Are we about to get better AI data from Google?

Episode transcript

Amelia: Hi, I'm Amelia.

Paul: I'm Paul.

Russ: I'm Russ.

Tonicha: I'm Tonicha.

Amelia: And this is Fin the Week. Welcome back. We had a little week off. How's everyone doing?

Russ: Yeah, good, cheers.

Paul: Good, yep. Back from half-term week in Ibiza, which was nice.

Amelia: Not too shabby.

Paul: No, no. It's a nice way to escape the bad weather we've been having over here.

Amelia: Well, it's nice to have you back, and it's nice to be back for another episode. Tonicha, you've heroically stepped in, haven't you, because Pat's not well.

Tonicha: Yeah. I'm the new and improved Pat.

Amelia: He'd better watch his back if you do too good a job.

Tonicha: Yeah.

Amelia: But we have got a brand new episode, and we're continuing our theme of signal loss. So far we've spoken about a growing trend of data sources going quiet, but an upcoming regulatory change could give marketers the rare opportunity to get some new data. Paul, what's this all about?

Paul: Yeah, so it all comes from something from the Competition and Markets Authority in the UK. Back in October 2025, Google was designated as having strategic market status in general search services. As a result of that, the Competition and Markets Authority has been working on documenting what its concerns are and what follow-up actions they need to see from Google. I think it all really centres on this idea that Google is by far and away the market leader in general search online, and some would say they abuse that position.

So anyway, on the 3rd of June this year, the Competition and Markets Authority released a paper about Google's general search services. It was about imposing what they call conduct requirements to address any concerns they have, and the key concerns were these. First, that Google gives publishers insufficient choice over the use of their content in Google's generative AI tools. So it's this whole argument that you've probably been hearing bubbling away, of large publishers and companies who are having their content effectively stolen, they would say, by the likes of Google and repurposed in their AI tools. The second one was that there's a lack of transparency about the use of content in Google's generative AI tools. So if they are using it, it's hard to know how and when. And thirdly, that there's ineffective attribution. So when content is being used, it's not always being attributed to the source like it should be.

Off the back of that, the Competition and Markets Authority basically said there were three things Google needs to do. It needs to address a publisher's lack of sufficient choice over whether their content is included or not in generative AI, and they actually widened the scope from their original notes to say that would include training and grounding. So not just having your content included in a result, but also having your content used as a training source for their AI tools. They wanted to see Google address a lack of transparency over how content is used, which basically means they want Google to provide better data and metrics, and that's what we're going to be diving into today, really. And then thirdly, they want effective attribution. They want to see Google taking reasonable steps to ensure that content is attributed clearly and accurately.

Google has already addressed one of these. They've fired up Google Search Console, their tool that lets you manage your site in search, and added a toggle that lets you opt out of their AI features. We've actually spoken about that on the podcast before, and how it's a bit of a blunt instrument, because you can basically opt an entire site out, and that would mean you're not featured at all. It probably needs some refinement, because I'd imagine some sites don't want to be part of the training but maybe do want to be part of search in general.

But the big thing is what's happening next, which is that the next milestone is on the 3rd of December. Google has to comply with the bulk of this, and one of the key things, to quote from the gov.uk website, is that they must provide publishers with clear and detailed metrics on user engagement with their search content in search generative AI features. Which is interesting for the theme of our podcast this quarter, which is about signal loss and how data sources are disappearing. This actually sounds like an opportunity for us to gain new insight, because they're basically saying to Google, you need to start telling publishers when you're using their content and when people are seeing it and clicking on it. At the moment, Google Search Console only gives us impressions on pages. So you can see when a page is shown in Google's AI features, but you can't see how frequently it's clicked or anything like that. And this sounds like it could be an opportunity to bridge that gap.

I think the first thing I was going to pose to everybody is: is it fair for marketers to expect that data? Should Google have to provide this stuff? I think you could probably argue either way.

Russ: Just to be clear, when it mentions generative AI, what kind of format are they referring to? Are they specifically talking about Gemini search and how content is summarised by Gemini? Or is it as widespread as, is it related to copy, or is it related to imagery as well? Let's say you were using an image generator, which is now within Gemini. Are they talking about if it was pulling on sources of imagery and using that to create new imagery, using what someone else has published as a foundation for that?

Paul: Yeah, I think this is primarily going to focus on the AI features within Google Search. So AI Overviews and AI Mode. Interestingly, I would imagine one of Google's defence strategies will be to almost treat Gemini as a completely separate thing. I mean, it kind of is. And I'd imagine that's one way they're going to wriggle out of it a bit.

