Our summary
Search data is more useful when it connects the words someone searched with the page they saw. Capital & Compute describes collecting those pairs from Google's Search Console service. Jev helps sort them into possible title fixes, wrong-page matches and cases where a good-looking result still gets no clicks.
Simple text checks remove pasted web addresses and special search commands before Jev reads the remaining rows. Jev estimates how likely a title is to promise the requested answer or a description to be specific. Code combines these estimates with search measurements to make work lists. Claude Code then changes titles and links.
The author reports more clicks after several changes, but that does not show which tool caused the gain. Two new pages supplied most of the clicks. Desktop traffic alone cannot establish that visitors are bots. Small hand-labeled checks help inspect questions, but are not enough to validate every cutoff.
Key takeaways
- Join search queries with pages and keep the raw data so later comparisons use the same basis.
- Ask about text you provide, such as whether a title answers a query, rather than asking the model to predict Google's behavior.
- Count new-page traffic separately from older pages. Record the starting measurements, decide when to undo a change, and set a date to check again.