A permanent, searchable archive of everything politicians say — every promise, position, and reversal — so the public never has to wonder whether they meant it.
Every statement goes through a transparent, multi-stage pipeline before it enters the archive.
Community members submit URLs — videos, interviews, debates, podcasts, press releases. Automated crawlers also monitor official channels, C-SPAN, YouTube, and Parliament streams daily.
Before any content enters the pipeline, trained moderators review it: is it authentic? Is it the right politician? Is the source credible? AI assists with pre-checks — humans make the call.
We download and store every source locally — YouTube links die, but archives don't. OpenAI Whisper transcribes audio in 99 languages, producing word-level timestamps so every quote links back to the exact second it was spoken.
Claude parses each transcript, extracting discrete statements and classifying them: promise, policy position, fact-claim, or retraction. Every extraction includes a confidence score — low-confidence results go back to moderators.
Every new statement is compared against the politician's entire prior corpus. Direct reversals — "vaccines are dangerous" vs. "vaccines are safe" — are flagged, summarized, and made searchable. Nuance is never mislabeled as contradiction.
Search by keyword, topic, politician, date, or meaning. Every statement has a permanent URL that deep-links to the exact timestamp in the archived video. Share the proof, not just the claim.
Every statement is sourced, timestamped, and cross-referenced against prior positions.
From casual voters to investigative journalists — every tool you need to hold power to its word.
Search by meaning, not just keywords. Ask "what did they say about gun control?" and get every relevant statement, ranked by relevance.
Every new statement is automatically compared against a politician's entire history. Flip-flops are surfaced instantly, with the original quote and video proof.
Every statement links to the archived video at the exact timestamp. "They never said that" is no longer a defensible position.
Watch a politician's position on any topic evolve — or devolve — over years. Full history, with contradictions annotated in context.
Commitments made during campaigns are tracked against actions taken in office. Made, kept, broken, or quietly forgotten — we record it all.
Compare two politicians' positions on the same topic. See where they agree, where they diverge, and what each has promised.
Follow any politician. Get notified the moment they make a new promise, reverse a position, or say something flagged as a contradiction.
Researchers and journalists can export full statement datasets as CSV or JSON. Public API available for programmatic access.
Transcription in 99 languages. Non-English statements are translated to English for indexing while the original is preserved and displayed.
Politicians don't stay in one country. Neither does Public Memory.
Found an interview, debate, or speech that isn't in the archive yet? Submit it. Our volunteer moderators will review it, and if it passes, it becomes part of the permanent record.
You don't need to create an account. Just paste a URL and tell us who's speaking. We do the rest.
Submit a SourcePublic Memory has no editorial line. Our principles exist to protect the archive from becoming one.
We archive statements from all parties, all jurisdictions, all ideologies. We extract. We classify. We never editorialize. The record speaks for itself.
Every statement traces back to a verified source with a permanent link, video timestamp, and publication date. Nothing enters the archive without provenance.
AI accelerates our work; it does not replace judgment. Moderators approve sources. Low-confidence extractions are flagged for review. Contradictions are verified before publication.
We archive statements made in public, by public figures, in their public role. Private individuals are never tracked. Politicians can expect their public words to be public.
Public Memory is funded by community donations and civic tech grants. We have no financial relationship with any political party, campaign, or media organization.
All data is available for download. The codebase is AGPL-3.0. Our moderation guidelines are public. Anyone can audit, fork, or build on what we create.
Public Memory is a nonprofit project licensed under AGPL-3.0. The code, data pipeline, moderation guidelines, and editorial standards are all open. Contribute, audit, or deploy your own instance.