Stop Using AI for Output — Use It as Your Audience

“How much more can we produce with AI?” is the wrong question.

Everyone is measuring AI adoption by output: more content, more copy, more campaigns, more. Mordy Oberstein and Miruna Dragomir argue that framing has teams convinced they’re getting value from AI when they’re actually getting the least interesting thing from it.

The smarter move? Use it as your audience.

Miruna built a custom GPT trained on her brand’s target personas — not to generate content, but to push back on it. The AI’s job wasn’t to help. It was to simulate a skeptical reader and flag where the messaging wasn’t landing. The result? More useful than any creative brief.

From stress-testing a strategy pitch against a simulated CEO to understanding how LLMs reflect internet-wide perception of your brand, this episode makes the case that AI’s real value isn’t in the output it produces — it’s in the thinking it forces.

Amanda Natividad, VP of Marketing at SparkToro, joins with her own take on using AI for audience research and strategic input rather than task automation.

Transcript

Mordy Oberstein (00:00)
You hear that? Of course not. That’s the whole point.

Welcome to Say It Anyway, where we say the things about digital marketing that you’re supposed to kind of keep quietly to yourself. And we say it — wait for it — out loud.

Each week, we’re taking an honest, maybe uncomfortable look at something marketers need to know, but might be too afraid to say anyway.

Mordy Oberstein (00:27)
I’m your host, Mordy Oberstein, the Head of Brand at SE Ranking, and I’m joined by the impeccably better-in-every-single-way-than-I-am CMO at Planable, Miruna Dragomir. How are you?

Miruna Dragomir (00:37)
I’m good. And that’s a lie. But hey, happy to be flattered. It’s going to give me the confidence I need to record this first episode in a neat and perfect way.

Mordy Oberstein (00:46)
Right. Well, I said it anyway. Get it? Said it anyway?

By the way, not all jokes will be that horrible, but some of them will be. I’m sorry in advance. This is what I do.

First things first, I think we have to pay our respects to the former SE Ranking podcast, DoFollow. This is not a pure SEO podcast. This is about anything in marketing that you just don’t hear people say. We will say it. No filter.

So, a big shout-out to the DoFollow podcast. Go look for old episodes of that. But this is a brand-new podcast. This is Say It Anyway.

Today, we’re taking up: stop using AI for output and use it as your audience — and why that’s just an obvious point you should be doing.

But first, Say It Anyway is brought to you by SE Ranking and Planable, two sister companies helping brands grow their presence across channels. SE Ranking helps you understand your search and LLM visibility, and Planable helps you create and collaborate on the content that improves it across social, website, PR, and beyond — so you can track, distribute, and validate the impact of your data in an AI world.

So Miruna, this whole topic got started because you were working on something with the team using AI that was super cool. It got us thinking that we should just talk about it, because it’s crazy how our natural AI affinity is not to do this — meaning not to do what you did. So, what in the world did you do?

Miruna Dragomir (02:07)
Right. So I built this custom GPT, basically, where I trained the chat to impersonate our personas. I gave it a lot of details about the three personas that we have, and I asked it — in many words — to objectively represent those personas.

I asked it to try not to make me feel better and not to help me with whatever I’m doing, but actually to show me its inner thoughts, its train of thought, while reading my output as a brand, as if it were my persona.

And that got me thinking: this is a very good use case. I’m not saying it’s the only one — it’s one of many — but it got me thinking that when AI first appeared, we all thought, “It can help me do so many things out of the things I was already doing.”

Then AI improved and improved, and we constantly thought, “It can do all of the things that I do,” or at least so many of the things that I normally do.

And I think that’s a box we very often find ourselves in. We get confined into this idea of: these are my tasks, so this is what I can do with AI.

We have that classic trick: ask it to be a senior copywriter and edit your copy, or a senior whatever specialty that comes out of your existing tasks.

But there is this gold mine that we are sitting on: ask it not to be you, but to be whatever you need it to be to stress-test your ideas.

For example, personas — that’s one thing. You have a landing page, copy, an ad, a campaign, whatever it is, and you can ask it to be your audience. Instead of paying thousands or tens of thousands of dollars for a focus group or user test, you can get a glimpse.

I’m not saying it’s the same quality, but it’s going to give you a glimpse. And you’re not going to be able to user-test every ad that you have anyway. So that’s going to help you get closer.

