LLM Value Chain

I found this image provided by the UK’s Competition and Markets Authority (CMA) very helpful for understanding the current LLM value chain. In the image they use FM which stands for Foundation Models. Specifically you can see where the big tech incumbents are currently playing. Some interesting pieces to me are:

  • The FM release will be how other enterprise and start up companies are able to leverage these FMs at scale for proprietary data. Perhaps startup FMs will run in tandem with these as well. I am excited about innovation here because it opens up the door for a lot of improvements to SEO and CRO executional improvements and changes. 
  • It will be interesting to follow how the bottom row continues to play out since that is where the end users. I am excited to see how AI makes its way into productivity software like Gsuite, Android, Microsoft Office, or  iOS. 
  • Notice where Apple is playing? They’re developing models and with its mobile ecosystem of users has a chance at shaking things up. 
  • When people talk about GenAI favoring incumbents, this image makes that clear. 
LLM Value Chain

AI, Blockchain, and Search

I recently finished Read Write Own by Chris Dixon. I found the brief section titled, Artificial Intelligence: The New Economic Covenant for Creators the most interesting because it most closely ties to search. He quickly summarizes historical context around content creators bringing supply and distributors like Google putting that content in front of relevant users. The creator base was fragmented back in the 90s and early 2000s and still today. Dixon argued the fragmentation of content creators prevented them from gaining any sort of power in the relationship between supply and distributor. Had creators been able to band together, the Google we know today might have been different. Instead the result was Google gathering more users to amass a 80%+market share in search today through moving to compliment products it gave away for free such as an internet browser (Chrome) and mobile OS (Android) then moved most of the commerce to search with ads and built a great product that kept users coming back. 

The connection both AI and blockchain have to this history of the internet is that it is a chance to reset the power dynamics. His big argument is that blockchain is the only way to do it because of the network design. The root of the design is to give more power to people who contribute value to the network, not the management teams of those networks. If a traditional corporate network tries to do this, over time it will just turn into what you see today as the network attracts users and then extracts as much value as it can for the benefit of the people who own the corporate network (founders, shareholders,etc.).  

It is a short part of the last chapter on pages 216 to 222 with some thought provoking questions at the end.

My thoughts are that I want the creative work I am a part of on the internet to have a chance to perform in search and get credit when value is generated. I also want to be able to find and point to the creative work of others as a user. I am really excited about that end-state. The design or technology that powers that discovery is less of my expertise but I find it interesting to stay on top of new technologies and ideas that might gain traction like blockchain.

Search Market Predictions

Coming off OpenAI’s announcement last week about adding more links out within its answers, I am going to spend the blog post trying to predict what the search market will look like in 5 years which will take us to the year 2029. At the end, I’ll give my thoughts on how to think about building skills now to be relevant in this potential world I am predicting. 

I am trying to remember to take a step back more often and think about what this means for the businesses I help and the career trajectory I am on. Here’s the predictions I have: 

LLM powered AI assistants become used for deep researching or more intimate feedback or help (e.g. writing essay, memo, or idea generation). 

Traditional search like Google has LLM technology embedded into it basically becoming another SERP feature. 

The lines between AI assistants and traditional search get blurred. You may use traditional search then move into the AI assistant experience. Rand Fishkin touched on this in a post. He was torn on categorizing OpenAI as productivity or search. 

I predict these blurred lines will result in search becoming more fragmented over the next 5 years. Google keeps the majority share but loses the 90% share it has today. Imagine something like 60 to 70% and the other 30 to 40% is spread across search engines and AI assistants. 

To date, search has been a winner-take-all market but I don’t see it staying that way with the rise of LLM technology and the momentum of others in the space. 

At the end of the day, this is all about consumer behavior and if people have incentives to make the switch. I remember when my ridesharing behavior changed in 2017. Prior to that, I stuck with Uber to avoid more apps on my phone. I favored simplicity. But in 2017, I started using ridesharing a lot more than usual while doing my part-time MBA. When class finished at 9PM, a rideshare was much easier than public transit. I downloaded Lift and was committed to finding the cheaper option for each ride so I would switch between apps and order the ride that had the lower price.

My incentive was a lower price. What may be the incentive for consumers to switch their search engine preferences? I think the incentive will be that the alternative options could be that much more helpful to have in your phone and at your fingertips.

