When you typed who is on snl tonight? You used to get 10 pages of clickbait that made you scroll down endlessly for the answer. The best product for most questions is to just give you the answer. That said, the death of discovery is worth mourning. I hope someone solves it but this was dying before ai came along. SEO killed the web, AI killed SEO, and so it goes…
If you want to win the war on AI search you just have to have a browser and an LLM answering all search queries. I don't really care who's behind the answer as long as it is right (which 99% of the time it is) so there is no loyalty at all, no need to switch, no need to try the competition, nothing. AI is the answer and it will get better with time
I use Brave, I don't know if it uses Leo (brave search) or gemini, or chatgpt, I just know the answers it gives me are the ones I am looking for and most of the time I extend the conversation with some of the options offered, hardly click on any link below the answer provided (which most of the time those links are garbage)
If Google didn’t put those there, people would be going to other LLMs to get their answers quickly. I don’t know how this ends but I strongly doubt any unilateral effort to withhold llm service by Google changes the medium term outcome.
I think it's actually wise of Google to disrupt themselves this way. If they don't the AI companies will do it for them, and they'll have no control over the process nor chance to recover lost ad income. As per the article, Google managed to pull this off without hurting ad revenue (surprise!)
Disclaimer: I wrote the blog post and submitted it
Is the future of providing content hosting markdown files and images surfaced with a sitemap xml file exposed to LLM providers, or perhaps just offering an mcp?
It’s a broad representation of US search, not a spam-focused list. But it does have some biases.
Google Ads and Bing Ads services. In practice, they grow the list by asking Google’s Keyword Planner for related terms and collecting things like “related searches” from results pages. But sourcing keyword queries like this tends to skew to queries with commercial value, reflecting what advertisers care about.
That was my guess too, but it seems like Google pulled it off so far. Clicks on ads didn't change, so Google is taking care to not cannibalize their own revenue.
Did they do unsustainable things that hurt their advertisers in order to temporarily goose ad clicks and earnings? Almost certainly, and I've heard anecdotes about that.
When you typed who is on snl tonight? You used to get 10 pages of clickbait that made you scroll down endlessly for the answer. The best product for most questions is to just give you the answer. That said, the death of discovery is worth mourning. I hope someone solves it but this was dying before ai came along. SEO killed the web, AI killed SEO, and so it goes…
If you want to win the war on AI search you just have to have a browser and an LLM answering all search queries. I don't really care who's behind the answer as long as it is right (which 99% of the time it is) so there is no loyalty at all, no need to switch, no need to try the competition, nothing. AI is the answer and it will get better with time
I use Brave, I don't know if it uses Leo (brave search) or gemini, or chatgpt, I just know the answers it gives me are the ones I am looking for and most of the time I extend the conversation with some of the options offered, hardly click on any link below the answer provided (which most of the time those links are garbage)
If Google didn’t put those there, people would be going to other LLMs to get their answers quickly. I don’t know how this ends but I strongly doubt any unilateral effort to withhold llm service by Google changes the medium term outcome.
I think it's actually wise of Google to disrupt themselves this way. If they don't the AI companies will do it for them, and they'll have no control over the process nor chance to recover lost ad income. As per the article, Google managed to pull this off without hurting ad revenue (surprise!)
Disclaimer: I wrote the blog post and submitted it
Is the future of providing content hosting markdown files and images surfaced with a sitemap xml file exposed to LLM providers, or perhaps just offering an mcp?
https://github.com/gordonbrander/zim-mcp as an example reference implementation
My local search interface can then be models updated over time with diffs via torrents, plugged into mcps for grounding and context not yet in models.
https://pirateface.co/articles/pirate-face-cli
Good. Searchers get the information they are looking for faster.
“US keywords DataForSEO tracks”
As a user, I am happy to see content spam trash to die and burn in hell.
It’s a broad representation of US search, not a spam-focused list. But it does have some biases.
Google Ads and Bing Ads services. In practice, they grow the list by asking Google’s Keyword Planner for related terms and collecting things like “related searches” from results pages. But sourcing keyword queries like this tends to skew to queries with commercial value, reflecting what advertisers care about.
Disclaimer: I wrote the article
This is temporary. It's against their business model to provide a service that's useful to its users.
That sounds bad for the ad revenue
That was my guess too, but it seems like Google pulled it off so far. Clicks on ads didn't change, so Google is taking care to not cannibalize their own revenue.
Did they do unsustainable things that hurt their advertisers in order to temporarily goose ad clicks and earnings? Almost certainly, and I've heard anecdotes about that.
Disclaimer: I wrote the article
It didn't affect their ad revenue, which is what Google wanted to get out of the program:
Clicks on ads didn't change, which suggests Google takes care not to cannibalize its own revenue.
In an ironic twist, OpenAI and others are placing advertising in their AI responses.
Those poor websites^H^H^H^Hcontent sources