How do models deal with assessing the quality of content and its accuracy/veracity when recommending products currently? What do the providers do to avoid a situation where more content === more traffic? Would love to see links to relevant research on this, if you have them. much success to you, appreciate your ai slop risk awareness.
There is the preselection, which depends on the fanout queries the model comes up with and the contents performance across those queries on the search index.
After that content is actually assessed by the model. This paper tried different strategies to improve performance for this last step: https://arxiv.org/pdf/2311.09735. Adding statistics, sources, original data are all strategies that we apply.
In classic SEO, creating more and more content leads to "cannibalization". Generally this hurts performance of all overlapping content so much that it is not worth it.
That's a super valid question, we get it a lot. There are a lot of overlaps.
In our view Profound and Airops are aimed at existing marketing teams. Our goal is to be more hands-off, so you don't need a team. With many of our clients we act more like an agency, communicating via Slack and automating step by step. That's the experience we want to create. We aren't there yet though.
Our view on Peec is that it is an analytics solution. They recently did launch an actions feature. But they do not take any actions (yet). Creating content takes a lot of resources. And agencies are expensive.
We currently simply integrate with your Google Analytics and filter by Source. This tends to be a lower bound, since it's not always set correctly. Coming from some of the native apps, users might be categorized as direct visitors.
There are other data sources we want to enable in the future like Cloudflare.
We think about it like this: all of these agents will be most useful to users if they provide valuable answers. So they will be looking for valuable content for grounding their answer.
There are exploits, you can overfit on whatever they currently use as an objective function. But those tend to be temporary. So in the long run, valuable content will win. That's what we aim to create. It's a fine line.
Do you doubt the statement on how to maximize usefulness? Or do you mean that the companies behind the models might not optimize (exclusively) for usefulness to the user?
> Do you doubt the statement on how to maximize usefulness?
Yes; the customer here is the site using it, not Google end users, who'll tend to accept whatever's the top search result even if it's deeply wrong or complete slop.
The wellbeing of search users isn't really the priority here, right?
Yes, that is correct. We help the brands, not the end user.
Let me try to rephrase the line of thinking:
To maximize value to the end user, the models generally aim to be helpful. The companies building these models are incentivized to make the model use helpful content.
Our goal is to be aligned with their objective function long term. And that incentivizes us to create helpful content.
Not all of this is a given. We don't know for sure how it will play out. There will always be ways to game the system. But we think those will get fixed over time.
Regarding the topic of ambient agents, what’s the impact of your product? It’s hard for me to imagine the impact but I guess it must be a necessity if we have ambient agents to get discovered at all right? Nice to see a player from Europe on the market too!
How do models deal with assessing the quality of content and its accuracy/veracity when recommending products currently? What do the providers do to avoid a situation where more content === more traffic? Would love to see links to relevant research on this, if you have them. much success to you, appreciate your ai slop risk awareness.
There is the preselection, which depends on the fanout queries the model comes up with and the contents performance across those queries on the search index.
After that content is actually assessed by the model. This paper tried different strategies to improve performance for this last step: https://arxiv.org/pdf/2311.09735. Adding statistics, sources, original data are all strategies that we apply.
In classic SEO, creating more and more content leads to "cannibalization". Generally this hurts performance of all overlapping content so much that it is not worth it.
What do you guys do differently than Profound or Airops?
That's a super valid question, we get it a lot. There are a lot of overlaps.
In our view Profound and Airops are aimed at existing marketing teams. Our goal is to be more hands-off, so you don't need a team. With many of our clients we act more like an agency, communicating via Slack and automating step by step. That's the experience we want to create. We aren't there yet though.
Add peec to that list.
True, it is very competitive.
Our view on Peec is that it is an analytics solution. They recently did launch an actions feature. But they do not take any actions (yet). Creating content takes a lot of resources. And agencies are expensive.
As an analytics solution it is a good option.
how do you track where users are coming from?
We currently simply integrate with your Google Analytics and filter by Source. This tends to be a lower bound, since it's not always set correctly. Coming from some of the native apps, users might be categorized as direct visitors.
There are other data sources we want to enable in the future like Cloudflare.
Ugh. The worst of SEO, but a bunch more of it? Noooooo.
I get it, there is a lot of worry about slop.
We think about it like this: all of these agents will be most useful to users if they provide valuable answers. So they will be looking for valuable content for grounding their answer.
There are exploits, you can overfit on whatever they currently use as an objective function. But those tend to be temporary. So in the long run, valuable content will win. That's what we aim to create. It's a fine line.
> all of these agents will be most useful to users if they provide valuable answers
This is a bald assertion.
Do you doubt the statement on how to maximize usefulness? Or do you mean that the companies behind the models might not optimize (exclusively) for usefulness to the user?
I do share doubts about the latter.
> Do you doubt the statement on how to maximize usefulness?
Yes; the customer here is the site using it, not Google end users, who'll tend to accept whatever's the top search result even if it's deeply wrong or complete slop.
The wellbeing of search users isn't really the priority here, right?
Yes, that is correct. We help the brands, not the end user.
Let me try to rephrase the line of thinking:
To maximize value to the end user, the models generally aim to be helpful. The companies building these models are incentivized to make the model use helpful content.
Our goal is to be aligned with their objective function long term. And that incentivizes us to create helpful content.
Not all of this is a given. We don't know for sure how it will play out. There will always be ways to game the system. But we think those will get fixed over time.
Regarding the topic of ambient agents, what’s the impact of your product? It’s hard for me to imagine the impact but I guess it must be a necessity if we have ambient agents to get discovered at all right? Nice to see a player from Europe on the market too!
Please don't override the browser's default scroll behavior. It's so jarring and basically never a good idea.
Thank you for the feedback. We'll launch our new site soon where this is fixed.