HackerNoon CEO on AI in Publishing: Transparency, Traffic and Human Reporting

For publishers, AI’s newsroom role is taking shape in two places at once: behind the scenes, where it can speed up routine work, and in front of readers, where it can change how stories are found and who receives credit. David Smooke, founder and CEO of HackerNoon, sees room for both uses, but draws a clear line around attribution: readers shouldn’t be led to believe that AI-created material came from a person.

HackerNoon is a technology publisher with more than 45,000 contributing writers and 4 million monthly readers. Founded by Smooke in 2013, it publishes op-eds, tutorials, interviews, columns, research papers and some journalism. Smooke describes himself as more of a writer and product manager than a journalist. His comments offer a view into how a community-driven publishing system can use AI while keeping human judgment involved.

AI as a set of publishing tools

At HackerNoon, AI appears across the writing and reading experience. The company has added a custom ChatGPT layer to its editor for rewriting, uses several image generation models, and generates summaries tailored to the character limits of different distribution channels. These applications put AI to work on specific tasks around a story, rather than treating it as a replacement for the writer.

The tools also extend access to published work. HackerNoon uses Google machine learning to translate stories into other languages and create audio versions of posts. Smooke also sees potential in helping writers brainstorm and fix grammar, helping readers find relevant stories, and helping editors manage content.

His operating principle is that AI can support research, but journalists must check its output. “Never fully trust the AI, and always always verify,” he says. The point applies even when a tool is useful for search: an efficient answer is not evidence that the answer is correct. For publishers, the practical value depends on pairing automation with a person who can assess what the system returns.

Disclosure and author identity

Smooke’s standard for transparency is direct: work made by AI shouldn’t be presented as human-made. Publishers should indicate where and how AI contributed to the reader’s experience. HackerNoon uses emoji credibility indicators to show when AI assisted with a story, giving readers a visible cue about the production process.

That policy addresses a basic trust issue: a byline makes a claim about who created the work. Smooke points to the relaunch of TUAW with AI-generated posts credited to people who had written for the site in the past. The approach drew criticism, he says, because writers don’t want their names attached to work they didn’t make. Readers, too, may value a post differently when they know the human behind it.

Disclosure doesn’t settle every question about AI use, but it makes the contribution easier to judge. A rewrite suggestion, a generated summary and an AI-written article are distinct kinds of involvement. A clear indicator can help audiences understand those differences without obscuring the author’s responsibility for the finished work.

Scale, speed and the cost of discovery

Smooke sees risks in the volume of generated material, including deepfakes that can attract billions of views across social media. Detection and labeling are improving, he says, while production is also becoming easier. The result is a growing supply of synthetic material that can compete for attention with work produced through reporting and other human effort.

Automated writing itself predates the current generative AI wave. Financial websites and tools have used natural language processing to produce headlines quickly, where speed matters. That history suggests the important question is often what task automation serves, and whether readers can understand its role.

Another pressure point is search. If a generated answer on Google resolves a question that previously sent a reader to a publisher’s page, that publisher may lose the visit. Smooke sees a possible counterweight: publishers can use AI through a single Application Programming Interface (API) call to improve their own search and discovery tools. Whether that keeps readers engaged will depend on the quality of the experience and the value of the underlying stories.

AI-curated search products can also help people find material. Smooke says curation adds value and that, with reliable and detailed rules, AI can sometimes curate as effectively as people. He isn’t sure AI alone explains the appeal of Perplexity; its design choices matter too. He notes that Meta AI and SearchGPT have used similar scrolling or source display patterns. Google’s use of generative AI in search results, in his view, points to AI becoming a regular part of internet search.

Attribution, training data and source links

Smooke is skeptical that the New York Times and OpenAI dispute will determine how creators are compensated for material used to train AI systems. He expects practices around training data to face greater constraints, while doubting that this particular case will set the terms for licensing content across the internet. The dispute followed the failure of a content licensing agreement between the parties.

He argues that reproducing another source’s content word for word is plagiarism, and expects OpenAI to pay the Times more before the matter ends. Those are Smooke’s predictions, not a settled account of the case’s outcome. His broader point for publishing is that AI-generated material doesn’t become source material simply because it appears in a generated answer. Sources still need to be cited and linked.

That distinction matters for search and for the publisher’s relationship with its audience. AI summaries can make research faster, but they can also keep readers away from the original page. Links let readers inspect the underlying work and give publishers a path to be discovered. As AI assistants become more common in search, the quality of those connections will shape how much visibility human-written work retains.

Why human stories still matter

Smooke’s advice to writers is to focus on storytelling rather than treat AI as an unbeatable rival. He expects the web to fill with everything from poor material to remarkable work, and believes strong storytellers will continue to find readers. The routes to reach audiences around the world, he says, are more numerous than before.

That optimism doesn’t remove the practical challenges. Publishers still have to decide which tasks suit automation, how to verify AI-assisted research, and how to label contributions clearly. They also need to adapt as search products answer more questions directly. Smooke’s approach at HackerNoon is to use AI for specific editorial and distribution tasks while keeping people responsible for judgment, attribution and the story itself.

His closing argument is simple: people want human stories. For publishers, the durable advantage may come from combining useful AI tools with original work, clear credit and links back to reliable sources.