The Big Debate: Non Keyword Signals - AI Chats & Privacy

Summary

---

Podcast Summary:

The Digital Marketing Podcast
Episode: The Big Debate: Non Keyword Signals - AI Chats & Privacy
Hosts: Daniel Rowles & Ciaran Rogers
Date: December 8, 2025

---

Episode Overview

In this episode, Daniel Rowles and Ciaran Rogers dive deep into the evolving world of digital marketing signals. They debate the increasing importance of "non keyword signals" in a landscape shaped by AI-powered search, shifting privacy boundaries, and changes within platforms like Google and Meta. Central themes include the decline of traditional keyword-focused SEO, the rise of AI-driven behavioral insights, and the complex ethical questions marketers now face regarding data privacy and user consent—especially related to AI chat interactions.

---

Key Discussion Points & Insights

1. The End of Keyword-Only Targeting and Rise of AI Signals

• Context: Google Ads' "AI Max" campaigns and similar AI-driven approaches now focus on a wider range of signals beyond traditional keywords.

• Observation:

  • There's a dramatic increase in search impressions but a corresponding drop in clicks as AI answers more questions directly in search results.
    • "If you look in your search console reports... massive increase in impressions, but quite a big drop in the actual number of clicks that you're getting." – Ciaran [00:41]
  • Even with effective SEO, click-through rates are plummeting because users get their answers in summaries or featured snippets, bypassing publishers’ sites.
    • "Our data... our impressions have pretty much doubled over the last 12 months. ... Our clicks at the same time have halved." – Daniel [01:21]

• AI’s Expanding Dataset:

  • Google is correlating not just search queries but a broad array of user behaviors (e.g., video watching, email activity, location via Maps).
  • AI employs this data to infer intent and interest in real-time, enabling "almost anticipating what you might be interested in or looking at next." – Ciaran [02:38]

• Campaign Performance:

  • AI-maximization campaigns are leading to "20, 30% uplift in results, basically by using ... keywordless signals." – Ciaran [03:39]

---

2. Privacy Implications and Ethical Dilemmas

• Core Debate: How far should platforms go in aggregating and leveraging behavioral data to personalize and target ads?
• Hosts’ Approach: Actively invite listener engagement and debate via LinkedIn, emphasizing the need for community perspective.
• Quote: "We want to find out how you feel about the privacy implications of these things." – Daniel [04:28]

Google vs. Meta Approaches

• Meta’s Evolution:

  • Meta (Facebook/Instagram) has shifted from using "connections" and explicit likes to heavily AI-driven content recommendation based on non-keyword signals—especially passive video consumption behaviors.
  • "The algorithm then has to change hour by hour, day by day because you might be in the mood for watching recipes at one moment and the next week you might be interested in fitness." – Daniel, referencing Alex Schultz [05:49–06:45]
  • Platforms dynamically adjust what users see using these complex behavioral metrics.

• AI Chat Data and Privacy

  • Meta announced a policy change: conversations with AI agents (e.g., in Messenger, Instagram, WhatsApp) may be used to target advertising.
    • "Those conversations, what they learn from those conversations and the insights they can glean can then be used to target advertising." – Daniel [07:04]
  • Ethical and practical uncertainty: how much privacy and anonymity should users expect from AI chat?
    • "Do we expect our AI chats to be anonymous?" – Daniel [07:29]
  • Distinction raised between user expectations for paid chat platforms (ChatGPT Pro) vs. open platforms.

Brand-Level Exposure & Data Control

• Brand Chatbots on Meta/WhatsApp:

  • Ciaran highlights concern over businesses using Meta-hosted chatbots: will AI-driven recommendations leak sensitive purchase intent signals to competitors?
    • "If any interactions with like AI chats within the meta platform, like how far does that extend? ... Does that mean those signals are going to be used by my competitors and available to buy within the ad space?" – Ciaran [08:24]

• Defining "Sensitive" Data:

  • Meta claims sensitive data will be filtered, but there’s ambiguity:
    • "How do you define sensitive data, actually? Because it's not a black and white thing." – Ciaran [09:10]

---

3. Notable Hypotheticals & Grey Areas

• Daniel's Hypothetical Business Model:

  • Creating a free expert chatbot for sites, with a catch: all data input could be used for ad targeting.
    • Raises the question: Is it ethical or even legal if buried in T&Cs?
    • "All I've done is in my terms, conditions. I've just said you can have this for free, but in exchange we get to use the data for targeting." – Daniel [10:43]
  • Further debate on whether website owners and visitors are truly aware of the implications.

