Cadmore media player playing video How Researchers Are Embracing AI Search: A Growth Opportunity for Journal Publishers
Video Player
Accessibility
Using forms mode, you can utilize keyboard shortcuts to make usability more efficient.
Space bar will play and pause the video. Pressing escape will close all displays and pressing it again will jump to the end of the player.
You can press A to open the Keyboard Shortcut Dialog which also provides an audio description of the player and all of its features.
Language
1x
Play Speed
Adjust Volume
Transcript
Search
Result Count
of
Click to jump to result
Search
Re-center
Segments
Result Count
of
Click to jump to result
Search
Name:
How Researchers Are Embracing AI Search: A Growth Opportunity for Journal Publishers
Description:
How Researchers Are Embracing AI Search: A Growth Opportunity for Journal Publishers
Thumbnail URL:
https://cadmoremediastorage.blob.core.windows.net/1ba18afb-3f19-452d-a18c-23f135d3b633/videoscrubberimages/Scrubber_1.jpg
Duration:
T00H34M06S
Embed URL:
https://stream.cadmore.media/player/1ba18afb-3f19-452d-a18c-23f135d3b633
Content URL:
https://cadmoreoriginalmedia.blob.core.windows.net/1ba18afb-3f19-452d-a18c-23f135d3b633/SSP2025 5-28 1415 - Industry Breakout - Consensus.mp4?sv=2019-02-02&sr=c&sig=v8I9o%2FHsFWPkZ61vfjHHThfA3j3nd%2B12CYilh4EMHKQ%3D&st=2025-07-01T09%3A14%3A40Z&se=2025-07-01T11%3A19%3A40Z&sp=r
Upload Date:
2025-06-06T00:00:00.0000000
Transcript:
Language: EN.
Segment:0 .
Thank you, everybody, for taking time out of your lovely, beautiful afternoon in Baltimore and joining us here today. I'm Eric Olsen. I'm one of the founders of consensus. Throughout today I'll definitely talk a little bit more what consensus is. But for those who don't just in a sentence, consensus is an AI native Search engine for academic research.
The purpose of today's talk is to talk about the very, very fast changing landscape of scholarly search. If you haven't heard. AI is changing things, and we're going to talk about how we're looking at the landscape and what we think it means for everybody in this room, including ourselves.
I'm going to start with a quote. I think everybody knows we are in the middle of a change. And I think at least we think on this side of the table that the truth is that this change is just in the earliest of innings and it's going to be a constant change, just like everything always has in every domain. But it's going to be a constant change for the next 2, 3, 5, 6, 7, 8, 9, 10 years.
And we think the change in this space is perfectly captured by these two graphs. On the left, you see exponential growth, and on the right, you see about a 30% decrease. We have purposely obfuscated the titles of these graphs throughout the course of this presentation. I'll share what these graphs are, why we think they're really important, and what we think it means for us on this side of the table.
And lots of you guys in this room here today to just introduce ourselves. Again, I'm Eric. I am the co-founder and CEO of consensus. Sitting at the table here, we have our two other consensus team members who are with us all week. Here at SSP. We have Christian Salem, my co-founder and chief product officer.
And then at the end of the table, we have JL Needham, our partnerships lead, who's the former partnerships lead for Google Scholar and Google Books. Is a quick, super, super quick background story of consensus, I met Christian about 13 years ago. We actually were members of a college football team together at Northwestern, and we are the children of academics and teachers and scientists, and we had always felt like outsiders to this world, but had a deep passion for wanting to contribute to science and wanting to see science diffused more into the public and made more accessible.
So, chart number one. Unsurprisingly, it is related to AI. So what that chart shows is the estimated weekly active users of ChatGPT. That shows if it is estimated that there are 800 million weekly active users today of ChatGPT, g.p.d. That means that 1/10 of the world uses ChatGPT every single week. Now, in our world of consumer tech, we've seen lots and lots of exponential growth stories, but nobody has ever seen anything quite like this.
It surpasses everything that comes before it. If you look at that first notch right there, that is going from 1 to 50 million in about four months, I believe, which is the fastest that a product has ever done that. And it looks like a blip on the radar of how it has continued to progress today. And then on the right side of the screen, we see what that growth has done in the markets that we all care about.
The adoption is incredibly rapid where we're seeing the most in younger and earlier career students, and then in the Fortune 500, in an industry where people are also rapidly adopting these tools. As a consumer facing startup, we are always, always, always, always trying to iterate and improve on our product. So what that forces us to do is talk to lots and lots and lots of people who are using our product and who are the earliest of adopters of these tools.
So we have a really good pulse of who are these people that are adopting these tools early, and specifically, who are these people who are adopting them in the spaces in the markets that we all care about. So we're going to go over the who, what. And then it is obfuscated by my subtitles. But the why, it's very simple. So I'll be able to explain it. The who is kind of what we just showed on the last slide.
It is mostly think of it as earlier career researchers, whether that means people who are still students or people who are just beginning their academic research career journey. Then the other place we really see this technology start to disseminate is in industry, whether those in the Biosciences, whether that's in consulting world or even clinicians in the doctor's office. And then in what are people in our space using AI for.
If you think about research, at the Super highest level of abstraction is a three step process of learning, creating and sharing. What you really see is people using it for the first step in that process, in a little bit of the early stages of that second step of the process. So the learning and creating. We see people using AI in conversational and fluid, flexible search and discovery and learning of academic content.
We see them using AI models and products to summarize, ingest, consume, and analyze that content. Then on the latter two bullets, you see people using it for the early stages exploratory kind of kicking off of these deliverable creation writing creative tasks. And there's multiple things in each one of these buckets. But the y at the bottom, there's only one thing you hear if you ask these users over and over again, why are you using these products and why are you using them for academic work.
The answer is it saves them time. We're going to talk in this presentation about some of the threats and the downsides that this can all bring. And I'm sure all week in these next two or three days, you're going to hear a lot about the negatives and the threats. And those are all 100% real. And part of this talk is to talk about how to mitigate these downsides.
