We had a rousing discussion with a group of publishers on how they are using – or not using AI, with special co-host Evan Young, co-founder of HeyNota.
Here’s the audio version:
If you want to skip the discussion you can also watch HeyNota’s demo is here.
Transcription: Introduction
Alisa Cromer: All righty, so this is an experiment. I’ve invited publishers primarily to share the prompts that they are using at work, and I include myself because I’ve been using, prompts and building an AI prompt library just for niche media companies. So this is an opportunity for you to talk to each other because we have a small group.
And also I’ve invited. Evan Young, who is the co-founder of a very interesting company called Heynota, and he knows more than me, so we’re going to start out by cohosting, and then I’m going to leave some time where he can show you the tools that he has developed.
I just want you to see a couple of things.
One is the website where you can access the AI prompt library. I’m reorganizing the navigation, but it is inside the Media Project Library in the right navigation, and the AI prompt library is first on the left.
So that’s where every prompt and tool I test and the prompts from this session will go.
What kinds of AI are useful for Niche Publishers?
Alisa Cromer: I am trying to divide things into buckets. So I’m thinking for a publisher, there’s kind of six basic buckets that we would use AI to automate tasks. So one is editorial, one is communications, and one is sales, SEO, tools, and workflow.
So Evan, are these the buckets publishers should be looking at for using AI, or have I left something out here?
Evan Young: No, I think this is a good start. Maybe something also around reporting or insights from, let’s say, your website traffic, your social traffic. Just how do you understand the success of the performance of the different mechanisms that you use to output data? So that would be one other one.
Do you have an LLM hero on your team?
Alisa Cromer: So, the other thing I wanted to share is this slide. So I get this text it’s, it’s built on a platform called Subtext and it’s from the editor of the Triangle Business Journal every morning. Here is the one from yesterday:
Well, I don’t live in the Triangle, but I love his texts. So yesterday, he had a meeting with some high-power CEOs, talking about what they were thinking about.
And one of the things that they said was, “Get someone in the company to learn LLM; it will pay dividends later.”
I mean everybody here is super motivated, that’s why you’re here. But I do think, and I’ve heard this many times, that media are not spending enough on R&D and, getting those competencies within the company.
So if you want to use HeyNota, for example, who within your own company will give you the heads up and tell you this is a great idea? So somebody should be in there to focus on researching LLM, Large Language Models and tools you can apply.
What do you use AI for now?
Alisa Cromer:
So from here on, this is what we’re doing: I’m just going to ask you to introduce yourself and say what you’re working on. If you have a prompt, great. If you don’t. Don’t worry about it.
So, Betsy, you’re with the boom Athens. Is that correct?
Besty Boom: Yes, and I’m just listening this morning.
Alisa Cromer: No pressure. No pressure. Okay, Chad, do you want to tell us where you are with your journey?
Chad Morrison: Similar to Betsy, I’m just kind of using AI for a little bit of email scripts. And I’m just right in the basic version of it. So I don’t have any apps or anything, but, I’m enjoying the basic Chat GPT.
So I thought I would jump on to learn from another publisher’s perspective on the good tools that you could be using.
Alisa Cromer: And what is your company, Chad?
Chad Morrison: It’s called Craig Kelman & Associates, and we’re based in Winnipeg, Canada. So we’re an association publisher. We partner with associations to publish their magazines for about a hundred different titles.
Alisa Cromer: Greg Martin, you want to jump in and tell us who you are and where you are in your journey?
Greg Martin: I am the associate publisher of ASHRAE Journal, and I am not very competent at AI, but I have been using it for competitive analysis to see where AI would rank us versus our competitors. ASHRAE is a standards writing organization, so we’re very interested in AI’s use of our content to answer engineering questions.
Alisa Cromer: Okay. That makes sense. You know, it’s funny. I do see people using it for things I would never guess. So, Jeff, do you want to, introduce yourself and tell us how you are or are not using AI?
Jeff Gayduk: Sure, sure. Good morning, everyone. Jeff Gayduk here from Premier Travel Media. We’re a niche travel publisher. We have a series of fully-owned titles and do some custom publishing for different travel associations.
I’m probably not the best person to answer prompts because that falls onto various levels as far as exact prompts that we’re using. However, how we’ve used it successfully, for example, for a series of media kits for different products of ours. We use AI to begin looking at the existing structure of the media kits, how we’re wording certain things and running that through AI prompts to maybe clarify the language a little bit, and I think we’ve been pretty happy with that.
Then we’ve taken that next step down to sales proposals. That’s something that I think my salespeople always struggle with is that they’re really excited when somebody wants to buy something – and then it’s like, Oh gosh, we’ve got to put a proposal together.