A whole different thing you can use it for is asking it to be your client and pitching your content to it. Train it to be your client: “He doesn’t like this, he doesn’t like that,” whatever.

I’ve asked it to be our CEO once, because I had this whole pitch and strategy, and I wanted to explain the thoughts and ideas that our CEO has, and the thoughts and ideas that I have.

I said, “Here’s my pitch. What are they going to not like? What are they going to challenge me on?” And I got trained for that conversation and improved the pitch because of it.

So yeah, basically, that’s my rant. You can use it for so much more.

Mordy Oberstein (04:41)
So basically, what you want to do with your AI is what I do when I go to my wife and say, “Hey, there’s a local neighborhood WhatsApp group. Someone posted this. Can I reply back with this snarky response?” And she says, “Are you crazy? No, don’t do that.”

Miruna Dragomir (04:55)
Yes. And ask it to actually be your wife and be honest, versus telling you, “Yes, Mordy, you’re right. You’re so good for feeling that way.”

Mordy Oberstein (05:07)
That’s crazy, by the way — that AI does that.

And that’s one thing you need to realize about the output you’re getting, especially in LLMs. We’re talking about this on another series we have called Let’s Chunk. We did an episode with Barry Schwartz — look for it on the SE Ranking YouTube and social media accounts — where AI will try to give you whatever you want.

You ask it, “Are the Jets good?” Well, you know, they’re good. They’re good at losing.

So that’s one thing you need to be careful of. What you’re basically trying to get back, I think, is some kind of honesty out of AI — which, by the way, is inherently novel, because no one’s getting anything honest out of AI.

Miruna Dragomir (05:44)
We have to keep in mind what it’s geared towards. It’s geared towards making you use it more.

How are you going to use it more? If you are happy. If it makes you feel good when you use it.

So if you want a different output — if you don’t necessarily want to feel good about yourself, which, why would you? — you have to ask it not to.

You have to ask it to be intentionally critical, or nitpick at your ideas, or be your very critical and annoying audience that doesn’t care about you as much as you’d like to think, for example.

Mordy Oberstein (06:17)
Back to the whole thing of always focusing on output. It’s always about output — how to do more with AI.

Which, I’ll say it anyway: marketers, when they do that, are inherently frustrated. Teams won’t say this to you, but this is what they are saying: “We’re using the AI thing, but it just ends up being more work than it was to do it the other way, the first time, by myself.”

And I know there’s an initiative to use AI, but maybe it became way more labor-intensive because you’re focusing completely on output.

But here, you’re just asking it to help you think — which may or may not end up in an output that you use.

That’s novel and crazy: asking AI to help you think. Now that I say that, it seems like, wait, what are we advocating?

Miruna Dragomir (06:56)
There’s something there. Have a think about it.

I think there’s a good parallel we can draw. Naturally, before AI, before technology, before you go into an argument, for example, many of us have that internal stress-testing anyway.

You imagine what this person will respond, then you respond back, yada yada, and you go through that. But instead of your internal voice, which can sometimes — many times — reflect your point of view, AI can help you be that voice.

Mordy Oberstein (07:28)
It’s a great idea. Sometimes all you need is something to get you out of your head a little bit so you can see a different perspective, or how someone else might take something.

And even if you say, “Miruna, when the AI fights back with me, it’s going to give me whatever AI argument. It’s not going to be a great argument.” Let’s say you’re right. Let’s say it’s not a great argument.

But the mere fact that it’s showing you a different point of view, or a different angle that maybe you could easily answer, can spark your brain to start thinking differently.

That whole motion of the mind is now different. Instead of being totally inwardly focused, you have an outward entity — an AI — questioning you, which gets your brain thinking in a completely different way.

Miruna Dragomir (08:08)
Yeah, 100%. And it doesn’t matter whether the arguments are good or bad. It matters that, as you say, they’re brought to you.

At the end of the day, you want to resonate with your audience. So if your audience criticizes you, you’re not going to get into a fight. You need to counteract that criticism as much as you can.

Mordy Oberstein (08:24)
It’s like when politicians prepare for a debate. They have all the fake questions first. Or if you’re a baseball player or whatever, you get fake interview questions from the PR team, so you know how to think about the real questions when they come up.