But it is important to remember that there are real positives of this. We all got into this space because we care deeply about science and believe in the power of knowledge, and believe that knowledge increasing improves society. And if these tools can speed up the day to day of the people who are involved in this knowledge creation loop, that absolutely has the potential to be incredibly net positive for society and something that we should all be really excited about.
Chart number two from that very first slide is the search interest traffic for Google Scholar. So those troughs are the summer. But the net decrease is about 30% if you map it back to 2022 around when AI products started to disseminate. And when we think about this chart and you think back to what I just showed of who are these users of AI products and research.
It's a lot of those early career researchers and students. And what we're thinking a lot about a consensus is those students, those people who are just beginning their careers, they may never learn the old habits of people before them, and they may never be tried and true Google Scholar users like the generations before them. And I think when you think about that means that graph might only just be starting.
And this is just the early innings of starting to see this downtrend. So that brings us to the next question is if that is true, what we just said, what we just showed is that there's going to be a continued decrease of using legacy traditional academic discovery tools for scholarly search.
What are the downsides. What does this mean specifically if they're going away from those tools. And like what we're seeing today, they're starting to use general purpose AI chatbots to do their scholarly discovery. We think that two core things happen that are problematic when people stop using research tools to do research.
The first is quite obvious sourcing, attribution and citations are no longer first class citizens of the products that they are using. What I mean by this is there are some high percentage of requests that go to these products that are just handled by sending text back to the user, that is just text that is generated from internal model knowledge. Model knowledge that might have learned from some of your content and there are no citations to be found.
The second part that can happen. And why they're an afterthought, is even if they are shown in the product and they're used in the response, they are not displayed prominently in the UX. They're not designed to be search engines. They're not search engines first, and they are very rarely ever interacted with, digested, and consumed by the user.
Even if these models do use a search function to bring in citations to their responses. The second big bucket of thing that happens is when those citations are used. We're typically seen to be lower quality content being used. The reason why we think this is happening is we love to crap on Google sometimes, and the people in the AI world do, but they're the greatest search company of all time.
And Google Scholar is a search algorithm that was tuned over decades of people using it, and that is trained to try to surface the best and highest quality content. These products are not Search Products. They're not search companies, and they're not learning from researchers, using them to try to surface the best possible content. So even when citations are used, not only are they an afterthought in the UX, but many times they are lower quality content than we're traditionally used to seeing in dedicated research tools.
And then yeah, the can't really see it with the titles. But the downstream effects of this, we think are honestly pretty scary. We think that number one, all of that amazing things that could happen of science progressing, that is at risk and lower quality science and research happens. And then number 2, for us and for everybody who produces content, that is a real risk to our respective businesses.
So the logical next question to us is what do we do about it. How do we reduce some of these downsides that can come from people starting to use not research tools for research. We are biased of what the solution or what the path to figuring out the solution should be. But as a consumer facing technology company, we always think it comes back to the user's needs and start with the user of what are they looking for so that we can get to the solution to our problems.
I mentioned this earlier, but we have genuinely done. We pride ourselves on being really, really user driven. We've done across our company of about 1,000 user interviews over the past year and a half, and we're doing these user interviews with the people who are the earliest adopters of AI in our space. When you talk to those users, we see five clear themes emerge when they ask them of what they are looking for in their research discovery tools.
Number one is a high quality, comprehensive corpus of content. That is why people still do use Google Scholar, because they trust that it has that the most. Number 2, they want to ensure that their tools have access to the fullness of that content so that they're not missing things. And it is the most accurate. It can possibly be. Number 3, they want products that have dedicated features for their workflow.
So think about reference management, saving papers, generating citations, maybe kicking off writing tasks, things that enable their workflows to move faster. And then 2, you see the new two attributes emerge. This is they want effective and they want effective. And that it's time saving. And it does what it says it's going to do and then accurate. And that it's not making things up and giving them incorrect information.
They want both of those in their AI functionality in the products. And then finally the magic on top that we hear that you have to infer this more when you see people use products and you hear them talk about their products, but it is now the expectation. And we've seen this in how people's usage of consensus has even shifted since ChatGPT has disseminated so much that it is now the expectation in 2025 that your products that you use will have a flexible, delightful, magical and AI native UX that users are now accustomed to.
So when you look across the landscape of what is out there to do scholarly discovery, and you layer in those five attributes, I think a real insight emerges as to why the only company, the only product that is truly exponentially going in 2025 is ChatGPT. These first three buckets are all more about academic rigor and academic workflow, and users are now showing their hand.
At least the early career researchers and some folks in the industry are now showing their hand that they're willing to compromise on some of those academic integrity and academic rigor features in lieu of making sure that their products have effective AI capabilities and a delightful AI, UX. And I think we can many people in this room probably can be honest with themselves as well, that we've all used AI to do a shortcut in something, and it's really, really hard to resist the power of a magical, intuitive, incredibly easy to pick up and powerful tool that is these new AI tools.
So where do we go from here. I think when you look across that landscape of tools and think about what users want, I think it's pretty clear. I think what we need to do to ensure that there are still high quality outcomes happening in sciences, and I think for everyone in this room, preserving our respective businesses, that we need, there to be tools that are academically aligned, that are our academically rigorous, that do all the things that the academic tools before us did of having a comprehensive index, of having the fullness of access to content, of having dedicated academic features, but they need to also do the things that are now the expectations of the new people in the market and the people who are demanding effective and accurate AI in order, demanding delightful, flexible, magical UX experiences in their AI native products.
So I'm going to transition now to talk a little bit about consensus. I'll share a little more detail. We'll have Christian give a short demo of the product. We're going to share a little more detail on what we are and how we think we fit into this landscape, and how we think we can solve some of these problems together. So first I said this in describing consensus, but I want this to be crystal clear that consensus is a search engine that has AI in it.
We are not an AI lab. We are not a chatbot. We are not an LLM. When I described what we are to people, the simplest way to do it is to say that we are what Google Scholar is, if it was rebuilt in 2025. We are a search product that has LLMs to do tasks within the product. We are not an LLM, just generating new text back to you.
