Research
Things I measured while building this site, written up with the data behind them. Each study publishes its aggregates as a file you can download and check the page against.
- Searches for online tools follow the school year. In Australia, it runs six months the other way.
Eight years of monthly searches for 2,387 tools in five countries. US tool searches are lowest in July and highest in October; Australia's year runs six months the other way. School tools swing 2.9 times as hard.
27 September 2026 - Tools that answer a question are losing their searches. Tools that do a job are gaining them.
Seven years of monthly US searches for 5,716 online tools. Since the year before ChatGPT, the median tool that answers a question lost 28% of its searches; the median tool that does a job gained 30%.
27 September 2026 - People don't convert files. They escape them.
A year of US searches to convert files, read for their direction. For every search to make a WebP there are 14 to leave one; nobody searches to make a HEIC. The formats people leave are young, and the ones they go to are more than 30 years old.
27 September 2026 - Keyword difficulty is a weak signal. I checked 12,512 search results pages to see how weak.
Every keyword tool sells a 0 to 100 difficulty score. I bought the actual first page of Google for 12,512 queries and counted who was standing on it. A low score was wrong 28% of the time, and the number that did predict a way onto the page was the weakest site already there.
5 September 2026 - Every kind of browser-tool search grew after 2023. Developer utilities did not.
Yearly search volumes for 6,823 tool queries with a full five-year history, by the job they name. Six categories rose +16% to +92% from 2023 to 2025. Developer utilities rose +5%.
5 September 2026 - What Google's first page looks like when you search for a tool.
4,123 first pages read for the boxes Google adds and 12,512 for who the results are: an AI Overview on 42%, YouTube on 30%, and no major brand at all on 37%.
5 September 2026