• Trust in Conversational AI:

  • The more natural the AI-human chat, the more likely users are to disclose sensitive information—raising the stakes for responsible data handling.
    • "You implicitly trust it a little bit more. Like you start having a much more human conversation with it." – Daniel [11:15]

• Organizational Data Risk:

  • Daniel recounts practical risks: with organizational AI tools like Copilot accessing legacy cloud storage (e.g., OneDrive), poorly set permissions can accidentally expose sensitive information.
    • "Suddenly it's got access to their OneDrive and they haven't got their settings set up right…suddenly I can say, how much does Kieran earn?" – Daniel [12:56]

---

Notable Quotes & Memorable Moments

• On the impact of AI summaries:

  • "The click through rate isn't this. It was necessarily. And that's again, people aren't just aren't clicking right. Because they're getting, giving the answers..." – Ciaran [01:35]

• On platform privacy policies:

  • "How far does sensitive go?" – Ciaran [09:10]

• On the gray zone of free tools and user consent:

  • "Could you though? Is that explicit concept? Was that made clear? Did I opt into that?" – Ciaran [11:09]

• On human trust in AI chats:

  • "Because you implicit when you talk to somebody, you implicitly trust it a little bit more..." – Daniel [11:15]

---

Community Engagement & Listener Challenge

• Debate Invitation:
  • Listeners are encouraged to join the debate on LinkedIn (search for "Target Internet" or "Daniel Rowles") and via the podcast website for a chance to win books and merchandise.
  • Aim: Gather opinions and real-world experiences related to AI, privacy, and non-keyword signal marketing (see [14:43]).

---

Timestamps for Key Segments

• Introduction to Non-Keyword Signals – [00:10]–[02:38]
• AI Summaries and Decline in Organic Clicks – [01:21]–[03:39]
• Privacy Debate: Google, Meta & Chatbots – [04:28]–[08:24]
• Data Use Cases – Hypotheticals and Concerns – [10:25]–[12:21]
• AI Tools, Organizational Security Risks – [12:56]–[14:11]
• Listener Engagement and Call to Action – [14:43]–[15:13]

---

Tone and Final Thoughts

The episode is analytical but conversational, with a healthy skepticism toward big tech’s use of data and a strong call for marketers and consumers alike to critically examine the boundaries of privacy and consent. Both hosts openly question the implications of current trends while encouraging a broader community dialogue.

“I think it's just an interesting area... Big questions.” – Ciaran [14:11]

---

Join the debate and share your perspective:

• LinkedIn: [Target Internet / Daniel Rowles]
• Podcast website: targetinternet.com/podcast

---

Transcript

[A]

Hello and welcome back to the digital marketing Podcast brought to you by targetinternet.com My name is Kieran Rogers.

[B]

And I'm Daniel Rolls.

[A]

And today we have a big debate on non keyword signals.

[B]

Okay, so non keyword signals have become a big thing. Initially where I first came across this was actually you a few months ago. You first pointed out to me about these AI Max campaigns within Google Ads. And what it basically says is you go in there and it says we're not just targeting our keywords anymore. So just talk to us about that for a second.

[A]

Kira. So with the advent of AI summaries and like AI questions and AI answers, you know, Google has sat on not just what keywords did Google people type in, but also what did they do next. And we all know that, you know, there's been a big drop for a lot of us on like organic phrases that people used to click on. You know, if you look in your search console reports, very common thing I see in lots and lots of different brands, search console is massive increase in impressions, but quite a big drop in the actual number of clicks that you're getting. And that makes sense, right? Because actually people are getting the answers given to them. They're not necessarily having to kick around anymore.