An iPhone user would need to download a second app and not use the Safari browser. An Android user would do the same and avoid the Google search bar likely prominent right on their home screen. 

What this means for businesses with target customers using search is that I predict organic search traffic becomes more fragmented. It will come from more sources. Google still has majority share in your organic search but not the 90%+share it has today – something a bit lower than that and you see organic traffic coming in from AI assistants that begin to get more usage.  

For people building careers in search today, I recommend leaning into these changes by leveraging the skills you’ve built to date and following the users. Specifically, the way you’ve understood customer behavior and then figured out how to act on it in accordance with what Google rewards. You can apply that to other platforms as well. Build strategies and business cases that provide a compelling reason to invest and go execute. 

In a fragmented search world, managing the strategies and performance on more platforms will be challenging but also will mix things up in fun and interesting ways.

No matter the platform, I believe sending traffic out for users searching is fundamental to helping users finish the task. OpenAI’s announcement supports it. We’ve learned a while ago from answer boxes that it might not always be needed but it will continue to be a big part of search. 

Do you think this might happen or something else? Hit me up. I’d love to hear your thoughts and why.

Web3’s Impact on Web2

I’m reading Read Write Own by Chris Dixon which is about blockchain technology and a good refresher on the history of the internet. 

It is getting me thinking about the impact Web3 could have on the industry I work in(search), the skills I’ve built, and the companies I help.  

Similar to what GenAI did in 2023, if it gains steam and becomes reality it’s going to create a bunch of threats and opportunities for the Web2 companies I’ve helped over the last decade. In tandem with the threats and opportunities will also be distractions. 

The more I read the book, the more interesting and powerful Web3 sounds since the power will be in the user’s hands. At the same time, I am curious to see the businesses that emerge in a potentially more popular Web3 world, what Web2 companies do if this trend happens, and if and how it impacts internet search. 

If you don’t want to splurge on the book just yet, this is a podcast episode where Chris covers it some. Another good high level post on Web3 here.

Product Discovery Framework Applied to SEO

A video from 2016 made the rounds amongst the Wayfair SEO Product team this week and I found it very useful. Teresa Torres shares insights into her experience working with Product & Engineering teams as well as defines what product discovery is and how to measure success.

As the discipline of SEO gets intertwined more in product management, there is a lot we can learn from product discovery frameworks. Curious technical and analytical minded SEO practitioners can naturally excel at product discovery because we’ve been doing details of it for years. By that I mean actions such as competitive analysis, keyword research, data analysis, etc.

While we’re naturally good at the details, we flock to solutions for our collaborators. I have found from experience that flocking to solutions is not as effective. What is effective is defining outcomes and problems for teams to discuss, refine, and rally around first. Then diving into solutions & experiments.

See a very simple example below…this can certainly get more complex with other strategic SEO pillars such as internal linking or schema but starting simple is key.

  • Business Objective: Grow our base of new customers
  • SEO Objective: Begin ranking for new keywords tied to priority business topic with a minimum of 2K new pages.
  • SEO Outcome: Drive new customers to site by increasing organic traffic to new pages with new keywords by H2 ’22.
  • SEO Opportunities / Problem(s)
    • Potential new customers are searching Google every day for what we sell. We identified 5K relevant keywords with search demand which we aren’t ranking for.
    • No clear system for programmatically generating new pages.
    • Manual creation would take far too long.
    • Discuss more with your team…..

Stopping here to call out that this is Product Discovery. If you are in an SEO Product Manager role, I recommend trying to frame your idea or project in the context above first prior to going into what we get to below which is known as Product Delivery. This is so important due to the cross-functional nature of an SEO Product Manager. You need to give your team ownership and context of the business objective and problem, then invite them to use their expertise on solutions / experiments.

What I like about this approach too is I hope you can feel less alone on your SEO island where you feel like the only one driving ideas and presenting solutions.

  • Solutions:
    • Consider technology vendor who has an out of the box solution
    • Build scalable system for programmatic page generation
    • Leverage machine learning to pursue a mix of programmatic content creation and manual curation
    • Discuss more with your team…..
  • Experiments:
    • Run 3 month pilot with vendor solution and measure impact.
    • Use GPT-3 machine learning model to produce content for 25-50 pages and gauge if it comes back descriptive and quality enough for customers.
    • Discuss more with your team…..