In a sponsored E&P Magazine webinar, former Google news lead and now Nota product chief Anntao Diaz tried to do something that’s still rare in AI conversations: acknowledge the fear, map the structural problem and then show, in concrete terms, how AI can help newsrooms publish more original journalism without losing their soul.
Diaz, who spent 15 years at Google working with publishers before joining Nota, made it clear early on that his loyalties lie with news. “You are 100% correct. I joined Google 15 years ago because I was, and I am still, in love with the open web,” he said. “For those 15 years I discovered news… no one in my family ever worked in news, and that’s what I decided to focus on.”
From that vantage point, he argued, AI isn’t creating a new crisis so much as exposing a long-running one.
AI anxiety and the structural bottleneck
Diaz didn’t pretend that AI is landing in a calm, confident industry. “Yes, I think it would be very untransparent or just not fair of me not to address the anxiety,” he told attendees. “Very much like I’m a recent dad and I go from very happy to very anxious throughout the day about a hundred times, I think there’s a direct analogy here with AI across newsrooms and, to be honest, across every single industry today, which is: AI as a topic drives anxiety.”
Behind that anxiety, he said, is a structural problem that predates ChatGPT and Gemini. “What I think AI does, right, is that it highlights and underlines problems or situations that have been going on since the beginning of the digital age. AI is just accelerating how deeply we feel that issue as opposed to creating it from scratch,” he said. “And this is what you see here, which is what I call the structural bottleneck.”
On one side of that bottleneck is content creation — all the friction and time it takes to turn raw material into publishable stories. On the other side is distribution — all the platform-specific demands that now sit between a finished story and a real audience. Diaz summed up the result in one phrase: “This is summarized — or I summarize it very humbly — as the low-velocity publication engine. And I believe AI really stresses the situation as opposed to creating it from scratch.”
In other words, if you’re already struggling to move ideas through the pipeline into stories, and stories into the right channels, AI-driven “content slop” and zero-click platforms make that weakness fatal.
Digital gold, not AI slop: why cadence still matters
If the web is flooding with machine-written content, Diaz argued, local newsrooms still hold an asset AI cannot synthesize on its own: lived, reported, on-the-ground originality.
“Where we’re being flooded with content, originality — the fact that every single one of you has content that no one else has because you are on the ground, because you’re creating from scratch things that do not exist online or offline — that is the ultimate value,” he said. “That is what we call digital gold these days.”
The challenge, he said, is to make more of that “digital gold” visible, without diluting quality. “The goal is, or the solution is, to publish more and better content,” he told the audience. “Now, don’t take out words from that sentence. It’s not just more, it’s better as well. More content does not lead to better content.”
Still, the industry can’t ignore volume or cadence. Diaz cited work from the Google News Initiative’s Local Experiments to illustrate the scale of the problem. In at least one experiment, he said, “a good publication frequency to ensure growth for a local news outlet is 15 — one-five — stories per day.” He stressed that it’s “a reference point, not a benchmark,” and quickly added that two deeply reported originals beat a flood of repackaged content. Diaz added: “In a world where readers now scroll past thousands of pieces of content, staying part of the conversation requires showing up frequently.”
That’s where distribution discipline also matters. “Most newsrooms that I’ve worked with across the last 15 years, right, tend to do Google, Facebook and, because of low resources, because it’s extremely hard, maybe an Instagram, maybe a YouTube,” he said. “We need to adopt a full omnichannel strategy to make sure we increase the potential of the reach of our stories.”
There’s a similar tension between the “fewer, better” school of thought and the realities of the algorithm. “Original content beats and trumps all,” Diaz agreed, responding to a question about whether more stories really are the best growth strategy. But, he added, “yesterday you had the issue of your original content… facing maybe 4, 5, 6, 7 different stories. Today and tomorrow that story is facing a hundred, a thousand, two thousand other pieces of content.”
The question for Nota, he suggested, is how to help a one- or two-person newsroom get from one or two solid pieces a day to five, then “seven, and nine, and ten” — without sacrificing the originality that makes their work valuable.
From rat committee to ready-to-edit story
To move the discussion from theory to practice, Diaz walked attendees through a real civic meeting in his own backyard: a municipal committee session on rodent problems in Somerville, Massachusetts, where he lives.