Christian, off to you. All right. I'm going to play a little video here. That is going to be a demo. It's going to go pretty quickly to keep the pace going. So don't feel like you have to catch every single detail. The best way to learn about consensus is just to go to our website, consensus app, and start searching.
But anytime a user enters in a query in consensus, the first thing we do is find the most relevant academic research papers according to your question. We also synthesize the results using I. But we have a huge focus on the actual sources. So not only do we have inline citations at the top, but we have an entire dedicated section like a traditional academic database. With the research papers below.
We enrich those papers with cool AI enhanced metadata. We have pages for each papers and easy links to the full text. Our goal is to use AI to get users to the right papers, but in addition to some of those traditional Academic Search features, we also have cool AI powered features. For example, consensus conversational. You can ask follow up questions of the results, and consensus knows instantly what to search next with your entire context from above.
We also have cool AI powered visualizations that let you see across the literature what many, many papers say, and you can use AI to interrogate the research quality and authority of the different positions on a given topic. For each paper, we try to extract things like the study design and the sample size and the population. Things that researchers or doctors or students our users need to care about, to know which paper they need to go read next.
And of course, always easy links to the full text for access or in this case for purchase. So that's the demo and you can go to consensus app and try some searches. I didn't actually purchase that paper. I just put it in my cart. Maybe later tonight. Oops, sorry. Let's see if we can get this back.
Awesome and yeah, thank you Christian. What that has resulted in is we have captured some of this exponential growth that we're seeing in AI tools. Now, these are by no means ChatGPT numbers. Everyone is dwarfed, dwarfed by them. But we do have a spark of something. And if growth holds to any extent, by the end of the year, we will have passed many decades old products and be the number two most used interdisciplinary scholarly search tool on the market, with just holding and even potentially decreasing growth from what we've seen from the beginning of the year.
But there is a very core difference, again, between us and general purpose chatbots. We are designed to try to engage users with the source material, and it shows up in the data. 45% of consensus searches result in somebody going to the paper page of going to one of the full text links of a paper. This is in huge contrast to general purpose products that have a sub 1% click through rate of their source material.
And when we look at the data time and time again, will we see is our best users. The users who keep coming back and back to consensus are the ones who interact with the results. The probability of somebody retaining skyrockets if they interact with results. We want our users to engage with the source material. So how we fit into this landscape view that I showed earlier.
Our hope in what we are striving for is to be the product that checks all five boxes that can capture the magic and the exponential growth that we see in general purpose chatbots and AI tools, but also be academically rigorous, academically aligned, and caring about the underlying source material. But in order to do that, you'll see the one that does not have a yes and a. Yet it will take work in with people in this room and thoughtful cooperation of how we can all work together to try to lead to better outcomes in using AI in scholarly search.
And I'm going to pass it to JL to talk about how we're doing that today with publishers. All this talk of AI and the future of scholarly search is heavy. So I'm going to take a detour and tell you a love story 10 years ago, I found myself unexpectedly single, and so I was also faced with the apps and learning how to run a search there and being trained in publishing and information science.
Of course, looked for the advanced search option, and I found one and was able to determine that in a college town where I was occasionally working and was hoping to get a date, I could filter by prospects with a PhD. And so that's what I did. There were 6, and I persuaded persuaded one to go out with me. And as we sat having coffee, nervously holding our coffee cups, making conversation, learning about each other, I dropped the line that I knew would probably get attention, and that is that I helped build and launch Google Scholar.
The PhD across me, from me. Her eyes bugged out and she paused and well, now a bit of my background. So I was involved with launching that service, and Doctor Thompson, who's now my spouse, was duly impressed that day, and I'm quite excited to be part of this effort that for me, is quite familiar. It was 20 years ago that I was involved in persuading publishers, just like you, to take a leap and enable Google to bring your content to a wider audience to make it more visible.
And that's what we're doing. It is just like the partnership you now have with Google, if you're a publisher or if you're involved with a publisher, it's a no fee based license that we propose, whereby we will index the full text, display that full text in snippet form only after, of course, indexing it, adding it to our search index.
And then, as Eric and Christian demonstrated there, using our tech to surface it in New ways and to make it more engaging even than a traditional tool like Google Scholar. What's notable is that different than other tools, we think it's quite important that users be engaged. I spent a number of the last years of my career in e-commerce, where it's all about getting a shopper to push the conversion button and conversion.
Here is that full color link to a publisher's platform, which is what we're driving toward the authoritative version of the paper. There are a half dozen, for me, compelling reasons why, as a former publishing professional and as a would be partner to you all, that you should strongly consider partnering with us. The foremost is the one that we focused on in this presentation, which is that we are losing ground right now.
You may not be seeing double digit drops yet, but the chart that we showed signals what's coming and we are not the solution to that problem. We're one of the solutions that you can implement to preserve the traffic you have today, to maintain the value you're providing academic libraries and your other customers. But then we're also a gateway to a new user that we have not historically had, which is one that's not technically trained, but who does want to understand the symptoms that are experiencing or have a business problem to solve.
And with the magic of AI, we can make your content relevant to them understandable, less technical. That is the growth opportunity. But again, the principal one is to preserve the value that we already offer our customers. There's also the obligation factor, the most common request we receive from libraries that we are partnering with as our customer to provide our service to their students and researchers is that if they've paid for your content, they want to see it in a service like consensus.
It's frustrating to them not to see you there. And so if there's one rationale, which is that your customers expect it. But then also, as I hope we've demonstrated, we're on the same team. We're a tech company focused on consumer tech, but we're also deeply committed to the science that we're all trying to disseminate.
And you can count on us to over time, innovate in that direction. You may have noted the announcement by Wiley a couple of weeks ago about their partnership with perplexity. You saw reference to the two Elsevier products that are now I find you have peers who are on the move now. It's no longer defensive actions trying to preserve copyright. Still important.