[B]

Just to give you an example, our data, so to interrupt, Kieran. But the if we look at ours, our impressions have pretty much doubled over the last 12 months. So we've done a really good job on SEO. Our clicks at the same time have halved. So that kind of gives you the scale that you can be dealing with.

[A]

Yeah, and the click through rate isn't this. It was necessarily. And that's again, people aren't just aren't clicking right. Because they're getting, giving the answers to lots of, you know, top of funnel questions through the AI. And that's bad news. If you're a publisher that spent years optimizing for, you know, the obvious phrases and fighting to get your way to the top, you could still be at the top, but no one's clicking. Why? Because the AI means that they don't need to. But you see there are lots of signals that Google are getting that tell it or who's asking that question. So I don't know whether you then went on to click because they didn't click. But I know not just what question did they ask then in that moment of time, but what questions have they been asking today and yesterday and over the last 21 days? And what can we infer from that? Now AI is brilliant at joining dots on data like that. And I think that's exactly what, you know, Google's engineers have turned their attention to teaching the AI how to manage it. But it goes beyond that, right, because actually they're also seeing what videos are we watching, you know, and what emails are we getting if we.

[B]

Where you going because of.

[A]

Yeah, where are you going because of Google Maps. And actually they're able to pull together literally, potentially, who knows, you know, it's a secret source, isn't it? But potentially millions of different data points for millions of different people and to build, you know, a really interesting map of like almost anticipating what you might be interested in or looking at next, potentially with that kind of level of data, if you've got enough data points, it's not too hard to make inferences on that. And I think, you know, this is all. It's all keywordless. You know, it's not stuff that people have necessarily searched on or clicked, but actually when you start joining multiple pieces of information together like that, it presents a really interesting picture and could, you know, indicate that certain people are in the market for certain different products or services. I think that's, for me anyway. That's what I'm assuming the AI max type campaigns are enabling and they won't necessarily get it 100% right, but actually they get close to it and close is good enough, actually, when you're in a competitive marketplace. And actually I've seen, I was very impressed with the sort of research campaigns that they'd run with various beauty brands and the uptick levels were. Were really impressive. They were sort of quoting 20, 30% uplift in results, basically by using the. This keywordless. These keywordless signals. And that's exciting, right? It's particularly exciting in an environment where, you know, we've perhaps seen the bottom drop out of our organic clause, should we say, from our search engine optimization. Actually, a new thing like this comes along and actually it makes sense to me that actually Google is not just using AI to generate, you know, very brief summaries, but actually it's using it to help join the dots together and work out, well, who can we help more based on what they've told us or what they've done or what they've searched on or where they've been. Let's join all of that together and present a unified, like, solution to it.

[B]

All right, so let's just say why this, we call this the big debate non keyword signals is we want to find out how you feel about the privacy implications of these things.

[A]

Okay.

[B]

Because already some people are concerned about privacy because. And it's not just Google, you know, we're going to talk about Meta in a moment. Lots of other people, they're all collecting lots of data about us. They're then using that for targeting stuff at us as well. But we want to get your opinions on this. So for the first time what we're going to do is we're going to set up a bit of a starting point for a conversation in LinkedIn. So you can go to the target Internet, LinkedIn, you can just find me Daniel rolls on LinkedIn. And actually if you just go targetinternet.com forward/podc, you will find the link through to the conversation. But we want as many of you as possible to go through and join the debate. And if you do join the debate to encourage people to do that, we have got some gifts and prizes to give away. We've got some books and merchandise, those kind of things. We'll just randomly pick some people that engage in the debate to get some of those as well. Now I said it's not just Google because I was interviewing Alex Schultz a couple of weeks ago, we released the podcast. He was the CMO of Meta and he'd written this amazing book called Click Here about digital marketing analytics. But one of the things he said which totally relates to this non keyword signal side of things, he said that our business has changed completely in the last six years. What we do is completely changed.

[A]

I said well, what do you mean?