So we’re really streamlining to give them a series of templates that articulate the product well, based on data that we’re giving it. So it’s not just making stuff up, you know, we’re giving it circulation, we’re giving it, sort of competitive advantages. It’s scouring through our website.
We’re also on the front end of building a directory of field trip locations across North America. And my head tech guy has been using AI to do that. What I can say with that is it meets human intervention, um, because it lies. So, you know, it’s the 80 percent solution maybe right now.
But we, we have to put eyes on it. We can’t just say, okay, here it is. It’s good. So I hope that helps folks.
Alisa Cromer: Yeah. So I am in a similar process with one client that’s a software company, and I am trying to kind of like dumb down or make their language more conversational on the website and, you know, all their marketing materials. The tone that people now want has shifted from “corporate speak” and anything that sounds “brochureish” to simple and authentic, and I am definitely using AI to do that.
And so the prompts there would be, I want a conversational tone. I don’t want a sentence more than 10 words, and, define your audience, those sorts of things. I’m just going to go page by page on the webiste, but I agree., it’s going to be human intervention because left to its own devices, it’s not going to work. However, I believe it can write better than me, with me there.
Jeff Gayduk: Yes, and I would say the big thing as far as media kits, you know, we’ve seen this major transformation for media buyers. It’s younger media buyers, typically female, that won’t pick the phone up. So we’ll literally put that into prompts of like, okay, so how would you say this to somebody who’s 28 years old, marketing, professional advertising agency, but their time is limited and, they need factual information. So, that’s where we have gotten, I think pretty good results.
Alisa Cromer: That means that if you send me a few of these prompts, you get a free membership because I know you’ve got those prompts there that publishers want, such as how to make your media kit more approachable, right? I’ll put them in the prompt library.
How to make a newsletter in two minutes
How many of you have tried to produce a newsletter from your own content using AI? And this is sort of segway to Evan, but has anyone tried that?
(Someone): No.
Alisa Cromer: Okay, so tried it to compare results from ChatGPT, Clod, and Perplexity. And the first thing that all the AI services – except Perplexity – said is that we don’t read links. But they can, so ignore that. The step-by-step way do it in two minutes is at this link.
I show exactly how I did it, but I will tell you now and type the prompts into the chat thread.
First, add feed to your URL. So if it’s niche publisher. biz, It would be nichepublisher. biz/feed/ . When you click that URL, what it’s going to produce is your content in a text editor, which you can cut and paste into the AI query.
Now I don’t write in HTML. I don’t code. I don’t know what any of this stuff means, but because what I publish is in chronological order, I can definitely look through and grab the code for the next few articles that I want, just cut and paste a chunk, and feed it directly into the AI engine.
So first I wrote a little prompt:
Please create newsletter content from the RSS feed for NichePublisher.biz for the dates in the feed below. Each of newsletter itesm should include these elements: a catchy headline linked to the article, and a brief one sentence summary of the content, and “Read More” with a second link back to the article.
After I wrote the prompt, I had to put the HTML on. AI is still saying I don’t read feed, you know how how fast it is. I don’t read feed, I don’t read feed I don’t read links. I don’t read links. Like, I know you don’t.
So I grabbed three stories from the RSS feed, I just cut and pasted it them in. So the order was, I put the prompt first, then I put, I cut and paste the RSS feed.
Yeah, it’s sort of ugly. If you don’t know what it looks like, it’s not pretty slight. It it looks like HTML, but you can read it as a human. It’s not, it’s not that difficult.
GPT and Claude both produced passable newsletters in seconds.
And I’posted the second prompt to write a short publisher’s note. So I wanted the publisher’s note to be writer at the top, but after I had the newsletter so that it could reference the newsletter content:
Please add a catchy one-paragraph intro letter to the top of the newsletter content, written in a conversational voice to an audience of Niche publishers. Elements should include Dear {first name}, one paragraph about this week we are exploring three key topics: What’s print good for, what’s AI suitable for, and how to promote your advertising sales. And a from line, Warmly, Alisa Cromer, Publisher, Nichepublisher.biz
All right, so, and then fed these prompts into the three different AI services.
Very simple. So here’s what happened. I graded each from A+ to, well, it just could do it. First of all, perplexity couldn’t do it. It flunked. Maybe I was doing something wrong, I don’t know, but I will not use perplexity again for this type of task. It’s very good, it does real time, everything is three months behind.
But since I’m using my own RSS feed and my own content, I just want my content. I don’t want to change that much in it.
ChatGPT was sort of flawless.
