Trust signals, publisher licensing and browsers built for agents

0:02Hello everyone. Welcome to our new episode of marketing to agents. And as always, I have Lalit with me. And we bring to you today a lot of updates and news from what's happening latest in the industry. Hey Lalit.

0:18Hey Ashwani. Good to be here as always.

0:23Fantastic, Lalit. All right. So, we have a full docket for this week, Lalit. And I think a lot of things are moving in this space now. So, why don't let we get started first with what's happening with Where is the judgment now on whether, you know, AI search is winning, AI search is losing. Last week we heard from Reddit CEO who was talking about that organic traffic is going down and it's not getting replaced by AI overview. This week we had businesses, like specific businesses, for example, Tripadvisor CEO. They said that they took a revenue hit because their SEO traffic dipped. And of course, AI overview is giving them importance, but it's not really covering up for the loss which they took on the SEO side. On the other hand of this argument, we have user-generated content sites like Yelp and Trustpilot who are actually seeing increase in their revenue. They are reporting higher and higher traffic and revenue. So, Lalit, where where do you think this segment is growing? Where does the segment going? Where is the usage falling in? Where will we eventually end up with? And then we'll of course look at it from how should a brand handle it

1:40From their side?

1:43Yeah. So, I think we So, we have to first establish that the the search behavior is fundamentally changing and it is significantly changed already, right? And there is still a lot of category specific nuance. Right? Categories where I think discovery is is happening on AI let's say the AI surface area versus let's say let's say transaction is happening on AI surface area versus let's say another category where the discovery is not that important, but let's say a lot of research also has to be done and there are the So, companies who deal in those categories, so this this category could be let's say software purchase, right? So, if you're basically evaluating software vendors, discovery is a very small piece of it, but since the decision is very risky, you want to spend a lot of time evaluating vendors and maybe Trustpilot gets the referral traffic rightfully so because people want to look at individual like individual reviews and maybe, you know, filter reviews by by the type of people who have reviewed. So, extra functionality which allows the user go deep into their own

2:59Research, all those websites and let's say platforms which enable that, they would certainly benefit from the referral traffic which AI engines are able to provide. Categories where it just ends on a well-researched answer and maybe the decision happens offline or on another surface area, let's say typically travel, right? Over there, I think, you know, the search surface area is somewhere the end of the gate, right? The end of the way the end of the tunnel is sometimes, right? Over there, I think the business model has to be rethought and as already we are looking at publishers getting still a little more innovative with respect to let's say licensing deals with AI engines and different business models. I think we have to sort of accept this reality and sort of business leaders have to just, you know, innovate around it.

3:55Absolutely. In fact, as you were talking about this, one more idea struck my mind. So, for example, now since we specifically are talking about we specifically are talking about TripAdvisor. So, I'm also assuming that with this agentic commerce increasing, now think about a a TripAdvisor or a Booking.com. So, when a user was earlier going to these websites and booking a booking a flight or booking a hotel, making a reservation, these sites would also have an opportunity to upsell to that customer. They would then get their, let's say, a cab booking as well. They would get some other bookings as well, and they will try to, you know, sell them experiences. Now, with the agentic commerce happening, if the user is just telling my if I'm telling my agent just to book me a flight, I think the business the publisher or the brand also loses the ability to upsell at least in that moment. Maybe there will be an opportunity to upsell later as well, but I think just the is there a drop in the ticket size as well is would be a question that I would it would be interesting to understand. Typically, when I go to Amazon or anywhere, when I'm buying one thing, usually Amazon shows me 10 other things to buy, and some of those are relevant as well, and I end up shopping more. And this must happen a lot more in different categories. If agentic commerce is taking that over, would the

5:13Average ticket size drop? Do you think that could also be a phenomenon we could start seeing now?