It is time for offensive moves. And we represent one of those. And among the moves you could make were actually a relatively simple one. And we have publisher partners already who have gone through our content license agreement to make sure it's friendly to a publisher. And we have one incoming partner who I had an interesting experience with 20 years ago, when they were the big holdout for Google Scholar.
I remember marching into their offices with Doctor Anurag Acharya, the mastermind behind Google Scholar, that some of you probably know where we were met by an audience of 20 publishing professionals who spent three hours grilling us on all the ways that we could do something wrong with their content. We persuaded them then that publisher is actually among the first to embrace what we're offering at consensus.
After about a half dozen meetings in the same sort of rigor. And so you can be sure that the path is now being smoothed for your content to be properly indexed and cared for by us as your partner. But wait, there's more. These are two analytics and business intelligence nerds.
You can count on them. And us too. I didn't mean to label you, but that's our background in part. We're going to offer you the best indicators of what's happening to your content of any partner you have. That's our commitment, our ambition. And we're already anticipating offering what you see illustrated here more than I think any partner of yours would offer today.
And that's so that we can learn together on how to improve upon that user experience. So there are several ways that you can connect with us here at SSP. I'm happy to talk afterwards. We are exhibiting booth 42. Come see us there. You can also join us in the conversation. On Friday at 12:30, as you see, there's an AI solutions roundtable.
You can come to our table and we can chat in a group setting. Or you can go to our page here for scholarly publishers and raise your hand so we can send you the info. Or you can just contact me directly. And with that. Awesome that is the end, I think we have. I think we have two or three minutes for Q&A. One thing I do want to say, the jail did not mention in that love story is that his spouse, who apparently was wooed by your work on Google Scholar, is now a daily consensus user and no longer uses Google Scholar and very, very proud of that.
Do you want to just open it up to any questions. Thank you very much. I just have a couple of questions. The first is do you scrape publicly available information if you don't have a licensing agreement with certain publishers. And the other is have you ever worked with an association or any other kind of publisher for the sake of marketing, to help as ambassadors to help spread word of consensus.
Yeah so to the first question, the vast majority of our data comes through partnership with Semantic Scholar. The only time we have gone to the open web to get more things is to fill in some gaps of metadata. We have never displayed text that we are told to not display. We have been asked by some publishers even to take down some of the abstracts in the product in display. And if you click around in our product, you can see if you click into some there'll be no abstract showing.
It will just direct link out. So we have never we have never scraped anything that we do not have the ability to then display in the product. In terms of partnering with ambassadors, no, I think it's something we'd love to invest more in honestly like we are. We were so heads down in our first. We're about two years old of a product, and we were so heads down in those first two years of building a product and getting traction and focusing on ourselves, like us, being here is trying to be the beginning of reaching out and being more a part of this community and being an active part of it.
So something we'd love to do, not something we've done today. Appreciate it. I'll ask a question that'd be a softball. Eric, your favorite feature of consensus.
Oh, man. Well, the obvious answer. OK I have the obvious answer is the consensus meter. Because I really do deeply believe in visual and delightful and engaging UX. And we think that's a unique take on it. That's been around for a year now. So what I'll say is if you saw in Christian's demo, there was like a little table in the analysis section.
And that's one of the next visual features we're trying to put into the product. So a way to no matter what your results are, visualize some of the most important parts of it in a nice UX table. I think we're continuing to be excited about how we can do more dynamic, engaging, delightful visual features like that.
I will share an anecdote in lieu of a question. This suggests just how quickly things have changed. I have an older daughter who just graduated from got her undergraduate last year, and it was only at her last semester when she was doing a senior thesis that I took it on myself to explain to her how to use Google Scholar or how to use other tools. Before that, she was left to her instructors and librarians because I didn't think she was ready to get into a full literature search.
A review I have a second daughter, who's now a junior at a land grant school and is less academically inclined than her sister. However, when she explained how she was using a ProQuest database to do something and I got a sense she was using training wheels, I thought, no, let's take her up a notch. And so I introduced her to consensus, knowing that she would have no trouble figuring it out, that she would be able to break through the technicality of interpreting the relative authority of papers.
So that's just in a two year time frame. A tool Google Scholar, that would have been wrong for a senior in college is totally right for a less scholarly oriented junior in college. That is the new user opportunity that we're talking about. I love it. That one last. Any final questions, please. We are basically we actually are overtime.
So this could be the last one. Yeah can you break down that phrase relative authority. What does it mean. And how do you determine it. Relative authority. Is that from the demonstration of the consensus meter. That is what you just said. Yeah I think we're talking about a consensus meter. Yeah, yeah.
Yeah so we in what determines what determines how the papers come back in. Our product is very similar to how Google Scholar has done it for years of you learn from the ways that people interact with the papers. And you learn from what are those signals that make a paper wanting people to interact with it. And typically what emerges from that is classic signals of quality like citation count, recency citation count relative to recency, the journals, the authors, the institutions we have metadata on all of those things.
And over millions and millions of data points of people interacting, you learn what are the kind of right ways to weight those things, and that is how we view our search ranking. That is how really sophisticated search companies typically do search rankings. Obviously for us, that is also how Google does consumer search. They just look at a bunch of different signals. We look at research oriented signals.
And then in the consensus meter of Christian said something like that. We just have some of those simple like quality indicator aggregation. So we do some estimations of the methodology of the studies and say which ones are meta analysis, randomized controlled trials. And so we give them a little tick. We look at citation count sj r scores of journals and just have aggregated statistics of those.
So by no means are they any of those be all end all authority markers. But they're helpful. What our users want and giving them some indicators toward more quality research. Great question. Again come see us at booth 42 or on Friday, the AI solutions roundtable. Thank you.
Thank you, everybody.
Segment:0 .
Thank you, everybody, for taking time out of your lovely, beautiful afternoon in Baltimore and joining us here today. I'm Eric Olsen. I'm one of the founders of consensus. Throughout today I'll definitely talk a little bit more what consensus is. But for those who don't just in a sentence, consensus is an AI native Search engine for academic research.