What he had to start with looked familiar to anyone covering local government: a long public Zoom recording and a PDF deck from the city’s health manager. “Essentially what I have here is a lot of raw inputs,” he said. “I have a very long video where they’re doing a Zoom and going through a lot of information. I have access to the PDF that was presented during the meeting… And this is actually from three days ago. So what can I do with this? And I’m gonna try to take this from raw input to a draft ready for me to review and add upon.”
The first step was Nota’s Clip tool. “Okay. So I’m gonna use our first tool, which is called Clip. And Clip essentially allows me to take any type of video content — so this could have been a livestream,” Diaz explained. He uploaded the meeting video, told the system to look for segments involving “rats” and “trash,” and let Nota’s proprietary LLM, Polaris, generate a series of 20–30 second clips and transcripts on those specific topics.
“So what you’re seeing here — you might have seen it pop up — the tool, the LLM, has worked, has processed the transcript, the video, and focused on the topics that I’m interested in, which are rats and trash,” he said. “And it’s organizing 30-second clips maximum… organized by relevancy. So this one on trash is probably the 28-second clip that I care the most about.”
From there, Diaz moved into the CMS. In WordPress, he opened Nota’s Draft tool and uploaded both the full meeting transcript and the rat-control PDF. “When you upload PDFs and they’re turned into text, they always look kinda weird — it’s super hard to process, right?” he said. “And this is — so this is the actual transcript. Every single thing that was discussed. And then I’m gonna click on Generate.”
Behind the scenes, Polaris takes over. “What the LLM is going to do — Polaris, our proprietary LLM — is take those two sources and apply the journalism standards that it has been trained on to create a draft article, inverted pyramid, AP,” Diaz explained. This isn’t a generic chatbot working off web-scale training data; it’s a model tuned for newsrooms. “We have very strong standards in responsible AI,” he said. “We also have a lot of collaboration with this industry specifically. So Polaris does create that inverted pyramid.”
Once the draft is on the page in WordPress, reporters can see more than just text. A preview mode shows the structure applied to the story, and an “Enhance” button lets them pull in context from their own archive. “Enhance is looking at my archives. Essentially it’s looking at past content that I’ve created, right, on that specific topic,” Diaz said, selecting an earlier story titled “Somerville’s Rat Problem Under Control with Innovative Pest Management” to enrich the new draft.
The result, he argued, is not a finished piece but a head start. On a good day, turning hours of public-meeting video and dense civic slide decks into something a human reporter can refine takes “in theory, a few minutes. It’s as fast as that, right?” Diaz said. “Obviously it depends on the topic and depends on the complexity of the story,” but it puts original, local reporting back within reach for understaffed newsrooms.
Access the Slides Used During the Broadcast
Draft, not destiny: keeping humans at the center
If there was a mantra running through the session, it was simple: this is a draft, not destiny.
Attendees quickly noticed that in one of the demo outputs, “Somerville” was misspelled — a reminder that even a tuned model can slip. Diaz didn’t dodge it. “It’s an error,” he said plainly. “With LLMs, the way LLMs are structured, they are probability models and not deterministic models. Hallucinations are not fully removable, right? We can get to a very, very, very low amount, which is, as you heard me say before, the case — but not fully, fully removable.”
Instead of promising perfection, he emphasized human responsibility and control. Inside Nota, the story-generation feature is literally called Draft. “We call it Draft for a reason,” he said. “Because we expect humans to be able to add their savoir faire — to be extremely French — right now, their knowledge on whatever we give them. I think to me it would be extremely worrying if you had a copy which is 100% perfect.”
That philosophy extends beyond typos. Diaz sees AI as a way to accelerate boring, repetitive work so journalists can spend more time on the parts that actually differentiate their coverage. In a question about how to integrate AI without “losing our soul,” he went straight to the heart of it. “I mean, I fully agree with the piece of the question which is in the middle: trusted human voices are essential to communication. I think that’s critical,” he said.