[B]

And he said well, essentially we, everything we showed was connected. I, we showed you some content across maybe Facebook or Instagram because you'd like to page you'd followed someone, you'd indicated you liked a piece of content, you were connected to someone in some way and we could use all those connections to weight things and work it out, you know, what it is you'd be interested in. Suddenly it was all about disconnected content. In what form? Well in terms of reels, short term videos. So suddenly they've got millions and millions of these videos and depending on what you search for, what you clicked on, but more what you clicked on and you've actually sat and watched that signal is what was then steering the algorithm. But the point that he made is the algorithm then has to change hour by hour, day by day because you might be in the mood for watching recipes at one moment and the next week you might be interested in fitness or whatever. And that's going to change as the day the week, the hours go on as well. So again, that's an example of meta using lots of non keyword signals to decide what shows up on your timeline as well.

[A]

So and it's a good example actually because I think for many years meta have been using non keyword signals. Right. Because it's not that well known as a search engine. I mean people do search on it, but the majority of their signals have always come from other stuff.

[B]

Convincing. What did you click on? Who did you connect to? All those kind of things as well. Now what I'd say is that there's a couple of things kind of here as well. What meta have come out and said, and this gets into the privacy side of things as well, is they are going to start later on this year. They've just changed their terms and conditions to say when you are talking to a kind of agent or a chat bot within meta, those conversations, what they learn from those conversations and the insights they can glean can then be used to target advertising and different things at you. Okay, now on first inspection to me that didn't seem any different to what Google are doing because Google are looking at the conversations you're having and then they're potentially going through and showing your content. So there's a discussion there about how much are we expecting these conversations to be private? Do we expect our AI chats to be anonymous? I mean, I know that if I'm within my paid for account for ChatGPT, that's locked down and it's private to that kind of platform, whereas if I'm searching Google, I'm expecting Google to be looking at that and using it for targeting ads at me. But you did point something out that's an area of potential risk and I hadn't got my head around this. So we should say this is our opinion. This is not something that's necessarily, yeah, what they're going to do with the data, but it's a scenario which could cause an issue.

[A]

I read the meta press release and we'll put a link to this in the show notes so you can read it too and form your own opinions. But my, I suppose the thing I was overridingly concerned about is okay, so it seems to be saying to me that if any interactions with like AI chats within the meta platform, like how far does that extend? You know, if I'm a brand that's implemented chatbot functionality within WhatsApp, for example, does that mean then that you know, any like purchasing signals, that chatbot that, you know, we've Worked with developers and had it built. Like, if it's given signals on what sort of products that people want to buy, does that mean that those signals are going to be used by my competitors and available to buy within the ad space? Because actually, that instantly has me as a marketer, has me worried. Oh, I don't know. I don't know how safe or secure that would be. I mean, I've been playing around with this actually. You can just have chats with llama within WhatsApp now. You know, you can just open up that box and just start chatting about stuff. But, you know, is the user explicitly aware of that? You know, how ethical is that? And I guess the answer actually comes down to how it's this is implemented. Like, this isn't going to be something they roll out until December and we'll have to take a look and see where that's at. But actually, if it's just like, is that okay if it's just in, somewhere hidden in the terms and conditions in small print? Like, these are all the kind of ethical questions I've got in my mind as a marketer, because actually as a, as an advertiser, if I'm making use of that data, that could be quite, quite sensitive data. And they did say within the press release that, you know, obviously sensitive data issues are going to be filtered for. And I applaud that. But then my mind's working overtime. How do you define sensitive data, actually? Because it's not a black and white thing necessarily from a consumer perspective, especially when you're aggregating millions of conversations up into a larger system. Like, where do you draw the line with that? How far does sensitive go?

[B]

Well, I'll give you. I've just had a business idea, right. So I'm going to take something like ChatGPT.

[A]

Yeah.

[B]

Do it for the API.

[A]

Yeah.

[B]

And I'm going to train it up on loads of stuff to do with digital marketing training. And then I'm going to go to all the digital marketing training businesses.

[A]

Yeah.

[B]

And I'm going to say join. Embed a chatbot for free.

[A]

Yeah.