The purpose of today's talk is to talk about the very, very fast changing landscape of scholarly search. If you haven't heard. AI is changing things, and we're going to talk about how we're looking at the landscape and what we think it means for everybody in this room, including ourselves.
I'm going to start with a quote. I think everybody knows we are in the middle of a change. And I think at least we think on this side of the table that the truth is that this change is just in the earliest of innings and it's going to be a constant change, just like everything always has in every domain. But it's going to be a constant change for the next 2, 3, 5, 6, 7, 8, 9, 10 years.
And we think the change in this space is perfectly captured by these two graphs. On the left, you see exponential growth, and on the right, you see about a 30% decrease. We have purposely obfuscated the titles of these graphs throughout the course of this presentation. I'll share what these graphs are, why we think they're really important, and what we think it means for us on this side of the table.
And lots of you guys in this room here today to just introduce ourselves. Again, I'm Eric. I am the co-founder and CEO of consensus. Sitting at the table here, we have our two other consensus team members who are with us all week. Here at SSP. We have Christian Salem, my co-founder and chief product officer.
And then at the end of the table, we have JL Needham, our partnerships lead, who's the former partnerships lead for Google Scholar and Google Books. Is a quick, super, super quick background story of consensus, I met Christian about 13 years ago. We actually were members of a college football team together at Northwestern, and we are the children of academics and teachers and scientists, and we had always felt like outsiders to this world, but had a deep passion for wanting to contribute to science and wanting to see science diffused more into the public and made more accessible.
So, chart number one. Unsurprisingly, it is related to AI. So what that chart shows is the estimated weekly active users of ChatGPT. That shows if it is estimated that there are 800 million weekly active users today of ChatGPT, g.p.d. That means that 1/10 of the world uses ChatGPT every single week. Now, in our world of consumer tech, we've seen lots and lots of exponential growth stories, but nobody has ever seen anything quite like this.
It surpasses everything that comes before it. If you look at that first notch right there, that is going from 1 to 50 million in about four months, I believe, which is the fastest that a product has ever done that. And it looks like a blip on the radar of how it has continued to progress today. And then on the right side of the screen, we see what that growth has done in the markets that we all care about.
The adoption is incredibly rapid where we're seeing the most in younger and earlier career students, and then in the Fortune 500, in an industry where people are also rapidly adopting these tools. As a consumer facing startup, we are always, always, always, always trying to iterate and improve on our product. So what that forces us to do is talk to lots and lots and lots of people who are using our product and who are the earliest of adopters of these tools.
So we have a really good pulse of who are these people that are adopting these tools early, and specifically, who are these people who are adopting them in the spaces in the markets that we all care about. So we're going to go over the who, what. And then it is obfuscated by my subtitles. But the why, it's very simple. So I'll be able to explain it. The who is kind of what we just showed on the last slide.
It is mostly think of it as earlier career researchers, whether that means people who are still students or people who are just beginning their academic research career journey. Then the other place we really see this technology start to disseminate is in industry, whether those in the Biosciences, whether that's in consulting world or even clinicians in the doctor's office. And then in what are people in our space using AI for.
If you think about research, at the Super highest level of abstraction is a three step process of learning, creating and sharing. What you really see is people using it for the first step in that process, in a little bit of the early stages of that second step of the process. So the learning and creating. We see people using AI in conversational and fluid, flexible search and discovery and learning of academic content.
We see them using AI models and products to summarize, ingest, consume, and analyze that content. Then on the latter two bullets, you see people using it for the early stages exploratory kind of kicking off of these deliverable creation writing creative tasks. And there's multiple things in each one of these buckets. But the y at the bottom, there's only one thing you hear if you ask these users over and over again, why are you using these products and why are you using them for academic work.
The answer is it saves them time. We're going to talk in this presentation about some of the threats and the downsides that this can all bring. And I'm sure all week in these next two or three days, you're going to hear a lot about the negatives and the threats. And those are all 100% real. And part of this talk is to talk about how to mitigate these downsides.
But it is important to remember that there are real positives of this. We all got into this space because we care deeply about science and believe in the power of knowledge, and believe that knowledge increasing improves society. And if these tools can speed up the day to day of the people who are involved in this knowledge creation loop, that absolutely has the potential to be incredibly net positive for society and something that we should all be really excited about.
Chart number two from that very first slide is the search interest traffic for Google Scholar. So those troughs are the summer. But the net decrease is about 30% if you map it back to 2022 around when AI products started to disseminate. And when we think about this chart and you think back to what I just showed of who are these users of AI products and research.
It's a lot of those early career researchers and students. And what we're thinking a lot about a consensus is those students, those people who are just beginning their careers, they may never learn the old habits of people before them, and they may never be tried and true Google Scholar users like the generations before them. And I think when you think about that means that graph might only just be starting.
And this is just the early innings of starting to see this downtrend. So that brings us to the next question is if that is true, what we just said, what we just showed is that there's going to be a continued decrease of using legacy traditional academic discovery tools for scholarly search.
What are the downsides. What does this mean specifically if they're going away from those tools. And like what we're seeing today, they're starting to use general purpose AI chatbots to do their scholarly discovery. We think that two core things happen that are problematic when people stop using research tools to do research.
The first is quite obvious sourcing, attribution and citations are no longer first class citizens of the products that they are using. What I mean by this is there are some high percentage of requests that go to these products that are just handled by sending text back to the user, that is just text that is generated from internal model knowledge. Model knowledge that might have learned from some of your content and there are no citations to be found.
The second part that can happen. And why they're an afterthought, is even if they are shown in the product and they're used in the response, they are not displayed prominently in the UX. They're not designed to be search engines. They're not search engines first, and they are very rarely ever interacted with, digested, and consumed by the user.
Even if these models do use a search function to bring in citations to their responses. The second big bucket of thing that happens is when those citations are used. We're typically seen to be lower quality content being used. The reason why we think this is happening is we love to crap on Google sometimes, and the people in the AI world do, but they're the greatest search company of all time.