5:20Yeah, so average ticket size might have some, let's say, affect whether most likely it's going to go down. But, if you look at many of these aggregator platforms, they have they've always had a significant advertising revenue stream. Right? So, if you look at Amazon or Booking.com, Amazon for sure has some sort of an advertising revenue where vendors or sellers can basically showcase their listing or, you know, pay for getting it displayed as an ad in the in the category list, right? So, that revenue for sure is going down because what's happening with the agentic commerce is that the decision and right right from discovery to decision to let's say even final purchase, that can actually happen in the entire agentic experience, right? With the adoption of UCP. So, I I'm I'm actually a strong believer that the the agentic era would make life very difficult for the aggregator platforms. and they can only survive if their let's say if they have a strong value proposition over and above aggregation, right? So, for example, if travel portals, they offer benefits like free cancellation or let's say assurance on trip, etc., right? Packaged in their experience, then there is still a reason for

6:38The AI agent to also prefer booking via the the aggregator platform. otherwise, what would happen is as the penetration of UCP and let's say, you know, better websites happen in every category, right? For example, Shopify as as an example compared with Amazon. Shopify has enabled each of their stores with UCP, which is Universal Commerce Protocol, right? And today a Hermes agent or a or a Open Travel agent can can basically discover and transact with a Shopify enabled store without [snorts] even needing any discovery. So, bypassing a meta discovery that happens on meta or discovery that happens on Amazon or any other player. So, unless the aggregator platforms offer a unique value proposition, which is compelling enough for the transaction to continue to go through them, the the leaf nodes, the leaf nodes on these marketplaces, they would stand to benefit from it.

7:41Fantastic point, Lalit. And yeah, I completely agree. So, if if a brand is there, let's say if I'm a brand and I'm not just an aggregator as well, I think there is value in me in of course getting as much as much of my number one is get as much testimonial, get as much customer feedback, reviews as possible. Of course, not just on G2 sites like G2 or Trustpilot, even maybe highlighting it on the brand sites. That would get picked up more often and would rate highly on would go a long way in in ranking you better on the AI searches. And at the same time, if as Lalit, you were agreed that if you are if you're an aggregator, then I think something to add value to the transaction would keep you relevant. If you're not adding a value to the transaction, then I think it will transaction eventually go to the end node, and that's where things will happen.

8:38Yeah. And and if you look at Ashwani, what are the trust signals that the the the individual seller can can sort of muster to bring together? On Amazon, they they definitely have the the benefit of the aggregated reviews that they've collected over time. On Shopify, no matter what you do, I think if you're just collecting reviews on Shopify store and then displaying it, it doesn't give give that trustworthy signal, right? so, platforms like Trustpilot, which enable individual seller, individual merchant, or individual vendor to to basically give an impression of trust with the review collections, that these platforms stand to benefit from this whole agent e-commerce. And that's why we are saying that Trustpilot is actually reporting significant growth in revenue year over year. Because if I was a Shopify merchant, right? And much of my reviews, let's if I'm a seller and I I maintain a store as well as I sell on Amazon, My Amazon reviews are there, but probably my Shopify store is getting discovered. And when the agent is evaluating whether my product is worth trusting or not or my brand is travel worth trusting or not, I really need third-party signals. And today Amazon reviews cannot be taken out. That signal is walled let's say walled garden under the Amazon

9:56Ecosystem. So, third-party let's say trust signaling platforms like Trustpilot or G2 in case of let's say software category. Trustpilot happens to benefit because it is agnostic to any category. It is sort of horizontal. I've seen right from services players to restaurants using Trustpilot. Right? So, great for them and because they're the you know the era that we're entering into, people would really need more and more services from Trustpilot just to signal trust.

10:28Agree with I think this completely aligns with the earning with the earning statements that they made. They're seeing a lot of enterprises a huge demand from enterprises as well to showcase their but it's just to just to follow up question on that. So, the demand for Trustpilot would be from brands from smaller brands to be able to to be able to integrate Trustpilot on their or is it going to help the the agents to be able to assess that my brand is is trust well trusted. So, is that for the end user or is that going to help the decision-making for the agent?

11:08So, the way it works Ashwani is that you can collect reviews on Trustpilot. But for you to display the Trustpilot rating on your website, you need to have a paid relationship with Trustpilot. Right? So, when agents are you know, sort of accessing your catalog, they they would appreciate if there is a third-party quality signal which is available in the API or the website, right? And that's where Trustpilot paid relationship comes in for the vendor.