Once again, he came back to originality. “What I strongly believe is the digital gold — I call it — that’s your originality. That’s what you bring in,” he told the audience. “So my answer to that is: one, define your soul, right? What makes you different? And that might be your tone, your voice. And use AI accordingly. AI is just a tool to accelerate what you’re doing, not a tool to replace your soul — to reuse the wording — or replace what you’re doing.”
Let the machines handle the plumbing
Where Diaz believes AI can safely take over is the “plumbing” work that many small newsrooms either don’t do, or don’t do consistently: SEO, metadata, social packaging and cross-channel distribution.
Inside WordPress, once the Somerville rat story draft is generated, Nota offers automated headlines, excerpts, tags, pull quotes, SEO titles, slugs and meta descriptions based on the text. “If I look at SEO, it’s the same logic, right?” Diaz explained. “An SEO title is not necessarily the same thing as a headline. So I’m gonna choose an SEO title. I’m going to choose slugs, which are very important, right? How is your URL structured? A meta description — again, very important. I see here the keywords that were selected right throughout the story.”
Those optimizations aren’t just theoretical. “On the SEO side of things, our average performance, when you use our SEO output, is three-digit growth on clicks,” Diaz said. “So said very simply, on average, our customers who use our SEO output see at least 100% growth on clicks coming from Google. Whether that’s Search, Google News, or Discover.” His advice: “The takeaway on this is: don’t spend time on optimizing for SEO. Let the machine do that part and focus on the source of your value, which is your original content.”
On social, the tool generates platform-specific captions — but with a twist. Diaz worries that if everyone uses the same “commodity LLMs,” they’ll all start to sound the same. “If we all put our content into a commodity LLM, chances are we’re all going to sound the same,” he warned. That’s why Nota includes tone modeling. “Because within Nota, one of the first things we do when you start working with us is we train our tool to replicate your tone — how you desire,” he said. “You could say, ‘I have a tone per content category. I have a tone for a given journalist or for me,’ right? Or actually it’s a brand tone… for the whole brand.”
Even here, the goal is reach plus distinctiveness — staying present in search, social and email without flattening every publisher into the same AI voice. As he put it earlier, “freshness of content is an extremely important factor” for surfacing in search and AI answers, but over time, “authority on any given topic is what allows LLMs, Google Search, and others to prioritize you… within an AI answer.”
Redefining what a small newsroom can do
In one exchange, Diaz described talking to a local outlet that proudly claimed to publish 100 stories a day. On closer inspection, “95% of that was just, you know, AP stories that they weren’t rewriting, they weren’t touching — they were just republishing,” he said. “You clearly understood the publication speed, but not the ‘better’ part of that sentence, right? It’s not original.”
For him, tools like Clip, Draft, Enhance and Nota’s SEO and social modules are about flipping that equation: fewer wire reposts, more local, original work, produced and distributed at a pace that gives it a chance to be seen. “Can you get to five? And then how do we get to seven, and nine, and ten?” he asked of one-person operations.
That doesn’t mean doing everything, everywhere, all at once. Diaz was blunt that most newsrooms are not truly omnichannel today. “There’s this idea that we are omnichannel, but the fact is we are not omnichannel,” he said. “Most news organizations are: Search; a little bit of Facebook; Instagram if you have the resources; and newsletters when you’ve managed to automate it. Right? So extremely far from being omnichannel.”
Still, he insisted that the combination of automation and human judgment can bend that curve. Reporters stay focused on interviews, verification, context and voice; the machine takes first pass at the rest. “In theory, a few minutes,” he said of turning a public meeting into a structured draft package. “It’s as fast as that, right?”=
No one on the webinar panel suggested that AI is “the entire solution.” Diaz didn’t either. “From a sustainability perspective, this is not the entire solution,” he said. “But I do strongly believe that what we’ve built is a significant piece of the answer.”
About Nota
Nota, is transforming how businesses communicate through technology. Nota is redefining business communication by empowering today’s communicators and building the tools they will need tomorrow.
Nota stands at the intersection of messaging and technology, offering assistive AI tools that amplify the reach, consistency, and impact of human-created content. Nota believes that when businesses combine the power of technology with the clarity of a strong voice, they do not just keep up, they lead. The result is high-quality content that informs and empowers audiences.
Comments
No comments on this item Please log in to comment by clicking here