In August 2026, the transparency provisions of the European Union’s 2024 AI Act took effect. The law includes disclosure requirements for certain types of AI-generated, public-interest text. Interestingly**, the law specifically exempts content that** “has undergone human review or editorial control” and for which a person or organization holds editorial responsibility (meaning that they are accountable for what gets published).
This is a crucial distinction. At NN/G, we use AI throughout the editorial process. AI may contribute words and even some ideas to our articles, but humans remain responsible for editorial judgment. In other words, people with relevant expertise, not AI, have evaluated and stand by the insights and practices communicated by that content.
The NN/G Editorial Process #
I’ve edited articles for NN/G since 2013. Almost every article we published went through my hands. It’s a rigorous, iterative process. The author proposes a topic, and we usually decide whether it’s the right one: is it something we don’t already have an article about, is it something the field needs, and is it a unique perspective or a clarifying framework that nobody else can provide?
And then the revisions come in. It’s very rare that an author’s first draft is published as is. Usually, two editors review for things like logical coherence, UX accuracy, and copy. (It doesn’t mean that all the published versions of our articles are perfect, but we strive to live by these principles.) Then rinse and repeat until the author and reviewers all agree that the version is publishable.
An article can have 2–3 revisions. Some have more than that. The process is more similar to academic publishing than to blog writing.
And everybody goes through the same process — guest authors and people high up in the organization, whether they’ve been with NN/G one day or twenty years. All are subject to the same rules.
This process hasn’t changed since generative AI entered the scene, but my life as an editor has been made easier in many ways.
AI Can Make Writing Clearer #
We have a wide range of authors, and most of them are not professional writers. Some people are better at writing than others. Some (myself included) speak English as a second language.
Before AI, I used to spend a long time rephrasing paragraphs to make them brief yet easy to understand. These days, Copilot in Microsoft Word or Grammarly help me say the same thing more concisely, and often also more clearly.
Of course, AI can also muddy the argument or miss an important nuance while shortening the text. The editor still needs to decide whether the rewrite is indeed better.
AI Can Satisfy Rigid Format Requirements #
Every NN/G article starts with a summary of a fixed length (today it’s 160 characters — the actual limit has varied somewhat over the years). It’s an art to write a paragraph that fits precisely within those limits. Most authors forget about the character count (or forget to count spaces). AI is great at generating article summaries that meet the constraints.
And summaries are not the only rigid format. Our study guides and glossaries also have rigid formats. For example, our study guides include links to articles, with a short description that follows a particular format (e.g., a full sentence with an active verb). AI is great at taking whatever the author generates and fitting it to the requirements.
AI Helps Adapt Existing Content to a New Channel #
We produce content across multiple channels (self-paced courses, reports, videos, and podcasts). Sometimes, an idea that lives in one of these formats needs to be repurposed for an article. For example, an interesting finding in a report, a topic in a self-paced course, a video, or a podcast episode. AI can help extract information relevant to that idea and adapt it to the new format. A human will always have the final pass and review, but much of the tedium of adapting that content can be handled by AI.
AI Can Assist Human Judgment #
AI is also useful beyond rewriting or formatting. It can help editors question both the evidence supporting an article and the logic of its argument.
For example, it can help check facts and verify claims. If an article cites a study or a law, genAI can identify potential misclaims or overgeneralizations that deserve closer attention. The language of the first sentence in this article was tweaked multiple times with AI help to make it reflect the specifics of the EU AI Act. However, AI never certifies that the claim is correct. When it flags an issue, it’s the editor’s or the author’s job to go back to the source, double-check the claim, and correct the error if they find one.
AI can play a similar role at the level of the argument. One of my hardest jobs as an editor is spotting logical issues, such as contradictions or assumptions that the article does not support or explicitly state. I have been doing this for a long time, but it’s still time-consuming. While a human editor will always do this work at NN/G, AI can help us.
For example, an earlier version of this article was centered around the idea that thinking should not be outsourced to AI. But elsewhere in the draft, I had also described AI as a useful critique partner that can participate in the reasoning process. When I asked ChatGPT to critique the draft, it noticed the tension between the two ideas. That forced me to articulate the final principle more precisely. This example is typical of how I use AI to help me critique an article: I share my own assessment of the draft and ask it to challenge or build on it. Sometimes it finds things I missed; sometimes it challenges my assessment; and sometimes it makes recommendations I disagree with. In every case, I weigh its suggestions rather than accept them as answers. I keep those that strengthen the argument and that I believe to be applicable, and I discard the rest.
Editorial Judgment and Responsibility Are Not Outsourceable #
Many writers warn that delegating writing to AI means delegating thinking to AI. One of the benefits of writing is that it gives us time to sit with our thoughts and organize them; in the process, as we discover weaknesses and flaws in our argument, we can make that argument better. It’s an opportunity to grow and practice judgment.
But thinking can happen outside of writing, too. It can happen in the form of a sketch, a discussion you have with someone, or while you dictate your unstructured thoughts to an AI tool. Prompting can be a part of the thinking process.
At NN/G, we don’t prescribe how authors should write or even think. They may work through an argument, discuss it with a colleague, dictate their thoughts and have an AI rephrase them, use AI as a critique partner. We believe that what matters is that a human expert has done the intellectual work needed to understand, evaluate, and stand behind the article.
AI can participate in the writing process, but we do not delegate editorial responsibility to it and we do not name it as an author. Human experts decide whether an idea is worth publishing, whether an argument is sound or useful, and ultimately whether NN/G is willing to put its name behind it.
If you’re curious, read our AI content policy. AI generated much of it using this article as input; humans reviewed, revised, and approved.