It’s good practice, at least, if anything.

I see this, by the way, coming from the SEO world, which is not really the SEO world anymore. People look at LLMs as, “Okay, it’s a place where I can get visibility.”

I look at an LLM as a place where I can see how the internet thinks very quickly. How does the internet think about a certain topic or a certain person, very, very quickly?

Like, “Hey, LLM, is Gatorade healthy to drink?” It doesn’t call me Mr. Oberstein, but it says, “Hey, Gatorade has a lot of sugar in it,” and yada yada yada.

LLMs portray how the web — as a proxy — thinks of you, your product, and your brand.

So it’s basically like looking at an LLM, similar to how you’re talking about it, as another person. How is this other person understanding you? And are you happy with how this other person, quote-unquote, is portraying you, understanding you, or perceiving you?

Miruna Dragomir (09:31)
Yeah, 100%. At the end of the day, it does get trained on all of the internet out there, including user perception, reviews, social, Reddit — it has a lot of the opinions anyway. So asking it to summarize that can give you a more objective reflection.

Mordy Oberstein (09:51)
Yeah. Just be careful with that also, as a point of order.

Again, like we talked about before, it tries to be positive on purpose. So know how to read between the lines. If it’s not overly positive, that means it’s being negative. Unless it’s gushing — like, “This is great” — there’s some negativity probably built in.

We actually asked Amanda Natividad, the VP of Marketing at SparkToro, about this: how do we start thinking about AI, not from a typical output or task-completion point of view?

We asked her: “What’s an AI use case you’ve personally used where you’ve asked it not to do a task or a job you would normally do, but for research into your audience, strategy, team, and so forth?”

Here’s what Amanda had to say.

Amanda Natividad (10:35)
I’ve used AI as my audience-learning assistant.

One of my favorite use cases is feeding it things like audience comments, interview notes, transcripts, or recurring questions, and asking it to surface patterns.

What are people confused about? What language do they use? What objections keep popping up?

And then I also use it to pressure-test how I respond to those frequently asked questions. The source of information is me. I’m just getting feedback on how I explain it and uncovering ways to improve.

Now, this is all useful because I’m not asking AI to come up with my strategy. I’m using it to help me see the audience more clearly, faster, and from a starting point of real data, real information.

I see AI as a magnifying glass for what’s there.

So if you’re prompting it with a vague understanding of your audience or market, or if you have no understanding at all of your strategy, it’s going to magnify that.

But if you go into it with true insights and your own focused strategy, it’ll magnify that.

Mordy Oberstein (11:38)
Thanks, Amanda.

It’s interesting, right? The way that it magnifies what it’s going to focus on is really dependent on how you use it.

And I feel like we’re — I don’t know, I’m terrible with time — two, three years into this whole AI environment, and we’re still not thinking about how we interact with it in a complex way.

I’m not saying our way is the only complex way to interact with it. It almost feels like a basic way. You’re asking it for input about how you’re doing, and then you act based upon the input it gives you about your strategy and your audience.

But we’re still not focused on that.

Miruna Dragomir (12:13)
Yes. I loved the magnifying-glass analogy. I think it’s so good.

I feel like many of us are still gushing over AI. And as you say, it’s at least two years into it, and we’re so focused on, “Oh my God, it can do so much, and it’s so smart.”

And it is. It is very impressive. I 100% agree.

But you have to control it. It’s on you to control it. Get over the gush and remember: it’s your tool, and your skills have to amp up fast to understand how it thinks, what it does, and that it is a magnifying glass of whatever you input.

And if your input is wrong, then yeah.

Mordy Oberstein (12:51)
Yeah. It’s like when you install tile and one tile is wrong on your new kitchen floor — all you notice is that wrong tile.

And that’s sometimes what AI does, which could be useful in trying to figure out gaps.

But it’s so funny. You’re right. By the way, you see ChatGPT starting — at the time of this recording, at least — to get nailed by all of this.

All of that hype eventually hits a wall, and you start getting into: what do I actually do with this? How do I actually measure this? Does appearing in an LLM actually bring anything to my business? How do I track that?

All those kinds of questions. Eventually, you need to do a real thing with it.

