As artificial intelligence tools proliferate across the internet, a parallel industry has emerged promising to identify AI-generated content. But recent findings suggest that many of these detection systems may be deeply unreliable, raising broader concerns about trust in digital verification.
Researchers warn that inaccurate readings from such tools could undermine confidence in legitimate content. In some cases, authentic material is dismissed as artificially generated — a phenomenon scholars describe as the “liar’s dividend,” where genuine information is discredited by simply labeling it as AI-produced.
Waqar Rizvi of NewsGuard noted that while much attention has focused on AI being used to fabricate misleading images and videos, a reverse dynamic is now emerging. Increasingly, authentic content is being challenged as fake, creating new complications for fact-checkers and audiences alike.
FCRF Launches Premier CISO Certification Amid Rising Demand for Cybersecurity Leadership
Testing the Tools: Errors Across Languages and Contexts
An investigation by Agence France-Presse (AFP) evaluated several AI detection platforms and found consistent inaccuracies. The tools tested — including JustDone, Refine.ly, and TextGuard — frequently misidentified human-written text as AI-generated.
In controlled tests, detectors flagged authentic material across multiple languages — including Dutch, Greek, Hungarian, and English — as containing high levels of AI content. Even excerpts from established literary works were incorrectly labeled.
In one instance, a human-written report on geopolitical tensions was classified as “88% AI-generated.” The tools also appeared to return similar results regardless of input quality, sometimes labeling even nonsensical text as AI-generated.
Monetization and Misuse Concerns
Beyond inaccuracies, researchers have raised concerns about the business models behind some of these platforms. Several tools not only flagged content as AI-generated but also prompted users to pay for services that would “humanize” or rewrite the text.
In certain cases, these services were locked behind paywalls, with pricing reaching up to $9.99. Experts argue that this pattern — detecting supposed AI content and then offering paid solutions — raises questions about whether some platforms are designed more for monetization than accurate analysis.
Debora Weber-Wulff, a Germany-based academic who studies detection technologies, described such tools as misleading, noting that they often generate incoherent or irrelevant outputs rather than meaningful analysis.
Some platforms have acknowledged limitations. One tool stated that no AI detector can guarantee complete accuracy, while also noting that free versions may produce less precise results due to simplified models.
Implications for Misinformation and Public Discourse
The implications of unreliable detection extend beyond individual errors. Researchers say such tools can be weaponized in political and social contexts to discredit individuals or narratives.
In Hungary, for example, claims circulated that a political document had been entirely generated by AI, supported by screenshots from detection tools. Investigators later found no evidence supporting the allegation.
Experts warn that as AI-generated misinformation spreads rapidly across social media, flawed detection systems risk adding another layer of confusion. Instead of clarifying authenticity, they may amplify doubt and deepen mistrust.
Fact-checkers increasingly rely on a combination of methods — including open-source intelligence, metadata analysis, and contextual verification — rather than depending solely on AI detection tools.
As the information ecosystem evolves, the challenge, researchers say, is not only identifying falsehoods but also safeguarding the credibility of authentic content in an environment where verification itself is becoming contested.