11:39So, it's both then. As a user also, I get more trust when I come to the site, and at the same time, for the agents also, for the AI agents also, it will be it'll be a additionally good input for them to consider the brand as more trustworthy.

11:54Yeah.

11:54Okay. Fantastic, Althea. So, I think this could be a good direction for brands to get affiliations from or build more trust signals either through UGC or through your customer reviews. I think all of that would help both in the consumer in the in the user building the trust with the users as well as with the agents getting the right signals.

12:17Yeah. Perfect. Moving forward, Althea, now you did touch upon a little bit about publishers. So, we did see a lot of news coming out of publishers this week. Starting with There are publishers who are now actively going and licensing with the LLM players, so that their content can be used for learning as well at the same time they get rightly compensated for that. So, it's a license deal. Now, again, yet to be seen whether these are exclusive or not, but that's a great, I think, win for the publishers where at least they're getting licensing deals with AI companies. AI companies are going to pay for it, number one. But, of course, there is also another side of argument where there are some publishers or maybe the smaller ones will always get left out. So, how does that happen then? Then what happens to them? And we saw an interesting news coming out of New York Times or New York Post. And what they've done is they have built a full AI recommendation and AI interface to be able to engage with the audience coming on their website. So, there are all these modes of engagement going on with with the publishers. Where do you think I think this finally lands at? Again, completely different. We've talked about how LLMs how AI

13:36Companies have have been up against publishers and publishers have not been very friendly, but this seems to be changing a little bit now.

13:46Yeah, Ashwani. I think So, see, totally depends upon the the publishers' individual pull, right? So, if a certain publisher has had significant [snorts] and dominant position, right? in terms of their quality of content and they they've always enjoyed direct traffic, right? Because of the the quality of content and the experience that they provide, then I think the best economic decision would be to sort of stick to let's say with you know, not to make their content available on AI engines and maybe continue with their own marketing and drive paid subscriptions. That would be the best and highest margin game for them, right? the second best thing would be to basically look at licensing their content exclusively with let's say one or two players. And then go sue everybody in every other engine which is sort of you know, accessing the content without licensing. So, this would again be another let's say you know, litigation led revenue. and the you you sort of raised the concern of let's say the small time publisher or the long tail of publishers. For example, let's say someone is an expert in Japanese recipes, right? Or someone who is let's say has in-depth content on

15:04Let's say you know, species who live in oceans. So, I think there has been some efforts in the open web to allow indexing as well as licensing for these publishers. So, Cloudflare Cloudflare has launched something where you can basically start with the blanket blocking of all AI crawlers and all AI agents, and then a programmatic way to sort of collect the payment. Kind of like a pay per crawl, right? And agents which are ready to pay per crawl, they can get access to the content. And there's a win-win situation there, right? There's another standard called RSL, which is really simple licensing. stands for RSL. And it is it is again kind of similar to Cloudflare, but slightly more open. There's no one vendor which is sort of jam Let's say Cloudflare is is the Whatever they're doing is very specific to Cloudflare, but RSL is an open standard. And any publisher can basically sort of implement an RSL-compliant load balancer or an you know, like an Apache plugin or an HTTP server plugin. And that would then ensure that their content is only accessible by licensed

16:22Crawlers. And it it also makes it super easy for acquiring a license. By sort of you know, offering a way to collect payments through smart contracts. So, I do believe that, you know, licensing the content is is the way to go if your content is [snorts] is really worthy of that. A large number of publishers, they are In fact, publishers is also a very loose term, right? Because almost everybody now is creating content, right? Now, what is their intention of creating content? Are they creating content for marketing? Are they creating content for the value of the content, right? is generating content is their core business, right? So, people who are generating content with a lot of investments in doing research and you know, digging out real insights, doing real research, and bringing real knowledge out in form of content. For them, I think licensing and continuing to have a paid relationship with their subscribers is the best way forward. Not doing licensing and maybe keeping their content open would significantly dilute the value of what they've done because people might get access to their hard-earned information right in their answers and no traffic would come to them, right? So and I do see there is an opportunity in