Russ: Because it does feel like, I mean, we've referred to AI models as a bit of a Wild West at the moment. The FCA had their public meeting and they were discussing that there's no FCA regulation within these AI models as well. So it was quite interesting to hear your thoughts on how far that would go, but it sounds like this is specific to Google Search. They're going to keep Gemini potentially out of it, at least for now.

Paul: I genuinely don't know how the crossover works at Google in terms of what's classed as Gemini and what's classed as a search AI feature. Knowing the way things go, part of their response to this regulatory requirement will probably be to keep the two very separate.

Russ: Yeah, because I think over time they'll probably merge, which means these publishing guidelines and these laws are going to become harder and harder to enforce, just because of how powerful generative AI is and how difficult it is to control the output. It's food for thought.

Tonicha: I think, with regards to your question, Paul, about whether Google should provide the information, it's an interesting one. From a consumer perspective, a lot of people don't want their search information being given out in that way, so maybe from that angle, no. But from a marketer's perspective, I would say yeah, especially if it's just information on what people are searching. Because in reality, that's not that different to the information they're already giving out on tools like Search Console. The main difference is just that the source people are searching on is AI, rather than Google or Bing or what they would have historically been searching on beforehand. So it's not actually giving out any more information than what it was already giving out when people were using Google. From a consumer perspective, it's not really giving out any extra information to what was already happening. But I suppose because it's a new tool, they have to treat it as a new thing. And I think these days people are a lot more aware of what data about them is being provided to marketers, et cetera. So with people being more aware of that and more protective of what they're searching, you might find there's a bit more uproar about the fact that Google's going to have to give this information out now, from a consumer perspective.

Paul: Yeah, it's an interesting point, because a lot of this focuses on Google Search Console, and within that, like you say, you can already access all of the search queries that people are searching for in classic Google Search. You can see how many impressions and how many clicks you get from those. But with their new AI performance section, you can only see the pages which are returned in those results, and you can see the impressions but not the clicks. It's kind of mad, in a way, that they don't show more information, because they already do it for the other type of search. So where there's missing information, you can't help but start to wonder what the secret reason is for not showing it. Why would they withhold it? Because even Bing Webmaster Tools, Microsoft's equivalent, shows you the actual search queries. For their AI report, you can actually see what search queries you show up for. They don't show you the number of clicks, but they do show you the specific things people are searching for, and that's quite powerful information. Tonicha, you do a lot of the reporting on our side. We rely quite heavily on that already, don't we? And it's only a few months old, that sort of data.

Tonicha: Yeah.

Amelia: Can I ask what you use that data for? Excuse my ignorance, but what difference does that actually make in your work?

Paul: You'll probably have some thoughts, Tonicha, but from my perspective, it helps to tell you if your website and your brand are properly understood by the search tool. If you're showing up for a load of irrelevant things, you know something's going wrong. And then it also tells you if you're serving your audience appropriately. Are they searching for things that you can genuinely help with? And are you serving them content that responds to what they need to know? We frequently use that information to help us say, here's an area of our website that we need to improve upon, provide more context, and stuff like that.

Tonicha: Yeah, to echo Paul's point, it's looking at what your audience is searching, and then you can often use that to see if there are gaps in what's on your website. For some clients, we can have a look at that report and see exactly what their audience is searching for them on, and if they find they're getting a lot of search requests around something they maybe don't have an article or a recent bit of content about, that means we can help the client with a new bit of content that will address a lot of the search queries coming from the audience. So that's one way we use it. But it can also be used the other way, like Paul said, to see if actually your audience don't really understand your website or your service, because the wrong sorts of people are landing on the website, or they're not searching what you'd expect to land on the website. So it just gives you a deeper understanding of your audience, where they're coming from and what they're looking for. That's mainly what we use it for at the moment.

Russ: Also, from a user experience perspective, I think there's quite a bit of crossover here, because understanding what features might need to go on your website, or what people are looking for, leads into the content strategy. At the end of the day, a website is content. That's what really makes or breaks it. There's a saying I've been saying for over a decade, which is good design can't save bad content. You can have a beautifully designed website with bad content and it's completely unusable, but if you can combine good design with good content, then you have a good user experience. So all of these terms that we're seeing lead into the content strategy, and also the feature set, which naturally makes a good user experience on a website.

Amelia: So this is potentially exceptionally valuable information. But looking at it realistically, do we think we are going to get this extra data? And what is that going to look like?

