entry on the drawing pad.
After a pass we rebuild the page, the inline preview data and the agent-facing files from the same source, re-deploy, and re-verify on the live domain that each page’s visible word count still matches its library file. That is how the numbers below were produced.
Change log
| Date | Libraries | What moved | Rule applied |
|---|---|---|---|
| 2026-09-20 | Easy, Medium | Movies tier rewritten: 98 + 98 distinct film titles out of the Movies categories (101 + 100 rows replaced by 101 + 100 original prompts). Measured at whole-file level: 96 + 97 prompts out, 102 + 100 in, plus 26 + 33 duplicate rows merged. The two counts differ because several removed titles are ordinary words too — UP, Brave, Sing and Taxi Driver still sit in other categories of those files. | Filter 1 (trademarks) and Filter 3 (duplicates inside one tier). |
| 2026-09-20 | 9 libraries | Brand sweep: 29 entries replaced one-for-one with generic words (kids (8), funny (7), office (3), bridal shower (1), easter (1), sports (1), charades for kids (1), easy (1), adults (6)). | Filter 1 (trademarks), family-facing lists first. |
| 2026-09-27 | Adults | 64 out, 66 in. Party-shock prompts, finance acronyms and therapy vocabulary. | Filter 2 (scope): the joke must come from social awkwardness, work or everyday life, never from a substance, a wager or a diagnosis. |
| 2026-09-30 | Site-wide | One overstated count corrected (charades “240+” to the exact 220) and every page’s word count re-verified against its library file. | Accuracy rule: a published count equals or understates the file, never exceeds it. |
Current snapshot
Corrections get published here. Use the contact page; if we change a library, this log and the affected page are updated in the same pass.
Who does this work
The entire process above is run by Jordan Mercer, the site's maintainer, who has hosted regular Pictionary-style game nights for families, classrooms, and workplace groups for years. Jordan personally plays every candidate word, writes the grading notes, and answers the suggestions inbox. There is no content team and no anonymous pipeline; the name on the site is the person doing the work. More background lives on the About Us page.