And that’s why, by the way, in the very beginning, enterprise companies were very lagging behind in adoption. And then eventually studies came out saying, “You know what? Enterprise teams lag in adoption of AI.” Gee, I wonder why — because they need to use it in a real way. Not just, “Oh, that is so cool.”

Miruna Dragomir (13:44)
Yeah, it’s less about playing. It’s about actual value added.

Mordy Oberstein (13:49)
And I think a lot of the actual value add is not on the output side. It’s in how you use this as a sounding board for yourself, or as a sounding board that represents what’s happening out there in the internet void, in terms of how you’re being perceived and what people are saying about you.

Miruna Dragomir (14:05)
Yes. And I remembered that a few weeks back, we did this post on LinkedIn and asked people whether they use AI for more volume, more quality, better results — what is it?

And everyone agreed that volume is not the key. But I think naturally, we go into that: let’s automate the post, let’s automate the content, let’s automate.

And volume is key in other areas. In product R&D, volume is a huge thing, and AI unlocks a lot of that. We know engineering resources have been a bottleneck for so long.

But on the go-to-market side, you cannot flood the world with your content. It’s not going to do anything.

Again, let’s not be confined to “we can do things faster.” It’s not just that. It’s not just the speed. It’s the quality.

Mordy Oberstein (14:56)
It’s the Achilles’ heel of internet marketers, and it has been for a long time. This is nothing new.

All the AI stuff — none of this is new. It’s just regurgitated into new problems because of AI.

That knee-jerk reaction of, “I can scale this.” I’ll say it anyway: just because you can scale it does not mean that you should scale it.

Miruna Dragomir (15:14)
I still think like that sometimes. I just made a joke about that in our planning session.

Our paid team was bragging about the good results, and obviously the natural joke is, “Can we double it?”

The easy logical shortcut is: if we double the output, we’re going to double the results. But experience — so much experience — teaches us that it’s not the case.

It’s not the case. That’s not going to double the results. Not just putting all our stuff out there.

Mordy Oberstein (15:41)
That age-old question of, as you increase volume, how do you ensure quality? Again, nothing new. They’ve been dealing with it since they invented butter factories.

Miruna Dragomir (15:51)
Yeah, but AI kind of tempts us. It’s a very big temptation to go into old patterns.

Mordy Oberstein (15:58)
Which I think is the perfect time for our little segment that we’ll do at the end of each show called Clearly Saying.

Because now that you’ve listened to all of that, let us tell you what we are clearly saying in one sentence.

What we’re saying in one sentence: if you’re using AI just for output — if your team is doing that — you are misusing AI.

Wow. It’s scary to have said that. Ooh. Sad but true.

Well, do you think your friends, families, and coworkers are also using AI the wrong way or misusing AI?

Miruna Dragomir (16:15)
One hundred percent.

Care about truth.

Mordy Oberstein (16:27)
Well, say it on social. Share your opinion out there in the world. Don’t forget to tag myself or Miruna, or comment on our posts. Or look for this podcast out there on the Planable and SE Ranking accounts, and tell us if you think your team is misusing AI.

We promise we will not tell your boss. You might be telling them on social, but we will not tell them.

Miruna Dragomir (16:48)
It’s a safe space. That’s the point.

Someone said this in a meeting yesterday: no one is an expert in AI because it’s too fresh to be an expert in it.

So that’s why we’re still in this golden era where we can say anything in a safe space, because we’re all learning at the same time.

Mordy Oberstein (17:06)
Yeah. Unless, by the way, you’re ChatGPT.

Remember when ChatGPT started running ads, and three seconds later there were people saying, “I will optimize your ChatGPT ads”? Those people are…

Miruna Dragomir (17:14)
Pitch. Pitch. Ship.

Miruna Dragomir (17:19)
Yes, it might be the next topic in one of our next episodes.

Mordy Oberstein (17:25)
Which I think means we’ve said enough for today.

So join us again next week as, again, we ask: do you have to be crazy to advertise on ChatGPT? The answer is yes.

Look for a new episode of Say It Anyway on the Planable and SE Ranking websites, and wherever you choose to listen to podcasts.

Please don’t be shy about leaving us a review or rating on iTunes and Spotify. Look for more from the podcast on the SE Ranking social media channels.

And until next time, don’t keep quiet. Bye.

Miruna Dragomir (17:54)
Bye.