Paul: Yeah, I mean, there were unattributed reports online that Google will be launching click reports as part of their plan to comply with this. Someone posted on LinkedIn saying, I heard people from Google saying this. So it was very much hearsay. And then Gary Illyes, who is an analyst at Google but very public-facing and quite well known in the industry, came out and responded to it. His direct quote was, we're super interested in giving more data, but we don't have anything to announce yet. So basically you've got this strange scenario where on the 3rd of December there's this Competition and Markets Authority deadline for them to comply, you've got some people online saying they are going to comply by providing this extra reporting layer, and then you've got official sources from Google being quite cagey. It's not like Gary Illyes came out and said, no, it's not happening. He said, that sounds like a great idea, but he didn't really qualify it in any way.

Amelia: Do you think it's because it's not there yet? Like, the functionality isn't there yet? Is it because they're trying to find a clever way of sharing that data in a way that works for them? Why is that, do you think?

Paul: Yeah, I don't know. This is where the vacuum of information creates conspiracy theories, in a way, because you think to yourself, why would they? They've got vast resources. They already show this data for other types of search. To me, it would be trivial for them to make this available. So there must be a reason.

And actually, in the next section I've got some notes about this. Alphabet, Google's parent company, has recently had its Q3 earnings call, and it said that AI Mode, this AI version of their search, has lower monetisation than traditional search. So I guess what they're saying is they make less money from it than they had hoped or would anticipate. And so you wonder, does that have something to do with their caginess about the data available, because it's not as rosy as they would like you to think? If you were a cynic, you would say what they've done is scraped all of the good content of the world, synthesised it into something that serves their purposes and keeps people on Google platforms, but doesn't send any meaningful traffic to the outside web. And they've said all along that it does send lots of traffic outside, but they haven't provided any data on that. This is that data. So I think there's that element to it.

Tonicha: I do wonder as well if it gives them a bit of a leg up from an advertising perspective, being the only ones that can actually access the data. Obviously Google are also their own advertisers, in a way. So I wonder if they maybe don't want to give that out because it gives them a leg up on everyone else that can't access the data and use it. And I think, like you say, Paul, monetisation is probably a big one. They're probably trying to figure out how they can get the most bang for their buck out of giving the data, rather than just giving it out without thinking about that first. Obviously all these things are still relatively new, so in the background they're probably working that out.

Paul: Yeah, exactly. Off the back of that Q3 earnings call where they talked about lower monetisation, the "therefore" from that was that they'll be scaling ads within AI Mode. So they'll be selling more ad inventory on that part of Google. And that's potentially interesting for a marketer, because it was that scenario which led to a lot of good data from classic search. When they launched Google Ads, or AdWords as it was back in the day, in order to sell ad inventory they had to provide data on search keywords and performance and clicks and click-through rates and stuff. That was the foundation of keyword research. And I wonder if this is a perfect storm, in a way, where they've got this regulator requirement, they've got low monetisation, they're going to be scaling ads. Is that going to see them launching a new type of tool into their ad system that lets you see click data and things?

Russ: Yeah, I think one of the challenges with AI is that these chat interfaces start off really clean. With my UI/UX head on, they start off really clean because it's a new way of interfacing with content. So what you find is it's normally just a search bar, and you've got history, and that's kind of it, because they're relatively new and people are still getting to grips with the learnability of this new way of speaking backwards and forwards to get the answers that you want. So advertising is naturally being left out of that, I think, to begin with. And we've been looking more at impressions, haven't we, on search? Because people aren't actually going through to websites, and I'm sure they're not clicking on sponsored ads, because they don't want to leave the chat window they have with the AI.

I do think, though, we're going to see much more advertising on these AI platforms. Not only Google, but all of the other models as well, including Claude and ChatGPT. They won't be able to include those advertisements in the response itself, because I think that would just be too jarring for everyone, but we'll probably see them popping up a lot more. Perhaps they're just figuring out at the moment what the next step is to monetise the AI features, for it to be as successful as traditional search. And perhaps that's why they're being a bit cautious in their next steps, because they won't want to put off advertisers, that's for sure, or anyone spending a lot of money on sponsored ads.

I do think, though, in the next couple of years we'll see traditional search and AI search just merge. I have this theory that we'll just stop talking about AI as a thing, because it will just be the way that technology works. At the moment, it's quite static, pulling content from websites in traditional search, and we've got this moment in time where those are quite quickly becoming legacy ways of finding content and getting results. Those will be phased out. Everything that would have been labelled as AI in the last couple of years will just be how we do things. And then I think we'll see a whole new suite of ways to monitor performance and look at all of these results. I think it could happen quite quickly. Maybe next year.

