AI Overviews: what changed, what did not, and what to do
Informational queries lost clicks. Commercial queries mostly did not. The response is structural, not a new set of tricks.
By Digiloom search team · Published · Updated
Where clicks actually went
The impact has been uneven. Broad informational queries — definitions, how-does-X-work, simple comparisons — lost a meaningful share of clicks to the answer panel. Transactional and local queries changed far less, because people still want to choose, book or buy.
If your traffic was mostly top-of-funnel explainers with no commercial connection, you felt this. If it was service pages and product categories, much less.
What answer engines actually reward
Content they can lift cleanly: a specific claim, stated once, in a self-contained passage, with a source or a date attached. Hedged marketing prose is unquotable.
A consistent entity footprint. Models assemble a picture of your organisation from many sources. If your description, service list and founding facts differ across your site, LinkedIn, directories and press, you are harder to represent confidently.
Conventional authority signals still matter. Being cited by an answer engine correlates strongly with being cited by other publishers.
Practical changes to make
Lead each section with the answer, then explain it. Invert the usual build-up structure.
Define terms explicitly rather than assuming context, and date your statistics.
Use headings that match how people phrase questions.
Keep FAQ and Article schema accurate and honest — markup that misrepresents the page is a liability.
Decide your AI crawler policy deliberately rather than inheriting a default.
What has not changed
Technical health, crawlability, site speed and internal linking matter exactly as much as before.
Commercial intent pages still need to rank in the conventional results, and still convert the same way.
Publishing a lot of thin content is still a bad strategy — arguably a worse one now.