And Google Scholar is a search algorithm that was tuned over decades of people using it, and that is trained to try to surface the best and highest quality content. These products are not Search Products. They're not search companies, and they're not learning from researchers, using them to try to surface the best possible content. So even when citations are used, not only are they an afterthought in the UX, but many times they are lower quality content than we're traditionally used to seeing in dedicated research tools.
And then yeah, the can't really see it with the titles. But the downstream effects of this, we think are honestly pretty scary. We think that number one, all of that amazing things that could happen of science progressing, that is at risk and lower quality science and research happens. And then number 2, for us and for everybody who produces content, that is a real risk to our respective businesses.
So the logical next question to us is what do we do about it. How do we reduce some of these downsides that can come from people starting to use not research tools for research. We are biased of what the solution or what the path to figuring out the solution should be. But as a consumer facing technology company, we always think it comes back to the user's needs and start with the user of what are they looking for so that we can get to the solution to our problems.
I mentioned this earlier, but we have genuinely done. We pride ourselves on being really, really user driven. We've done across our company of about 1,000 user interviews over the past year and a half, and we're doing these user interviews with the people who are the earliest adopters of AI in our space. When you talk to those users, we see five clear themes emerge when they ask them of what they are looking for in their research discovery tools.
Number one is a high quality, comprehensive corpus of content. That is why people still do use Google Scholar, because they trust that it has that the most. Number 2, they want to ensure that their tools have access to the fullness of that content so that they're not missing things. And it is the most accurate. It can possibly be. Number 3, they want products that have dedicated features for their workflow.
So think about reference management, saving papers, generating citations, maybe kicking off writing tasks, things that enable their workflows to move faster. And then 2, you see the new two attributes emerge. This is they want effective and they want effective. And that it's time saving. And it does what it says it's going to do and then accurate. And that it's not making things up and giving them incorrect information.
They want both of those in their AI functionality in the products. And then finally the magic on top that we hear that you have to infer this more when you see people use products and you hear them talk about their products, but it is now the expectation. And we've seen this in how people's usage of consensus has even shifted since ChatGPT has disseminated so much that it is now the expectation in 2025 that your products that you use will have a flexible, delightful, magical and AI native UX that users are now accustomed to.
So when you look across the landscape of what is out there to do scholarly discovery, and you layer in those five attributes, I think a real insight emerges as to why the only company, the only product that is truly exponentially going in 2025 is ChatGPT. These first three buckets are all more about academic rigor and academic workflow, and users are now showing their hand.
At least the early career researchers and some folks in the industry are now showing their hand that they're willing to compromise on some of those academic integrity and academic rigor features in lieu of making sure that their products have effective AI capabilities and a delightful AI, UX. And I think we can many people in this room probably can be honest with themselves as well, that we've all used AI to do a shortcut in something, and it's really, really hard to resist the power of a magical, intuitive, incredibly easy to pick up and powerful tool that is these new AI tools.
So where do we go from here. I think when you look across that landscape of tools and think about what users want, I think it's pretty clear. I think what we need to do to ensure that there are still high quality outcomes happening in sciences, and I think for everyone in this room, preserving our respective businesses, that we need, there to be tools that are academically aligned, that are our academically rigorous, that do all the things that the academic tools before us did of having a comprehensive index, of having the fullness of access to content, of having dedicated academic features, but they need to also do the things that are now the expectations of the new people in the market and the people who are demanding effective and accurate AI in order, demanding delightful, flexible, magical UX experiences in their AI native products.
So I'm going to transition now to talk a little bit about consensus. I'll share a little more detail. We'll have Christian give a short demo of the product. We're going to share a little more detail on what we are and how we think we fit into this landscape, and how we think we can solve some of these problems together. So first I said this in describing consensus, but I want this to be crystal clear that consensus is a search engine that has AI in it.
We are not an AI lab. We are not a chatbot. We are not an LLM. When I described what we are to people, the simplest way to do it is to say that we are what Google Scholar is, if it was rebuilt in 2025. We are a search product that has LLMs to do tasks within the product. We are not an LLM, just generating new text back to you.
Christian, off to you. All right. I'm going to play a little video here. That is going to be a demo. It's going to go pretty quickly to keep the pace going. So don't feel like you have to catch every single detail. The best way to learn about consensus is just to go to our website, consensus app, and start searching.
But anytime a user enters in a query in consensus, the first thing we do is find the most relevant academic research papers according to your question. We also synthesize the results using I. But we have a huge focus on the actual sources. So not only do we have inline citations at the top, but we have an entire dedicated section like a traditional academic database. With the research papers below.
We enrich those papers with cool AI enhanced metadata. We have pages for each papers and easy links to the full text. Our goal is to use AI to get users to the right papers, but in addition to some of those traditional Academic Search features, we also have cool AI powered features. For example, consensus conversational. You can ask follow up questions of the results, and consensus knows instantly what to search next with your entire context from above.
We also have cool AI powered visualizations that let you see across the literature what many, many papers say, and you can use AI to interrogate the research quality and authority of the different positions on a given topic. For each paper, we try to extract things like the study design and the sample size and the population. Things that researchers or doctors or students our users need to care about, to know which paper they need to go read next.
And of course, always easy links to the full text for access or in this case for purchase. So that's the demo and you can go to consensus app and try some searches. I didn't actually purchase that paper. I just put it in my cart. Maybe later tonight. Oops, sorry. Let's see if we can get this back.
Awesome and yeah, thank you Christian. What that has resulted in is we have captured some of this exponential growth that we're seeing in AI tools. Now, these are by no means ChatGPT numbers. Everyone is dwarfed, dwarfed by them. But we do have a spark of something. And if growth holds to any extent, by the end of the year, we will have passed many decades old products and be the number two most used interdisciplinary scholarly search tool on the market, with just holding and even potentially decreasing growth from what we've seen from the beginning of the year.