17:39Let's say aggregated aggregation play where long tail you know, there could be a market place. There are a few market places which are funded already, right? The who basically aggregate long tail players, long tail publishers. But they're sort of offering a cookie cutter solution to long tail publishers. But publishing as we know, I think it's it's nuanced, right? You might want to even in addition publishing also, publishing some books are way cheaper in let's say developing countries versus when you buy that book in a developed country, so people might want to have publishers might want to might want to have control on geography level licensing or demographic level licensing, right? So the cookie cutter market places, they do half the job, but I think as the open standards get more penetration we will see a really rich ecosystem emerging again, right? The web that we are so used to in the last 25 years, right? The the whole advertising led you know, business model of the web, it sort of flourished the content, right? I distinctly remember when Blogspot came about, right? I my first blog was on Blogspot. And it was obviously a pleasure to sort of publish your thoughts to get the your thoughts in a well-structured manner out and your network would be able to read

18:56It. If no nobody reads it, documentation itself was a sort of like a journaling and a documentation was itself value for someone who's publishing. But that whole activity of publishing was I would say turbo boosted with AdSense, Google AdSense, right? And the the way it incentivized let's say publishers to write content, that sort of I mean in fact we've all seen that entire evolution where there's almost no category out there which does not have content presence on the internet, right? So, that vibrancy of ecosystem, I think that is threatened with the AI search. And I think licensing is is again like a one of the one of the I would say a win-win way out.

19:45Absolutely. I think I was going to come to that exactly. So, I think the this whole publisher leaving aside the bigger ones, but I I think the overall democratizing of publishing or posting online started with AdSense when Google started you know when the blog posts started and then Google started allowing people to monetize that. But now since now the discovery will become a challenge now. So, you're right a lot of and a lot of people are going to go a different route. I think if the content is really good and insightful, subscription is a way to go. And and this I have seen a ton of new Substack subscriptions getting launched every week. And if you just follow Twitter, like every segment at least Twitter I think caters to more content which is more specific to certain segments. So, we've seen a lot of Substack coming out of Substack subscriptions coming out of that. And there are people like again, some like very young audiences who've been able to build a million-dollar worth subscription businesses alone just by just on Substack and in very short amount of time. So, I think that if the content is and again, again another thing which I think has happened is different content will now follow different paths. So, for example,

21:03If the content is more around lifestyle, if the content is more around fashion, I think the right place for them probably would be now YouTube and and creating more video or photo content. And I think that would have a different level of There are different monetization methods there. So, but yeah, this this goes into more creator economy. But like coming back to publishers, I think absolutely spot on that. I think it will be a mix of doing a licensing deal and then being able to s- w- being able to wall your content from other non-licensed partners. And then and then be able to and then be able to get enough traffic discovery or or money out of that licensing agreement.

21:50Yep. And I think it's it also opens the door for newer business models on the internet, right? or in fact newer new new type of businesses. So, there's a podcast called Startup Idea podcast, and I think the host there, he talks about an idea category called data refineries, right? Just like we refine oil, oil is present or minerals are present, right, in the nature, but we have to refine it so that the the the quality density of the the mineral that we want or the oil that we want, right, is high. And then you can put to use, right? So, quite like that, I think there are so many information categories where the data is very well like very well present on the internet, but extremely scattered, right? whether in form of reviews or some publication here and there, right? So, someone can actually build really good business for their category of information by building a data refinery, where they use agents to regularly collect information, put this in well-structured format, and provide a paid access or licensing access to to LLMs who want to use it. This could be, let's say, you know, venture funding data or analysis on tax data like in in in in in India, right? If someone wants to do an analysis because as

23:07It is, I think the data is present unstructured and it is too much hard work for like someone is trying to ask a question about let's say Indian Indian taxpayer division, right? Like what are the categories of Indian let's say taxpayers. So, that data is there but it but someone has to refine it, make it available, and then this answer could have a very like it let's say ready-made data to to sort of give it to the user.