But I will say Claude was better. Claude gave me fresher, more compelling language that I used, and I felt it improved my writing. And, you know, I’ve been writing and selling advertising for a couple or more decades, and it was better than me. So I was pretty impressed with Claude. So those are the two prompts and and, you know, don’t worry if it says I don’t read links because you’re just going to put the HTML fed in there, don’t bother with perplexity.
My feeling is Claude won hands down.
Has anybody, has any, any of you tried this exercise?
I’ll take that as a no.
(Someone): No.
AI to test email subject lines
Alisa Cromer: Okay, so the other thing I tested put this on the website was the uplift from email subject lines using AI.
So I use MailChimp, it’s not the most sophisticated ESP and it’s very expensive, and I understand that, but one of the things it started doing, is that when I send out an email it will say, Would you like me to rewrite the subject line using AI and send it to the non-openers?
Huh, yeah, I would. This falls into the category where I’m encouraging you to experiment with tools that you’re already using that now incorporate AI. Why not?
So I tried it, and this is what’s really interesting. Not only do my open rate increased by 50%, but without changing anything inside the email, my click through rate. increased by 150%. That 2.5x the clicks.
Now, it was not a straight AB test. I get it. I sent one out Friday at 12:25 and the other one on Sunday at 12:45 in the afternoon, both East Coast times. Both are not great days for a B2B email.
However, when I see these tools, I’m like, wow, they’ve already identified my non-openers and placed them into the queue. That eliminates a couple of steps. Then they had a couple of word choices that I didn’t think of. And I can get 150% lift on click through. I say, I’m going to do this again.
I know every ESP is different, but that’s how I’ve sort of manually been going at this.
So Evan, I’ve given a couple use cases of what I’m doing manually, that your tools automate among many other things.
And by the way, um, HeyNota only works with media companies, they don’t do AI foreverybody in the world. So he’s very dialed in to us.
Where should publisher’s start?
Alisa Cromer: So Evan, where do you think a publisher should start? I mean, I love the media kit and the proposal idea. I mean, that’s number on my list right now to add to the prompt library. What are you thinking that they should be thinking about?
Evan Young: Yeah, I mean, just to start is you just need to start. So, one of the things that we hear a lot is publishers are just fearful of A. I., but as Alisa at least has outlined, it can be extremely helpful and assistive.
So I think the main hoal that you should have is to just identify a use case or a pain point at your respective publication that you spend an inordinate amount of time on. Then identify, a tool that you can leverage to help with automating that process.
And again, I’m not suggesting that AI or machine learning is here to take anyone’s job. It is just here to hopefully help you make you more efficient and help the team that you’re working with do more with less.
If you’re a publisher, you’re likely wanting to talk to sources or write your articles or ensure that, you know, your copy is accurate and complete. You may not want to focus on how I best SEO-optimize this particular story, what the best social caption I can use to amplify this across my platforms, or how to write a newsletter and make it compelling and easy to understand.
So I think the starting point for many publishers should just be to first know that AI is not all bad and that it can be extremely useful. Then start with the test; find the thing that bothers you the most and find a solution that you can apply to that, decide if you’d like to continue to discover more about where else you can find efficiencies.
Alisa Cromer: But I do want to ask Jeff Gayduk because I think the question of personnel is really critical; who is leading the charge within your company?
Is it your IT person? Is it somebody, a young person in the sales department? Are you up late at night at home, like I am doing this weird stuff. Who, who is doing that for you?
One of the reasons I ask is that the publisher of ParentsCanada came to me, and she’s like, my IT guy wants to do generative AI to write the editorial, and the editors are having to fit. She has a fight on her hands.
I’m like, no, no, no, no. You have to get somebody in your company engaged in the right ways – and I think she has now. But you’ve done this so successfully. Who did you get engaged to write the prompts and test these different tools. Who’s doing that? Are you doing it?
Jeff Gayduk: Yeah. So I actually tricked them. When AI first came out, and I started playing with it myself , I created an article about a travel destination and I sent a note to my editors and said, Hey, here’s an article. Written by a created a fictitious name, give me a review on this.
And everyone kind of came back and they had some feedback here and there.
Um, this was Jasper, well before chat GTP. So I brought them into the room after we had this back and forth. I said, “So I have got some interesting news for you. This article was not written by a person. It was written by AI,” and all of them just were flabbergasted. And I said, “Rhis is the future of our business and we can either embrace this or we’re going to get run over by it.”
So at that point we got buy-in from them. Like, okay, this could be beneficial for fact-checking and getting us started article briefs, we use a lot of it for.
So I wouldn’t say there’s one person in our company, but our SEO people use it all the time to research article topics. Our editors use it for article briefs, again with some human intervention. I don’t see every piece of content that’s published by our organization, but I doubt very seriously, that anything gets published, that’s completely AI.
