Sam Altman says OpenAI has a fix for a telltale sign that you used ChatGPT (www.businessinsider.com)

OpenAI quietly changed ChatGPT’s behavior so the model will stop using em dashes when a user requests that in custom instructions — a stylistic tweak CEO Sam Altman announced on X. Users had complained that ChatGPT often ignored such formatting requests and that the em dash had become a recognizable “fingerprint” of AI‑generated prose. Altman called the fix a “small‑but‑happy win.” The change ties into broader customization features (custom GPTs, memory, and formatting hints) that let users bake preferences into future interactions. The update is significant because it reduces one easy-to-spot signal people and automated detectors used to identify AI writing, complicating attribution and detection efforts in editorial, academic and moderation contexts. Technically, it’s not a change to the model’s core architecture but to how generation obeys user-specified instructions and formatting constraints, so similar stylistic markers (clichés, phrase patterns) still remain unaddressed. For the AI/ML community, this underscores a trend: as models become more controllable at the surface level, detection will increasingly need to rely on deeper linguistic, metadata or provenance approaches rather than single punctuation heuristics.