The file nobody reads just got added to the test everybody runs
llms.txt barely gets fetched by AI, and Google says it doesn't use it. So why did Google's own Lighthouse start grading it? An honest, data-led take.
Blog
Notes from the table.
Written when there's something worth saying about search, GEO, AI visibility or the operator-with-agents model. No content calendars, no SEO bait. Just notes.
Google didn't say it. Its AI did. A German court says that's the same thing.
A German court just ruled that Google owns the words its AI Overviews generate. Why the ruling is right, why it's a mess, and why the big G will switch the answers off before it stands behind them.
The clicks aren't dying. They're emigrating.
Google clicks are falling and the panic industry is booming. Strip out everyone's incentives, mine included, and the data says traffic is moving, not ending.
Who AI tells UK founders to raise from
I ran 40 founder fundraising questions through ChatGPT, twice each. Over half named investors and cited no source. The data, and what it means.
LLM SEO is a contradiction in terms
You cannot optimise a model's frozen training data, and optimising what it retrieves live is just SEO. So LLM SEO is a contradiction, and this is what to do instead.
GEO is a bad name for it. It won anyway.
The industry settled on GEO for the work of getting cited in AI answers. It's an ambiguous, slightly inaccurate name almost nobody loves. This is how it won anyway.