But there is a very core difference, again, between us and general purpose chatbots. We are designed to try to engage users with the source material, and it shows up in the data. 45% of consensus searches result in somebody going to the paper page of going to one of the full text links of a paper. This is in huge contrast to general purpose products that have a sub 1% click through rate of their source material.
And when we look at the data time and time again, will we see is our best users. The users who keep coming back and back to consensus are the ones who interact with the results. The probability of somebody retaining skyrockets if they interact with results. We want our users to engage with the source material. So how we fit into this landscape view that I showed earlier.
Our hope in what we are striving for is to be the product that checks all five boxes that can capture the magic and the exponential growth that we see in general purpose chatbots and AI tools, but also be academically rigorous, academically aligned, and caring about the underlying source material. But in order to do that, you'll see the one that does not have a yes and a. Yet it will take work in with people in this room and thoughtful cooperation of how we can all work together to try to lead to better outcomes in using AI in scholarly search.
And I'm going to pass it to JL to talk about how we're doing that today with publishers. All this talk of AI and the future of scholarly search is heavy. So I'm going to take a detour and tell you a love story 10 years ago, I found myself unexpectedly single, and so I was also faced with the apps and learning how to run a search there and being trained in publishing and information science.
Of course, looked for the advanced search option, and I found one and was able to determine that in a college town where I was occasionally working and was hoping to get a date, I could filter by prospects with a PhD. And so that's what I did. There were 6, and I persuaded persuaded one to go out with me. And as we sat having coffee, nervously holding our coffee cups, making conversation, learning about each other, I dropped the line that I knew would probably get attention, and that is that I helped build and launch Google Scholar.
The PhD across me, from me. Her eyes bugged out and she paused and well, now a bit of my background. So I was involved with launching that service, and Doctor Thompson, who's now my spouse, was duly impressed that day, and I'm quite excited to be part of this effort that for me, is quite familiar. It was 20 years ago that I was involved in persuading publishers, just like you, to take a leap and enable Google to bring your content to a wider audience to make it more visible.
And that's what we're doing. It is just like the partnership you now have with Google, if you're a publisher or if you're involved with a publisher, it's a no fee based license that we propose, whereby we will index the full text, display that full text in snippet form only after, of course, indexing it, adding it to our search index.
And then, as Eric and Christian demonstrated there, using our tech to surface it in New ways and to make it more engaging even than a traditional tool like Google Scholar. What's notable is that different than other tools, we think it's quite important that users be engaged. I spent a number of the last years of my career in e-commerce, where it's all about getting a shopper to push the conversion button and conversion.
Here is that full color link to a publisher's platform, which is what we're driving toward the authoritative version of the paper. There are a half dozen, for me, compelling reasons why, as a former publishing professional and as a would be partner to you all, that you should strongly consider partnering with us. The foremost is the one that we focused on in this presentation, which is that we are losing ground right now.
You may not be seeing double digit drops yet, but the chart that we showed signals what's coming and we are not the solution to that problem. We're one of the solutions that you can implement to preserve the traffic you have today, to maintain the value you're providing academic libraries and your other customers. But then we're also a gateway to a new user that we have not historically had, which is one that's not technically trained, but who does want to understand the symptoms that are experiencing or have a business problem to solve.
And with the magic of AI, we can make your content relevant to them understandable, less technical. That is the growth opportunity. But again, the principal one is to preserve the value that we already offer our customers. There's also the obligation factor, the most common request we receive from libraries that we are partnering with as our customer to provide our service to their students and researchers is that if they've paid for your content, they want to see it in a service like consensus.
It's frustrating to them not to see you there. And so if there's one rationale, which is that your customers expect it. But then also, as I hope we've demonstrated, we're on the same team. We're a tech company focused on consumer tech, but we're also deeply committed to the science that we're all trying to disseminate.
And you can count on us to over time, innovate in that direction. You may have noted the announcement by Wiley a couple of weeks ago about their partnership with perplexity. You saw reference to the two Elsevier products that are now I find you have peers who are on the move now. It's no longer defensive actions trying to preserve copyright. Still important.
It is time for offensive moves. And we represent one of those. And among the moves you could make were actually a relatively simple one. And we have publisher partners already who have gone through our content license agreement to make sure it's friendly to a publisher. And we have one incoming partner who I had an interesting experience with 20 years ago, when they were the big holdout for Google Scholar.
I remember marching into their offices with Doctor Anurag Acharya, the mastermind behind Google Scholar, that some of you probably know where we were met by an audience of 20 publishing professionals who spent three hours grilling us on all the ways that we could do something wrong with their content. We persuaded them then that publisher is actually among the first to embrace what we're offering at consensus.
After about a half dozen meetings in the same sort of rigor. And so you can be sure that the path is now being smoothed for your content to be properly indexed and cared for by us as your partner. But wait, there's more. These are two analytics and business intelligence nerds.
You can count on them. And us too. I didn't mean to label you, but that's our background in part. We're going to offer you the best indicators of what's happening to your content of any partner you have. That's our commitment, our ambition. And we're already anticipating offering what you see illustrated here more than I think any partner of yours would offer today.
And that's so that we can learn together on how to improve upon that user experience. So there are several ways that you can connect with us here at SSP. I'm happy to talk afterwards. We are exhibiting booth 42. Come see us there. You can also join us in the conversation. On Friday at 12:30, as you see, there's an AI solutions roundtable.
You can come to our table and we can chat in a group setting. Or you can go to our page here for scholarly publishers and raise your hand so we can send you the info. Or you can just contact me directly. And with that. Awesome that is the end, I think we have. I think we have two or three minutes for Q&A. One thing I do want to say, the jail did not mention in that love story is that his spouse, who apparently was wooed by your work on Google Scholar, is now a daily consensus user and no longer uses Google Scholar and very, very proud of that.
Do you want to just open it up to any questions. Thank you very much. I just have a couple of questions. The first is do you scrape publicly available information if you don't have a licensing agreement with certain publishers. And the other is have you ever worked with an association or any other kind of publisher for the sake of marketing, to help as ambassadors to help spread word of consensus.
