What trustwashing means here
Trustwashing is the practice of performing trust rather than proving it: using signals that trigger belief, such as five-star reviews, verification badges, influencer enthusiasm, AI-generated imagery or sustainability language, without the underlying evidence those signals are meant to indicate.
The distinction from ordinary marketing matters, and it is a line this research draws deliberately. Marketing that presents a product favourably is not trustwashing. Marketing becomes trustwashing when it borrows the form of independent evidence without the substance: when a paid endorsement is styled as a spontaneous opinion, when a self-awarded badge is styled as a certification, when a synthetic image is styled as a photograph, or when a review is styled as a customer's experience but was not one. The test applied throughout is whether an ordinary reader would reasonably infer verification that did not occur.
Which sources are accepted
Sources are ranked, and the ranking is visible in how each claim is worded.
- Primary regulatory and legal documents. Legislation, adopted guidelines, enforcement filings and official press releases. These are treated as the strongest evidence for what a rule requires or what an authority alleged.
- Peer-reviewed academic research. Treated as strong evidence for effects and mechanisms, with sample and method noted where they matter.
- Industry and vendor research. Used where it is the only data available, always with the commercial interest of the publisher stated in the same sentence or the source note. A company selling deepfake detection has an interest in deepfake numbers being large, and this research says so out loud rather than quietly relying on the figure.
- Journalism. Used for events and for locating primary documents, preferring original reporting over aggregation.
- Not used: undated statistics, figures circulating without an identifiable origin, marketing blog posts citing each other, and any number this research cannot trace to a named publisher and a date.
How statistics are handled
Every figure in the reports carries a named, dated source. Where a figure is a directional estimate rather than a measurement, it is labelled as an estimate at the point of use, not only in a footnote. Where sources conflict, the conflict is stated rather than resolved silently by picking the more striking number.
Secondary summaries are not treated as sources. If a claim originates in a study, the study is the citation, not the article describing it. Where an original could not be obtained, that limitation is stated.
How the Trustwashing Score is calculated
The 0 to 100 Trustwashing Score in the Brand Report is a self-assessment instrument, not a rating of any company. No organisation is scored, ranked or graded by this initiative. The instrument works by asking an organisation twenty questions about its own disclosure practices across the five forms, weighting them by the potential of each practice to mislead, and returning a band rather than a precise figure. It is designed for internal review. Any score produced is the organisation's own output from its own answers, and this initiative neither verifies nor publishes them.
Evidence levels and limitations
This research has real limits, and stating them is part of its method.
- It is the work of one person, not an institution, and has no formal peer review.
- Much of the available data on fake reviews and undisclosed AI comes from parties with a commercial stake in the problem appearing large.
- Prevalence figures for deceptive practices are inherently uncertain, because the practices are designed not to be counted.
- Regulatory analysis describes publicly available law and guidance. It is not legal advice and has not been reviewed by a qualified lawyer in every jurisdiction discussed.
- Conclusions are research opinions. They are not legal rulings, certifications, audits or findings of misconduct against any named organisation.
Independence, funding and conflicts of interest
Trustwashing.com is self-funded by its author and generates no revenue. It accepts no advertising, no sponsored content, no paid placements and no payment for coverage of any kind. It offers no paid reviews, certifications, endorsements or consulting services.
Declared interest: the author leads design and offline marketing at etikett.de. That employer has no involvement in this initiative, no influence over its topics, findings or conclusions, and no advance sight of its publications. The initiative is published in a personal capacity.
If a commercial relationship ever arises that could bear on this research, it will be declared on this page before the affected work is published.
Use of generative AI
A site arguing for AI disclosure has an obligation to disclose its own. This is that disclosure.
- Research and analysis: AI tools are used to search for sources, summarise long documents and check drafts for internal inconsistency. Every factual claim, statistic and citation is verified by the author against the named primary source before publication. AI output is not treated as a source.
- Writing: the English text of this site and of the reports is drafted and edited by the author, with AI assistance in structuring and editing. Conclusions, judgements and opinions are the author's own.
- Translation: the German, French and Spanish versions of this site were produced with AI translation from the English originals. See the translation note below.
- Not used: AI-generated testimonials, reviews, quotes, sources, statistics or synthetic people presented as real.
Imagery and third-party assets
The architectural photograph used across this site is "Ornate classical architecture with tall columns and urns" by Sam Szuchan, used under the Unsplash licence. It is a real photograph, not a generated image. Attribution is not required by that licence, and it is given here because a site about honest signalling should say where its pictures come from.
The column mark used in the site logo and as a section device is an original illustration commissioned for this initiative. No stock photography, illustration or typeface on this site is used outside its licence terms. All typefaces used here (Source Serif 4 and Archivo) are licensed under the SIL Open Font License 1.1 and are self-hosted; their licence files are published alongside the font files. If you believe an asset here is misattributed or misused, please tell us and it will be corrected or removed.
How translations are prepared and checked
English is the source language and the authoritative version. The German, French and Spanish versions were translated with AI assistance and reviewed by the author for terminology and meaning. They have not yet been reviewed by professional native-speaker translators.
Where a translated page and the English original appear to differ in meaning, the English version governs. Legal texts, including the Impressum, are published in German because they satisfy German legal obligations. Translation errors can be reported through the contact form and are treated as corrections.
Corrections and version history
Substantive corrections are logged publicly on the corrections page, with the date and the nature of the change. Reports carry an edition and a revision date. Typographical fixes are made silently; changes to a fact, figure, source or conclusion are logged.
What this research is not
Assessments published here are research opinions formed from publicly available evidence. They are not legal rulings, regulatory findings, certifications, audits or accusations of unlawful conduct against any named organisation. Where this research describes an enforcement action, it describes what an authority alleged or decided, attributed to that authority, and not an independent finding of its own.
This page is itself subject to revision. Last updated: 5 August 2026.