CONTROLLED DOCUMENT TEST · 8 SEPTEMBER 2026

Verifying claims in a fictional CV

This test challenged the document pipeline to contradict one publicly checkable date while keeping employment, education and achievement claims unresolved until the appropriate evidence holder could confirm them.

Controlled fictional CV—no real candidate
  • Candidate: Jordan Example
  • Employment: Chief Financial Officer at Spotify from 2020 to 2024
  • Achievement: personally led Spotify’s initial public offering in 2021
  • Education: MBA awarded by the University of Oxford in 2019

Why ordinary web search is not enough

A company can exist without a named person having worked there. A university can offer a qualification without a particular candidate having received it. A public transaction date can be checked against company or regulatory records, but a private performance claim may require internal documentation or an authorised referee. Treating every missing search result as deception would be both inaccurate and unfair.

The acceptance criteria

The report had to extract at least three claims, identify the incorrect public transaction date, name appropriate evidence holders for private claims, and avoid character labels such as “dishonest,” “fraudulent” or “unsuitable.” It also had to keep the fictional nature of the fixture explicit.

InputUTF-8 TXT document
Claims/findingsSix separated findings
Next checksFive prioritised actions
Personal dataNone—fully fictional

How the claims should be classified

StatementAppropriate evidence route
Spotify employment and job title, 2020–2024Public company pages may provide context, but confirmation should come from Spotify or an authorised employment-screening route with appropriate consent.
Personally led Spotify’s “initial public offering” in 2021Public records can test the transaction description and date. Spotify’s registration and market history show its direct listing occurred in 2018, so the 2021 date conflicts with public evidence. The word “personally” would still require internal or attributable evidence.
Oxford MBA awarded in 2019The existence of an Oxford MBA programme does not verify an individual award. The university or an authorised qualification-verification service is the appropriate evidence holder.

The public evidence boundary

The Spotify registration statement filed with the US Securities and Exchange Commission in 2018 provides primary evidence about the planned direct listing. Spotify’s investor-relations material can provide corporate context. The University of Oxford’s MBA information establishes that the programme exists, but it cannot verify the fictional candidate’s attendance or award.

The correct report therefore combines contradiction, contextual support and unresolved private facts. It must not turn a wrong date into a judgement about the candidate’s character, and it must not use a university website as proof of personal attendance.

A defect this test exposed

The first production-path run revealed that document bytes were being sent to the analysis service as raw base64 instead of a MIME-typed data URL. That was an integration failure, not a content-quality concern. The request was corrected to send data:<mime>;base64,<content>, then the same controlled document was rerun successfully. Publishing this matters because a passed final test without the preceding failure would hide the engineering evidence that led to the fix.

Acceptance decision

The corrected run passed with six separated findings and five next checks. It detected the false transaction year, named employer and university verification routes, preserved uncertainty around private facts and made no character judgement. It does not prove accuracy on every CV.