[B]

Into your website. Okay. And if you do that, it will be an absolute expert on digital marketing. But I answer people's questions, it will find out what they're interested in for you. It's really going to be great for your website, really added benefit. Okay. And then what I'm going to do is I'm going to take all that data and I'm going to use it to target ads at people for me, because I know what they're interested in and all I've done is in my terms, conditions. I've just said you can have this for free, but in exchange we get to use the data for targeting. Right, so. And you could get away with that really easily.

[A]

Could you though? Is that explicit concept? Was that made clear? Did I opt into that? Well, if you signed up gray area.

[B]

If you signed up for a free tool in exchange for giving access to the data, I think that the brands have just given that away by accident. The people on the website would have had to go in and say, yes, I know that if I'm using this chatbot, but you know who's responsible for that? Is it the website that it's sitting on? Is it the platform that have provided the AI chat? So absolutely. There's some gray areas to get into this and I think that's the point is that I'm being a little bit facetious, but I do think that we're starting to. Because you implicit when you talk to somebody, you implicitly trust it a little bit more. Like you start having a much more human conversation with it. And I think that if people knew that everything they typed into ChatGPT or Google Gemini was going to become public, they'd probably be horrified. And that's not what we're saying is going to happen. But there is a debate here to be had about how anonymous should AI conversations be. And I think your example of I build an AI chatbot of some description to go into my platform.

[A]

Yeah.

[B]

And then whoever provided it could then use the data afterwards. Is anyone clear on that? So I think it's a really important one as well.

[A]

A lot of this is going to shake out in the final implementation of it. Right. That's the thing. I guess all I'm saying is actually for me it just had alarm bells ringing and I thought actually it's a really interesting like conversation that we could have and actually it's not something we've done a lot of like opening up the debate to all our listeners. We'd really love you to get involved with it. Absolutely. Like share your opinions on this. I'm really interested to know what people feel. I'm really interested to know if any of you have come across similar situations within your own implementations of systems and how you've overcome it. I think it's just an interesting area.

[B]

Big one that's coming up at the moment from a privacy point of view with AI is that people have bought a paid for version, they've locked it down within their organization. And maybe they're using Copilot, Maybe they're using ChatGPT, but suddenly it's got access to their OneDrive and they haven't got their settings set up right in their OneDrive, so people have got access to folders they probably weren't even aware they had access to. But also they haven't cleaned out their OneDrive folders in 15 years. And it's just a mess of files in there. And suddenly I can say, how much does Kieran earn? And somewhere on the thing, somewhere, it's got your salary listed and my salary list and that kind of stuff. And actually I can access that by accident. Now, this isn't AI errors, this is human errors that are in place already. But the AI has made that accessible. So there's lots of interesting things come out. What if I use AI agent mode and I give it the login to my LinkedIn and my Facebook and my X and I'm getting it to access those accounts and then someone hacks my ChatGPT account because I didn't set up two factor authentication, I didn't use a strong password, all those kind of things as well. So there's lots of interesting challenges. We would love to hear from you. What do you think about AI chatbot data being used to target ads? How have you got any concerns about security from an AI point of view at the moment? What is it that you think is acceptable and what's not acceptable of this as well?

[A]

I'd love to know what people think about how companies respond to changing their terms and conditions. That's the big thing, isn't it? I think, because actually you make a whole bunch of assumptions at the time and then suddenly an announcement's made or we're changing this as you might already be using it. You might, you know, how does that work for all the businesses that are already using this technology, you know, and they're actually already engaged with users. And again, all of it comes down to how it's actually rolled out and implemented. But I think there's some big questions there. Big questions.

[B]

So get over to LinkedIn. If you just search target Internet, you'll find it. If you just search for Daniel Rolls, R O W L E S, you'll find it on my platform as well. And then easy as well. Target.com forward/podcast and you'll find the link through to debate there. Get involved, lots of prizes to give away. We'll list all the prize there as well. Pick out some random people engaging and then we will feed back into another episode what the results of all that engagement have been. So thank you for listening to Digital Marketing Podcast and we look forward to you joining the big debate.

For more episodes resources to leave a review or to get in contact, go to targetinternet.com forward/podcast.