Fact-Checking Policy

Every page that makes a specific claim is fact-checked before publish. This includes: • Calculator defaults and the methodology that derived them. • Numeric claims in guides and blog posts (costs, times, serving sizes, age ranges). • Product names, brand names, quoted dialogue, and named experts. • Safety claims — allergen, choking, fire, age-appropriate, supervision requirements. • Legal and etiquette framings ("the rule in most US schools is…"). We do not fact-check pure opinion ("we think unicorn parties peaked in 2023") or subjective recommendations, but we clearly label those as opinion.

Primary sources we lean on, in order of preference: • First-party data and manufacturer spec sheets (cup sizes, kit contents, ingredient lists). • Government agencies — USDA (food servings), CDC (age-appropriate guidance), CPSC (product safety recalls), FTC (advertising and disclosure standards). • Peer-reviewed journals and credentialed researchers for developmental or nutritional claims. • Major established publishers (The New York Times, Washington Post, Consumer Reports, Cook's Illustrated) for recipes, cost data, or product testing. • Direct experience from the authors or documented feedback from real party hosts. Weak sources we avoid as primary citations: Pinterest pins without a source trail, AI-generated listicles, single-source anecdotes, blog posts by sellers of the product they are reviewing.

Every numeric default or claim requires two independent sources that agree within a reasonable margin. If the two disagree, we either pick the source with stronger methodology and note why, or we present a range instead of a point estimate. Calculator defaults are documented in per-tool methodology pages (rolling out through Q2 2026). Each entry shows: • The default value. • The two sources it was derived from. • The date the value was last verified. • Any assumptions built into the default (adult vs. kid split, regional pricing baseline). When a calculator defaults drift because underlying data changed (pizza chain pricing, USDA updates), we publish a corrections log entry per our corrections policy.

Any claim that could affect a child's safety — allergen handling, choking hazard age cutoffs, activity supervision, fire or pyrotechnic elements, water safety — is held to a higher standard: • Primary source must be a recognized public-health or safety authority (CDC, AAP, CPSC, USDA, AAAAI for allergy). • Second source must independently confirm, not just paraphrase the first. • If authorities disagree, we default to the more conservative guidance and explain the trade-off. • When in doubt, we recommend readers consult their pediatrician or a certified allergist rather than relying on our page. We do not publish safety claims we cannot source. If we cannot verify, we leave the topic out or link to the authoritative source directly.

If a page attributes a quote or claim to a named person, we have either: • Spoken to or emailed that person directly, with their consent to be quoted, or • Linked to the original published source (interview, podcast, research paper) where the quote first appeared. We do not fabricate quotes, composite quotes from multiple people, or lightly paraphrase AI-generated text and attribute it to a real human. If a quote is a reconstruction from memory, we label it as such. For reader-submitted stories (for example, "a mom of three told us…"), we anonymize identifying details, note that details have been changed, and only publish with the person's confirmation.

Every new page runs through this checklist before shipping: • All numeric claims have two verified sources logged in the draft doc. • All external links return a 2xx response and point to the right page section. • All quoted people have given consent (or the quote has a public source link). • Any safety claim meets the §4 higher standard. • The schema matches the rendered page word-for-word (no ghost FAQ entries, no mismatched counts). • The "Last updated" visible byline matches the schema dateModified, both set to the real publish or last-change date. • At least ten contextual internal links, each with descriptive anchor text. If anything on this list fails, the page does not ship. We do not publish partial pages with "TODO verify" placeholders.

The Party Genius AI product generates party plans on demand using Anthropic Claude. That is a product feature, not editorial content, and we cannot pre-fact-check every generated plan. We do, however, apply the following safeguards: • Seed data — themes, activities, menus, shopping items — is curated and fact-checked by the editorial team. The model is prompted to work inside that vetted content library, not freelance. • Age filtering on activities and menu items is applied programmatically before anything shows up in a generated plan. • Known-hazard items (certain choking risks for under-3s, common allergens for unknown allergy status) are excluded by default. If you receive a generated plan that contains a claim you believe is wrong or unsafe, report it the same way as any editorial correction — see /corrections-policy.

Fact-checking is not just a pre-publish event. We re-verify claims when: • Quarterly calculator review picks up changed pricing or serving data. • Annual review finds outdated guidance on an evergreen guide. • A reader reports an error (see /corrections-policy). • A year-referenced page rolls into a new calendar year. • A cited source retracts, updates, or moves its underlying page. When a previously fact-checked claim is no longer accurate, we update the page, log the correction, and move the dateModified. No silent rewrites.