Yeah so to the first question, the vast majority of our data comes through partnership with Semantic Scholar. The only time we have gone to the open web to get more things is to fill in some gaps of metadata. We have never displayed text that we are told to not display. We have been asked by some publishers even to take down some of the abstracts in the product in display. And if you click around in our product, you can see if you click into some there'll be no abstract showing.
It will just direct link out. So we have never we have never scraped anything that we do not have the ability to then display in the product. In terms of partnering with ambassadors, no, I think it's something we'd love to invest more in honestly like we are. We were so heads down in our first. We're about two years old of a product, and we were so heads down in those first two years of building a product and getting traction and focusing on ourselves, like us, being here is trying to be the beginning of reaching out and being more a part of this community and being an active part of it.
So something we'd love to do, not something we've done today. Appreciate it. I'll ask a question that'd be a softball. Eric, your favorite feature of consensus.
Oh, man. Well, the obvious answer. OK I have the obvious answer is the consensus meter. Because I really do deeply believe in visual and delightful and engaging UX. And we think that's a unique take on it. That's been around for a year now. So what I'll say is if you saw in Christian's demo, there was like a little table in the analysis section.
And that's one of the next visual features we're trying to put into the product. So a way to no matter what your results are, visualize some of the most important parts of it in a nice UX table. I think we're continuing to be excited about how we can do more dynamic, engaging, delightful visual features like that.
I will share an anecdote in lieu of a question. This suggests just how quickly things have changed. I have an older daughter who just graduated from got her undergraduate last year, and it was only at her last semester when she was doing a senior thesis that I took it on myself to explain to her how to use Google Scholar or how to use other tools. Before that, she was left to her instructors and librarians because I didn't think she was ready to get into a full literature search.
A review I have a second daughter, who's now a junior at a land grant school and is less academically inclined than her sister. However, when she explained how she was using a ProQuest database to do something and I got a sense she was using training wheels, I thought, no, let's take her up a notch. And so I introduced her to consensus, knowing that she would have no trouble figuring it out, that she would be able to break through the technicality of interpreting the relative authority of papers.
So that's just in a two year time frame. A tool Google Scholar, that would have been wrong for a senior in college is totally right for a less scholarly oriented junior in college. That is the new user opportunity that we're talking about. I love it. That one last. Any final questions, please. We are basically we actually are overtime.
So this could be the last one. Yeah can you break down that phrase relative authority. What does it mean. And how do you determine it. Relative authority. Is that from the demonstration of the consensus meter. That is what you just said. Yeah I think we're talking about a consensus meter. Yeah, yeah.
Yeah so we in what determines what determines how the papers come back in. Our product is very similar to how Google Scholar has done it for years of you learn from the ways that people interact with the papers. And you learn from what are those signals that make a paper wanting people to interact with it. And typically what emerges from that is classic signals of quality like citation count, recency citation count relative to recency, the journals, the authors, the institutions we have metadata on all of those things.
And over millions and millions of data points of people interacting, you learn what are the kind of right ways to weight those things, and that is how we view our search ranking. That is how really sophisticated search companies typically do search rankings. Obviously for us, that is also how Google does consumer search. They just look at a bunch of different signals. We look at research oriented signals.
And then in the consensus meter of Christian said something like that. We just have some of those simple like quality indicator aggregation. So we do some estimations of the methodology of the studies and say which ones are meta analysis, randomized controlled trials. And so we give them a little tick. We look at citation count sj r scores of journals and just have aggregated statistics of those.
So by no means are they any of those be all end all authority markers. But they're helpful. What our users want and giving them some indicators toward more quality research. Great question. Again come see us at booth 42 or on Friday, the AI solutions roundtable. Thank you.
Thank you, everybody.
Form Title Download
Transcript
Form Title Bookmark
Entire Media
Segment(s)
Custom Clip
Start At:
End At:
Form Title Share
Entire Media
Segment(s)
Custom Clip
Start At:
End At:
Form Title Accessibility and Keyboard Shortcuts
The cadmore media player is a professional video and audio player experience designed to deliver media as well as relevant transcript, segmentation, and metadata. On the screen is a traditional html video player with buttons such as play, pause, forward, captioning, language and play speed selection and more. Adjacent to the video player are generally tabs, one of which shows a scrolling transcript of the video and the other shows any segmentation of the media, sometimes referred to as chapters. At the top of the video is a menu bar containing an Explore button and Tools button. The Explore button allows you to dig deeper into the video by opening up an explore dialog which contains tabs for a visual navigation of the video using thumbnails, an about tab which contains rich metadata about the media, a segments tab which contains further information about the video, and a related tab which contains links to other media related to this media. The tools button opens a menu with options for sharing the media on social media, creating bookmarking links, creating embed codes for your website, generating citations, and more.
Keyboard Shortcuts
Keyboard shortcuts only work while you are in forms mode. By entering forms mode, you will be able to quickly access portions of the media player with single key commands.
Spacebar: Play/Pause Toggle
I: Info Menu,
O: Tools Menu
T: Transcript Tab,
S: Segments Tab
F: Forward 15,
R: Rewind 15
L: Languages,
P: Playspeed
M: Mute Toggle,
V: Volume Bar
N: Video Only Toggle,
B: Full Screen Toggle
C: Captions Toggle
Escape: Closes dialogs and menus / Exits Player
Z: Next Transcript Chapter
X: Previous Transcript Chapter
Q: Next Transcript Paragraph
W: Previous Transcript Paragraph
Entire Media
Segment(s)
Custom Clip
Start At:
End At:
Embed Size
-
You Size - Responsive
-
We Size - Responsive
-
Picture Overlay Popup
-
400 x 225
-
512 x 288
-
400 x 700
-
768 x 1024
-
1024 x 512
-
1280 x 608
Form Title Cite
No citation provided.
Info
Segments
Related
Thumbnails
Exit Clip Mode
You have requested to see a portion of the video outside of the clip.
Click OK to switch to the full video or click if you would like to stay inside the clip. You can always go back to the clip with Tools and Clip Mode