Tonicha: I'd agree. If I was to guess how it will all work in the future, I think it's not likely they would launch a whole separate way of advertising on these AI tools. It will be within the ecosystem that already exists of Google Ads and Google Search. And like Russ has said, I think people will start talking a lot less about AI search, and it'll just go back to being search, with AI as a part of that. I think already, talking about the two so separately doesn't make sense when you think that even if you type into Google now, you're still technically AI searching, because you get the AI Overview at the top. So I guess the main difference in my head at the moment would be people searching on Gemini or Claude versus using Google or Bing, but even then it's technically AI as well. So yeah, I would definitely agree that over time there'll be less chat around AI search and traditional search, and it'll just go back to being search, with AI as a part of that.

Paul: Just picking up on one of the points, that they don't want to put advertisers off. I wonder if part of the reason they're cautious about providing click data from their AI tools is that the number of clicks is meaningfully lower than traditional search, and therefore they don't want to scare potential advertisers off straight away. An advertiser might look at that and jump to the conclusion that advertising in Google's AI platform is useless because it won't send me any clicks. I wonder whether they're trying to shift the frame of reference, in a way, to say clicks aren't the thing to track anymore, track impressions and stuff like that. Because advertising in AI results is more like being on a billboard, as opposed to traditional search, which was like a directory where you would expect a direct response. So it's going to change the way people advertise via Google, I think.

Tonicha: Yeah, I think, like you say, pay-per-click won't be as much of a thing. It will be more about trying to get your content to be the one that's picked up in the overview, and whether or not you could pay for that in the future so that you're more likely to get picked up in that overview than other competitors in your space. I think that will probably be more what people are paying for. So impressions or citations will be what people are paying for, over actual clicks to their website, because that's just going to massively decrease. We've already seen that across the reporting for a lot of our clients: clicks to websites are decreasing, but you're seeing a lot more impressions or citations on AI search instead. People are getting the answers from your website, which is still the most important thing, but they're not actually having to click through to the website itself to get that response.

Amelia: So that's perhaps where things are heading. But what do we think is actually going to happen now? Do we think Google are going to comply?

Paul: I genuinely don't know. They have to comply, you would have thought. It's something they have to do. But will they use some sort of technicality to say, we already provide that data via this thing that isn't really an answer?

Amelia: And do you think the CMA will let them get away with that? How closely are they going to be on their backs with this?

Paul: It's all been taken very seriously. It's not a throwaway thing. They can't really opt out of this. I don't really know, though, what levers the CMA can pull to enforce it. I'd imagine Google will comply, but whether they comply in a way that's actually useful or not is probably the question.

Russ: Just looking at these three concerns and what Google could do about them. The choice over whether a publisher can have its content in generative AI, that would probably be hidden away through a dark pattern of settings somewhere, which is impossible to find, and there's a checkbox. You'd probably find it quite difficult to even put in place. Maybe they'll just do that as a bit of a checkbox, but whether people will actually know it even exists as an option is another thing. The people that do will probably have to be quite up with the news and what's going on, and really keep an eye on what the latest settings are. The attribution one, though, surely they can just have a little footnote attributing the source of the publisher. So perhaps that's one of the quickest wins to comply on their side.

Paul: Yeah. A lot of major tech companies over the last week or so have all agreed to a new standard that lets publishers opt in or out of AI training and things. And it's as simple as a line of text in a file that lets you opt in or out. So that element of it, I think, is under control. It's the performance data part which is interesting, because I don't know whether Google would argue that what they have in Search Console already is what has been asked for. But I think the regulator would probably say that's not enough. Certainly when you read the quotes from the people talking about it at the regulator, they seem to expect more depth in terms of the data that's available to publishers. So it's whether Google decides to argue that what they've provided is enough, or whether they do in fact add extra layers of information to it. Because right now, they do less than Microsoft does with Bing. Bing provides search queries, and even if Google provided that, from a marketer's perspective that would be really useful and interesting. And from a publisher's perspective, I think that would tell them a lot as well. So fingers crossed, by December we'll have that sort of data available.

Amelia: And do we think, in general, it's fair for marketers to expect this information from Google? Is it the least they can do?