Alisa cromer: So you, you did not assign a lead within your group:
Jeff Gayduk: No, no, just because it touches different elements of the business. Yeah, and I think about your comment earlier. It sort of made me think that maybe we should have somebody who knows more than other people do about this.
But I think we’ve embraced it organizationally just because there are so many efficiencies with it.
Let’s look at HeyNota….
Alisa Cromer: So I will ask Evan to share his screen and walk through his tool. I haven’t seen it in a while. I know it’s evolved. I liked it because You can create a newsletter in a more automated way than I did manually. Theirs is faster and can curate from third-party sites.
And also HeyNota has many SEO tools, which is something I’m working on at my company right now. I do not want to write another meta tag and get all my keywords. I do not want to do that anymore, and I have seen people on WordPress sites that have switched to other providers where they’ve gotten literally a 50 percent increase in traffic when they use machine learning, instead of human meta tagging.
So without further ado, I give you Evan Young (transcript below recording, as well).
Evan Young: Thank you for the opportunity to chat with the team today and thanks all for joining.
So, HeyNota was founded about two years ago by a handful of gentlemen, myself, and my other two cofounders who came out of the Los Angeles Times. There was a lot of frustration around the tools we saw in the market, and we really just wanted to build new tools and innovate for a space that’s traditionally been overlooked or taken advantage of by big tech.
So we created Nota, and we’re working with a handful of amazing publishers, from large publicly traded companies all the way down to extremely local, hyper-local news publishers and everything in between. From TV to lifestyle publishers to radio, etcetera. We’re trying to do everything to help with optimizing your day and your time.
Right now, we’re just driving efficiencies in as many ways as we possibly can think about for those tasks. Alisa just mentioned that she hates doing it and doesn’t want to do it ever again. We want to help you do that with ease.
Alisa cromer: To break this down, what exactly are you doing? Like, I know you put a browser extension on your top of your browser, and it will curate a newsletter from a link. Can you show us that?
Evan Young: Sure. Yeah. So we’ve got a handful of tools that I can show you so we can start with that one. Let’s say that you do want to create a newsletter on you’re on one of your articles that you have on your website. I’ll just use the L. A. Times is an example. The browser extension lives right here within the browser (see video above).
So you can preview what this would look like in somebody’s inbox. If you were to send it across, we’ve also got the, an AI policy that you can add to the bottom as well. If you want to inform your reader base or subscribers that you are using AI to produce these outputs, this is completely optional.
But if you do have an internal AI policy that you want it to share with them, that will be a solution.
Alisa Cromer: How do you get this from here to your ESP? Because everybody’s using different ESPs.
Evan Young: Yeah, so you would hit the export button and there’s a couple of options. You can copy the rich text and paste it into your email template.
What we see most of our publishers doing is copying html to clipboard. So you would just go into your existing email service provider template, add a block of html code as one of like the. Segments you could pull into your template, and then you would just paste the text that we generate for you here in our tool and everything would flow through into your email template.
So in a matter of minutes, you’ve created a net new newsletter.
What we’re seeing a lot of our publishing partners do is to create net new products that they didn’t create before because they just didn’t have the time or bandwidth from their team’s perspective. So maybe you wanted to create a sports newsletter, but didn’t have somebody who could focus on sports or a food or travel or opinion.
Now you have the opportunity to just look at the articles that you want to include in those newsletters and quickly create an output that someone can copy edit and push to their ESP.
Alisa Cromer: Any questions on any of this?
(Someone): It looks great.
Alisa Cromer: It does, doesn’t it? So I write content and I spent a lot of time talking to people and getting use cases, so it’s very expensive content. I can spend a week on one piece of content that I think will help 300 publishers, but I’ve just started a curated newsletter. If I could program it correctly, I think it could save a lot of time.
Chad Morrison: Evan, I was going to ask. So you’re on the publisher’s site here, correct? So let’s say we do some association publishing, and they have members who are either doing press releases or news. Can you utilize this tool on sites outside of your own domain?
Evan Young: Yeah, absolutely. I mean, ultimately, you would have to be mindful of any paywalls you might run into.
But yeah, as long as you have access to the article, you can pull it into the project. And then you also have the capabilities here to, you know, if you wanted to credit any, you know, other publication, you can add that text right into your project as well.
So, this doesn’t just have to be your own, content that you pull into the newsletter product.
Chad Morrison: Gotcha. So you would either have like an attribution there or I’m assuming it’s it’s it’s rewriting this so you’re not plagiarizing what somebody’s already written.