23:40Very interesting health, yeah. I that does seem to be an idea especially with the

23:45[clears throat]

23:46With content, with data, with content the explosion of content which will happen. I think the explosion of content and data making it making sense of what's what's relevant, what's not relevant, what's clean, what's not really original. I think it'll be it'll be an important task in general as well. Just to keep sanity on what's what's getting published and what's there on the web. Perfect. Moving on to our next item on the docket. AI browser. So, this has been an intriguing topic that and I somehow have not still jumped onto that bandwagon. Maybe I'm a I'm not an early adopter on this side. But Perplexity was the first one to come up with an AI browser comment and it had it had great reviews. People started using it and and now we are seeing two different newses news coming out this week around AI browsers. The number one is of course Perplexity making their offering completely free and trying to get more and more user acceptance and then thereby trying to get and on the other side get more publisher deals going so that they can have more access to information. Whereas on the other side we saw OpenAI

25:04Shutting down their, Atlas browser, which was a similar attempt. So, let I know you've been a very, active user of, Perplexity comment and just having just being aware of how the AI browsers have have grown over the past year or so. Would love to hear your thoughts on how do you see this change happening and why one company shutting it down whereas the other is trying to, you know, win the market with it?

25:32Yep. So, I think OpenAI hasn't shut it, like properly. Like when I say properly as in like they've not really given up on browser. They've just, changed the perspective. So, it is not available as a standalone browser, but it is available as a in-app browser for their Codex agent, right? Or their ChatGPT work agent. And, my personal opinion is that that is the right thing to do, right? Because the see browser is a is a fantastic work surface area, right? A lot of work gets done on browser, research gets done on browser, our let's say travel planning gets done on browser. In fact, in the modern web, almost everything has to have a web presence and then, browsers are way how we are the gateway to the web and then much of the transactions happen through the browser interface only, right? And as our work gets delegated to an intermediary, right? If today, let's say, if someone is doing all of their work by themselves, whether it's researching, booking a ticket, or, let's say, writing a document, producing code, producing documents for their work, pitching to clients, everything happens, let's say, with them as the primary

26:49Agent, then just having another browser would only be a delta in productivity. Like maybe a delta increase in productivity, right? Because they still are dependent on the human to use the browser. The moment we introduce another intermediary which I can delegate much of my work to, right? Which is happening through a Permeases or through let's say let's say Codex, right? I think this new user on the internet would require a special browser, right? Which does not block, which which is a lot more token efficient, which has let's say MCP interface or programmatic interfaces to get stuff done. In fact, one news which is even more significant than the two things that you talked about in browser world is Cloudflare launching their own agent browser. They have launched something called as Kite, which is a specific browser built for agents, right? And agent as a user on the internet is geared for efficiency, right? They can do like thousands of queries in a second. They can quickly consume so many pages, perform so many actions, right? So the traditional browser which was primarily built for the user may not be the perfect one for the agent, right? It does not truly unlock the agent when it comes to productivity.

28:08So any any half-assed effort where you still want users to be using it and also agents to be using it, that would only be a marginal 10% improvement in productivity. What OpenAI is doing is they probably are taking the Cloudflare route where the in-app version of their browser is not meant to be used by the user, right? Users can obviously troubleshoot a let's say session where agent is stuck. it allows them to troubleshoot it, but primarily it is the Codex agent which is going to use it. So I'm a believer that browsers have to be built or maybe rebuilt for agent as a primary user. And Comet probably has sort of missed that bus and ChatGPT is probably taking a good detour and going into the right lane.

28:57Understood. So, then what is the point of having Comet browser anyhow? Then why would I not just run, let's say Chrome, and then have the GPT have the Cloud extension in it and just use that for my browser mode or things to be done on the browser. Like, what additional value does Perplexity Comet provide today?

29:19So, Perplexity Comet was the first one to come up come, you know, like bring about this usage like the UX pattern of the agent taking over a tab and then you can monitor. The UX is pretty nice where the the color scheme of the tab also changes so that it distinguishes, you know, itself from a tab which is not being overtaken by when agent, right? So, they did some really pioneering work on the the browser the agentic browser. And the inbuilt assistant that Comet has, that I think is by far one of the best integrated experience when when you're trying to, let's say, summarize a page or summarize a YouTube video, right? It was much better earlier. Now they have sort of tried to optimize it more towards computer, right? They're just trying to push the computer product in the Comet browser itself. So, the experience is now different. But, Comet is in some sense it's their foot in the door for productivity knowledge work, right? The real product that Perplexity is wanting to push is the computer product. And maybe in a month or month or two, maybe in a quarter at max, we will start to see Perplexity computer will not be a separate app, but it will be a like the Comet will be the computer, right? So, you don't have to necessarily install a Comet app, right?

