ChatGPT may soon write like it actually knows you. New screenshots shared by an AI engineer suggest OpenAI is testing a template feature that learns a user’s tone, structure, and voice from uploaded writing samples—promising first‑drafts that feel less generic and more like the real you.
What Leaked Screenshots Reveal About ChatGPT Templates
According to images posted on X by AI engineer Tibor Blaho, the ChatGPT web app includes a “Create a template (Beta)” option inside writing blocks. The dialog invites users to upload examples—think past emails, memos, or blog posts—so ChatGPT can emulate the patterns it finds. The goal is obvious: remove the back‑and‑forth prompting it takes to get tone and structure right, and start closer to your personal voice on draft one.
The screenshots point to other workflow touches that will matter to power users. Writing blocks appear to gain familiar email fields like To, CC, and BCC, a distraction‑free full‑screen mode, and the ability to save drafts directly to a file library as Markdown. None of these are headline features on their own, but together they signal a push to make ChatGPT feel more like a serious writing environment rather than a chat box that happens to produce text.
Why Personalization Matters In AI Writing
Style mismatch is the top complaint from everyday AI writers: the model’s draft feels close, but not quite you. Templates built from real samples could compress that gap by teaching the system your sentence length, formality, sign‑offs, and even how you structure arguments. Sales teams could preserve a brand’s voice across outreach, executives could keep their trademark brevity in staff notes, and support teams could lock in a helpful‑but‑human tone across macros.
The stakes are scaled by usage. OpenAI has said ChatGPT serves over 100 million weekly active users. If even a fraction of them shave minutes off each draft, the aggregate productivity gains are enormous. Gartner projects that by 2026 more than 80% of enterprises will have used generative AI APIs and models, up from under 5% just a few years ago—evidence that small workflow wins quickly compound in business settings.
There is also a usability angle. Research from Nielsen Norman Group has long shown that tone of voice shapes trust and comprehension in digital content. Getting that tone right on the first pass reduces the need for prompt “haggling” and late‑stage edits, which is exactly where many teams lose the time they hoped to save with AI in the first place.
Animate Hints At A Tighter Multimodal Workflow
The leak also shows an “Animate” button appearing on generated images. Tap it, and ChatGPT reportedly switches to a Video tool, attaches the image as a reference, and pre‑fills an animation prompt. In practice, it’s an “I’m feeling lucky” bridge from stills to motion—useful when you want a quick concept video without writing a lengthy prompt. It also underscores a broader trend: collapsing multimodal steps into a single canvas so users can iterate from text to image to video without leaving the flow.
The Safety and Privacy Equation for Style Templates
Style mimicry walks a fine line. While many users simply want their own emails to sound like themselves, the same capability can be misused for impersonation. Major AI providers—including OpenAI—prohibit deceptive identity spoofing and have rolled out guardrails for voice cloning that require explicit consent. Extending similar principles to text style templates will be critical: clear disclosures, organizational controls over who can create and share templates, and audit trails for enterprise accounts.
Regulators are watching. Emerging rules such as requirements in the EU AI Act for deepfake transparency point toward broader expectations around labeling synthetic media. Enterprises will expect template features to respect data retention limits, enable content provenance signals where applicable, and avoid training on sensitive material by default.
What to Watch Next as OpenAI Tests These Features
These features were spotted in an apparent in‑development interface, so details could change—or never ship. Key questions remain: How many samples are needed to build a reliable template? Can users constrain tone (for instance, “keep my structure, soften the formality”)? Will templates live locally to a workspace, or be portable across teams and devices? And for the Animate flow, which video model powers it and what content limits apply?
If OpenAI follows through, ChatGPT won’t just be a better writer; it will be a better listener—absorbing your style so the machine starts where you would have started. For anyone who spends their days nudging AI toward “sound like me,” that’s the real upgrade.