Evan Young: Great question. So everything that we create requires human author content to start. So we’re just looking at the article that you provided us in this link and our tool is very extractive in the sense that with aI tools, you have this thing called temperature.
You can turn up or down in terms of how much you give it leeway to kind of embellish or or get creative. Our temperature is turned all the way down.
We’re not trying to get into a situation where anything that we produce feels like it was factually incorrect or made up. Obviously, we work with a lot of news publishers, so that would be bad for us.
So everything that you see is just understanding the article content and giving you a best practice for the outputs that you’re generating for that respective use case. And we have a bunch of different tools that do quite a few different things.
We have a summary tool that can give you headline suggestions with SEO, you know, for outputs for meta descriptions, tags, keywords, slugs.
We also have a brief tool… somebody, I think, Jeff, it was you who you say you do article briefs.
We can simply create a kind of an Axios-model brief where we give you a headline, a short summary, and some bullet points you can create to, you know, provide a brief. We’ve got video tools that you can use to create videos from your assets.
And we also live in the CMS, as well, so we’ve got CMS integrations with WordPress that live right here at the bottom of the article. So as you’re typing your article, you can just use our tools to get headline suggestions. And you can simply if you like it, you can simply just hit select and it’ll automatically update the article with that content.
Same with your SEO. Same with your social tags or social captions that you might want to create as well.
Which CMS systems does HeyNota integrate with?
Alisa Cromer: So what content management systems are you integrated with?
Evan Young: Yeah, good question. So we have a WordPress plug-in that also works with Newspack, which is kind of a subsidiary of, WordPress for news publishers.
We’re, with the CMS that McClatchy uses. We’re also talking to Blocks right now, and working to get our APIs integrated into their tools. And then we’re, you know, trying to elevate our conversations with WordPress VIP to, you know, get into some more of the, the larger publications there. We’ve also talked to publishers that are on Ghost and IndieGraph and a handful of others that are a bit smaller that we’re considering.
But ultimately, we can move a lot faster than a lot of these, you know, CMS companies can when it comes to integrations. And so ultimately, it’s just a matter of do they have the technology bandwidth or the engineering bandwidth take on a partner like us to, you know, build out what they need to build out so that we can surface our outputs for the customers that they serve.
Alisa Cromer: So let’s talk SEO. Okay, that is my cross to bear. I okay, so I use WordPress with a mostly paywall-blocked model. So you told me that machine learning was going to be better than me.
Evan Young: Yeah, we could test that out. So, if you use WordPress and you have our plug-in, you get page title suggestions. meta or meta descriptions, slugs. You can use all of these again are based on best practices in terms of how many keywords you’re using within the page title.
Evan Young: Ultimately we created our tools to ensure that a human is always present in the output process.
So we tried to minimize the amount of friction we create, but we do want to create a slight bit of friction just because we don’t want people to feel like, things are being published without any human eyes on it.
So as you go through these different outputs and look at, you know what you like and everything here is completely editable. But if you do make those selections, the last step you can take is with our proof tool. And so on the right rail, our proof tool will basically grade your article pre published to see if there’s any opportunities that you can improve it before you actually push it to site.
And so if this right now is good, but it’s not excellent, we want to jump into the analysis and understand what’s happening. So we’ll give you some insights on if it’s the optimal link for SEO, if it’s easy to read, or if there’s any misspellings that you might have that should be potentially looked at.
If you hit highlight suggestion, It’ll take you into the section of our tool where we give you headline suggestions.
One other thing you also notice is we have this microphone with L. A. Times food here. So, one of the tools allows you to create the tone of voice that you want the outputs to be generated in for your respective publication.
And so this has the L. A. Times food tone applied to it. And this is just the out of the box tone from Nota. And so you can kind of look at the differences and see which one you like best. Hopefully, the ones that have your tone or the ones that you would appreciate more and you can select that. And then once you refresh your analysis, you’ll now see that the score should go up quite a bit because you actually added your headline.
There’s probably a few other things that I could do to improve this. But, essentially, this was just a way to kind of have that final step to understand whether or not you should be comfortable with pushing your article to site and you’ve done everything you need in order to make sure that it gets most the most engagement possible.
Alisa Cromer: Well, and thank you. We’re we’re right at the end of our time. I really wanted to thank everyone for. Joining and sharing and, um hope to, uh, to see you all.
Evan Young: Well, thank you so much for allowing me to be here.
Alisa Cromer: Thank you. I’m really glad you came. I know it was short notice. I think I threw this together in a couple of days and I’m so happy everybody came and I’ll see you soon.
Evan Young: Yeah. Okay. Talk soon. Bye.