30:37And knowledge work productivity, if you someone really has to compete in category, they have to have a desktop installable product. And a browser is a is a good I would say packaging for high productivity product. People may not use Comet for the browser. People might end up using the computer which comes coupled with the the Comet. Yeah, so Comet got the limelight because it was the first one. and then everybody else copied the pattern of either having a Chrome extension or basically, you know, taking over the agent taking over the the entire the tab to get the work done. But still, you know, the way it works is either it is using a like a Playwright kind of a framework to understand the content on the page and then the LLM reasoning over what should be the next action on the page. If you want to do just a simple thing of let's say filing an expense report, right? So, it is basically screen-by-screen, frame-by-frame analysis or a page HTML analysis, right? And then the the user agent has to then hijack the JavaScript to sort of move the cursor and get the click done. It is not very token efficient, not the best way. Right? So, what we eventually need is a age a browser which is specifically

31:54Built for the agent where agent gets I would say direct access to controls and actions that they want to take rather than, you know, agent behaving as a user. Like a human user. So, agent should not be trying to behave or emulate a human user. They should remain as an agent and the browser should adapt to the agent capabilities.

32:16Interestingly, so you you believe that the browser has to evolve to then accommodate agents or like the browser it's a different browser which then works according to how agents operate. Then that would be that we in your mind is the right approach to solving this problem.

32:33So, Ashwani, I think so long term I I future is that the UI will be dynamic and humans would not be looking at a browser. What humans will be looking at is something that the agent outputs and that will be totally dynamic, right? And today we have like first versions of the agent GUI, right? AGUI is a is an open standard which Google has championed where there's no fixed set of markup depending upon what the output agent wants to give, the markup is generated dynamically. If it is a graph, then it will be a graph. If it is a list, then it's going to be a list, right? So, in a year or two, right? we will mostly be on our voice interface with the agents either through mobile or through our airpods or any new device, right? And we will be seeing the output dynamically on our screens or on our glasses which which the agent has produced and agent is then using the agent native browsers, right? So, there might still be some very archaic application which might require, let's say, a human to open up a browser. So, that compatibility has to be maintained. But, significant portion of our work will be done through voice command, voice orchestration, and looking at, you know, the dynamic AGUI.

33:46Perfect. Yeah, I think that that's a vision and something strong that that will empower the agents to actually run a lot of things that we are already envisaging and seeing happening on the web today. Fantastic. little bit that I think we have covered most of the docket for for our this week. would love to hear any any remarks, anything that you would like to cover, anything around AI specifically or anything that you've heard, seen in your in last week around AI search that you that's something could be captured today.

34:29I think we have covered enough. I I think I'm seeing AO becoming e- more and more significant in conversations. so there was a there was an agency which did a survey with CMOs in India and close to 98% of CMOs they're actively looking for an AO solution, right? Because this is now a board level conversation. Like what is our AO let's say visibility? And in fact a couple of weeks ago I came across a ad on LinkedIn which was surprisingly an ad for G2, right? G2 which is a like a software review website where someone can go and review their software tools. And in that ad almost every G2 investor, right? Almost every G2 investor was on the cap table, they talked about how AO is important and how G2 helps you crack AO. So I think that was something good to watch. And yeah, we at Airpr I think we are we are we are seeing a surge in pipeline. We are seeing, you know, happiness and satisfaction in customers. So pretty bullish about AO and yeah, looking forward to this AI AI-enabled information access world.

35:44Fantastic as I am yeah, that's a great great note to end today's episode as well. So yeah, team listeners please let us know if you have any questions, if you have any feedback about this episode. And yeah, as Lalit mentioned AI search AI monitoring your AI mentions is becoming more and more critical, business critical. Every boardroom conversation has this covered now. So yeah, please feel free to check out airpr.ai around this and do let us know if you want to get more in touch and details about it.

36:20Thank you.

36:21Thanks Lalit.

36:22Thanks a lot. Thank you.

36:24You.