Also in the episode: fan-out queries running site: searches on domains that don't exist, and how his team turned that into an expired-domain finder.

Full episode: youtu.be/ZntIGuNx29c?...

Curious what you've measured. Anything on this list working for you?

llms.txt. He's never seen one increase visibility in an AI answer. Its real use is pointing actual agents at your API, MCP and docs.

What does work is unglamorous: describing your brand the same way on your site, LinkedIn, G2, press releases. LLMs want consensus.

Structured data as a fact pipeline. In a published test, when the page text and the schema disagreed, the text won.

When the fact lived only in schema, ChatGPT answered ~20% of the time and Claude 0%.

Schema still helps. Just not as a place to hide information.

Markdown copies of your HTML pages. He ran it across hundreds of thousands of chats and found zero cases of an LLM citing the .md version over the HTML page.

Cost of shipping it anyway:
- duplicate content
- twice the crawl budget a diluted internal link graph

Full episode: www.youtube.com/watch?v=nItA...

Full episode: www.youtube.com/watch?v=yxpi...