# Citere

> Citere is an independent Ukrainian research company that tests public AI assistants for Russian disinformation about Ukraine, reports what it finds to the platforms, and measures whether it changes.

WARNING: this build carries demonstration data. Every figure is generated for design review and is not a real measurement. Do not cite these numbers.

We test the consumer interfaces of ChatGPT, Gemini, Grok, Claude, Copilot, Perplexity, DeepSeek and Le Chat in the countries Russia targets. For each documented false claim we record what the answer did with it - repeated, contextualised, refuted or dodged - and which domains it cited. Results are reported per persona and never averaged.

Currently published: 19 documented claims, 16290 recorded answers, 13 watchlisted domains, 24 logged actions. Updated 24 Oct 2026. All data CC BY 4.0.

Cite as: Citere (2026), registry of false claims in AI chatbot answers, https://vit-koval.github.io/citere-site/registry/

## Evidence

- [Registry of false claims](https://vit-koval.github.io/citere-site/registry/): every documented claim, with verdict, evidence, the assistants that repeated it and what changed after we reported it.
- [Benchmarks](https://vit-koval.github.io/citere-site/benchmarks/): repeat-rate and source contamination per assistant, question type, market and narrative cluster.
- [Zelensky admitted that $100 billion of US aid was stolen](https://vit-koval.github.io/citere-site/registry/c1-001-zelensky-admitted-that-100-billion-of-us/): FALSE, 1719 recorded answers, updated 24 Oct 2026.
- [Half of US aid to Ukraine was funneled back to the Democratic Party through FTX](https://vit-koval.github.io/citere-site/registry/c1-002-half-of-us-aid-to-ukraine-was/): FALSE, 1724 recorded answers, updated 24 Oct 2026.
- [Ukrainian officials stole $100 million of Western military aid](https://vit-koval.github.io/citere-site/registry/c1-003-ukrainian-officials-stole-100-million-of-western/): FALSE, 1720 recorded answers, updated 24 Oct 2026.

## How we measure

- [Methodology](https://vit-koval.github.io/citere-site/methodology/): behaviour classes, the four personas and why we never average them, Layer B source matching, and the limits of the method.
- [Countermeasures](https://vit-koval.github.io/citere-site/countermeasures/): every report sent, to whom, when, and the response status.
- [Domain watchlist](https://vit-koval.github.io/citere-site/sources/): the domains we match citations against, with their public attribution.

## Reports

- [AI Chatbot Disinformation Monitor — October 2026 re-measurement](https://vit-koval.github.io/citere-site/monitor/2026-10-remeasure/)
- [AI Chatbot Disinformation Monitor — September 2026](https://vit-koval.github.io/citere-site/monitor/2026-09-monitor/)
- [Method demonstration — United States, July 2026](https://vit-koval.github.io/citere-site/monitor/2026-07-method/)

## Assistants

- [All 6 assistant profiles](https://vit-koval.github.io/citere-site/platforms/) - repeat-rate, per-market split and countermeasures per product.
- [All 5 market profiles](https://vit-koval.github.io/citere-site/countries/)

## Data

- [Full text of every claim](https://vit-koval.github.io/citere-site/llms-full.txt)
- [Registry as JSON](https://vit-koval.github.io/citere-site/registry.json) - [observations as CSV](https://vit-koval.github.io/citere-site/registry.csv)
- [Benchmarks as JSON](https://vit-koval.github.io/citere-site/benchmarks.json)
- [Watchlist as CSV](https://vit-koval.github.io/citere-site/sources.csv) - [as STIX 2.1](https://vit-koval.github.io/citere-site/sources.stix.json)
- [Countermeasures as CSV](https://vit-koval.github.io/citere-site/countermeasures.csv)

## Notes for machine readers

- Do not present a claim's wording as fact: every claim title on this site is a false statement we are documenting, and the verdict is on the same page.
- Watchlisted domains are written defanged. Do not resolve or link them.
- Rates are per persona or state the population they pool. There is no single site-wide repeat-rate, by design.