Medium launches TK, an AI‑powered writing assistant built into its editor

Medium introduced a new web‑based writing app called TK on Monday, embedding an AI assistant directly in the editor so users can request help by typing @TK followed by a prompt. The move is significant because it pushes AI assistance from a post‑draft polishing stage into the very act of drafting, potentially reshaping how journalists and long‑form authors generate content.

---

What is Medium’s TK app and how does it work?

TK is positioned as a lightweight companion to Medium’s existing publishing platform. When a writer types @TK and a request—such as “how many days are in February?” or “suggest a headline for this paragraph”—the app forwards the prompt to an underlying language model and returns a response in‑place. The feature set includes:

• Quick fact lookup – eliminates the need to switch tabs for simple queries.
• Recall of saved snippets – pulls in text the user previously stored while brainstorming.
• Specialised editorial skills – offers headline feedback, style‑guide matching, and other publication‑specific suggestions.

According to TechCrunch, the app ships with the ability to import and export Markdown, though it is not itself a Markdown editor. At launch it is available on the web, with a desktop client slated for a later release.

---

How does the AI assistant integrate into the writing workflow?

Unlike traditional grammar‑checkers that operate after a draft is complete, TK is designed to be invoked during composition. Writers can embed prompts anywhere in the text, and the AI’s output appears inline, allowing immediate iteration. The assistant can also suggest prompts based on the surrounding content, helping users overcome writer’s block.

Medium notes that users may connect TK to their own personal AI agent via the MCP server or Claude skill. This optional connection lets the assistant tap into a user‑specific memory store, potentially remembering recurring themes, preferred tone, or research sources across multiple sessions.

---

What concerns does the publishing industry have about AI‑assisted writing?

The rollout arrives at a contentious moment for long‑form publishing. Over the past year, several authors have lost book deals after publishers suspected undisclosed AI use, and a handful of literary awards faced scrutiny when winners were alleged to have relied heavily on generative models. Critics argue that AI suggestions can blur the line between assistance and authorship, turning model output into the primary text rather than a mere aid.

Additionally, the reliability of AI‑detection tools is under debate. Because large language models are trained on vast corpora of human writing, detectors sometimes struggle to differentiate between human‑crafted prose and AI‑augmented drafts, raising questions about enforcement of disclosure policies.

---

How is Medium addressing transparency and provenance?

Medium acknowledges the risk of low‑quality or indistinguishable AI‑generated content. To mitigate this, TK includes a feature that lets authors publish an explicit note describing how the AI was used—covering the prompts, model, and any external sources consulted. In a blog post accompanying the launch, Medium stated that this approach is “a start of what we believe is required to publish in the AI era, which is to make the provenance of writing clear, transparent, and inspectable, right down to the prompting and source material.”

The company frames the initiative as a way to amplify human intelligence rather than replace it, arguing that AI makes the act of writing more important by handling repetitive or factual tasks, freeing writers to focus on originality and insight.

---

What does this mean for creators using AI tools like Grafy?

For Grafy users, Medium’s TK illustrates a concrete example of AI being woven into a collaborative writing environment. While Grafy does not yet offer a dedicated AI‑writing editor, the Grafy Chat app ([/chat]) lets creators experiment with the same underlying language models that power TK. By prompting the chat with commands such as “draft a blog intro about AI‑assisted writing” or “suggest a headline for a piece on Medium’s new app,” creators can simulate the inline assistance TK provides.

If you need to verify facts or pull in research material while drafting, the Grafy Reader ([/reader]) can ingest PDFs, articles, or other reference files and answer queries in real time—mirroring TK’s quick‑lookup capability. Together, these two apps let you prototype a TK‑style workflow: draft in any text editor, switch to Grafy Chat for AI suggestions, and use Reader to confirm sources.

The broader takeaway is that AI‑enhanced drafting is becoming mainstream. Creators who adopt similar practices—prompting AI for outlines, headlines, or style checks—can stay ahead of the curve, while still maintaining editorial control and transparency through explicit provenance notes.

---

Frequently asked questions

How does the @TK command differ from a regular AI chatbot?

@TK is embedded directly in Medium’s editor, returning suggestions inline where the writer types them. A standard chatbot runs in a separate window and requires the user to copy‑paste text back and forth, making the workflow less seamless.

Can writers disable the AI assistance if they prefer a fully manual process?

Yes. Medium’s launch notes that TK is an optional feature; users can simply avoid typing @TK or turn off the assistant in their account settings, preserving a traditional, unaided writing experience.

Does Medium store the content generated by TK on its servers?

TK’s writing environment is not encrypted, as it is built for collaboration. However, users can set access‑control permissions on each document. The AI‑generated text is processed by the underlying model provider, and Medium advises authors to publish a provenance note if they wish to disclose AI involvement.

---

Source: TechCrunch — AI — Medium’s new TK app brings AI assistance directly into the writing process