Paul: Yeah, I always wonder this. What is it reasonable to expect? It's not personally identifiable information. It's not data that you could use to target an individual person. And it's data that's generated off the back of Google using a publisher's content. So I think, yes, it probably is fair to expect that data, because ultimately Google has benefited from someone else's work here. It's the least they could do to provide that stuff. That's my thoughts on it.

Tonicha: I think I'd agree. As long as it's nothing sensitive that you can target individuals by, I don't see why it would be unfair for marketers to expect it, because ultimately it's basically the same information that we used to get from Google Search Console, but for a new tool that people are using to search. I don't think we're asking for anything extra. It's probably still going to be a lot less than what was provided on Search Console. But I think it comes back to that point from before. From a consumer perspective, what people are worried about is, is what I'm typing into AI going to be attributed back to me? And I don't think that's ever really going to happen. So overall, I think it is fair for marketers to expect it, as long as it's within reason and people's information is still kept private.

Amelia: And Paul, before we wrap up this conversation, let's talk about the wider context. What can you tell us about that?

Paul: Yeah, so it all centres on that thing I mentioned before, that it looks like AI Mode in Google is harder to monetise, and therefore their answer to that is to scale up the ads they're running. And I think that could open the door for more data, because in order to sell those ads, advertisers are going to want to know what they're getting off the back of it.

Then there's some other stuff. A writer from the website 9to5Google said that Google is touting the fact that AI Mode and AI Overviews keep people on Google for three times as long as traditional search. That's another bit of context worth keeping in mind, because it feeds into this conspiracy theory that they're not giving you click data because the clicks just aren't there. If they provided a report that showed you clicks, it would highlight in stark reality how poorly that's performing. So putting the regulatory argument to one side, there are also commercial things that are going to drive change here, because I think Google is still trying to figure out how to monetise AI Mode. That's ultimately what they're worried about, isn't it? Let's be honest. They want to know how it helps them dominate the next era of search.

Amelia: So it'll be interesting to see what happens. It's the 3rd of December, that next milestone, right?

Paul: Yes, that's the date by which they have to comply. So keep your eyes peeled.

Amelia: I'm sure it's something we'll come back to after that date to see how we're getting on. But for now, shall we do a bit of Jargon Busters?

Paul: Yeah, we've had a good run of it, haven't we?

Russ: Yep, let's go for it.

Amelia: Yeah, we've done pretty well, actually. Was it Dead Cat Bounce we had last time? That was a new one for everyone.

Paul: Yeah.

Amelia: Just in case you've never caught this before, every week I give the guys an industry term from our list of jargon busters to see if they can explain the definition. And this week, I don't know if I'm going to say this right. I've looked up the many different ways to say it, and it's one of those that I think British and American can say differently. The term is EBITDA. Have you heard of this?

Paul: Oh, EBITDA.

Amelia: Yeah, EBITDA, that's it. Spelt E-B-I-T-D-A.

Tonicha: No idea.

Russ: Yeah, I've heard of it. Just from an accounting perspective, I think I've heard of it.

Amelia: And do you know what those letters stand for?

Russ: Paul might have to complete it for me, but is it earnings before tax?

Amelia: You've got an I first.

Paul: I think that's it. Interest.

Russ: Is it income taxes, or is it interest, tax?

Paul: Interest.

Amelia: And then you've got D, A.

Russ: Depreciation.

Paul: Is it?

Russ: Sounds about right.

Amelia: Yes, you did it.

Russ: It's used in accounting. It's basically what the business earns before you take all of these fees, essentially, off the table.

Amelia: Yeah, so the acronym is earnings before interest, taxes, depreciation and amortisation. I had never heard of this term.

Tonicha: No, I hadn't. Thank God you guys are on this podcast. I was just sat here going, I don't know.

Russ: I don't think we completed it very well, but I understood the gist.

Amelia: I mean, that was a bit of a team effort between you and Paul there. We got the gist.

Tonicha: I think that was good, yeah.

Amelia: And we've all learnt something.

Paul: It's a shame Pat's not here, because it's the type of one he would embark upon a monologue with.

Tonicha: Yeah, this is what happens when you get Pat's understudy.

Russ: Pat heads up the finances around here, so he would have answered that perfectly, probably.

Amelia: Maybe he'd have taken us through a little presentation about the history.

Russ: I'll mention that he should tune into this bit, especially if we're talking about him.

Amelia: Oh, absolutely. Come back soon, Pat. Thank you so much, everybody, for today's episode. Really interesting conversation, and as I said, I think this is probably something we will come back to in future episodes. But we will be back next week, so we'll catch you then. Thanks for listening.