First, rule out actual breakage
Real SEO breakage has fingerprints: rankings dropped for pages that used to hold, indexation fell after a site change, Core Web Vitals collapsed, or a manual action landed. Check Search Console for coverage and position changes, run a crawl for technical regressions, and compare year over year by page. If rankings and indexation are stable while clicks fall, your SEO is not broken. Something upstream of the click changed.
The upstream change: answers without clicks
Two forces take clicks from stable rankings. On the results page, AI Overviews resolve more queries in place, so position three earns fewer visits than it did a year ago even when nothing about your page changed. Before the results page, buyers confide their situation to ChatGPT, Gemini, Claude or Perplexity, receive recommendations, and only search to validate. Both shrink the click pool your rankings draw from, which shows up in your analytics exactly like broken SEO: same positions, less traffic, weaker leads.
The diagnostic split: if impressions hold while clicks fall, AI Overviews are absorbing your queries. If branded validation searches rise while non-branded discovery falls, assistant conversations are pre-framing your buyers, the pattern detailed in GEO vs SEO.
The fix is addition, not repair
You do not fix this by doing more SEO; the rankings were never the issue. You extend the program to the answer layer: measure which of your category’s questions and buying moments get answered with competitors across the four assistants, fix the sources behind those answers, and publish for the confided situations no keyword tool shows you. Your existing SEO becomes the foundation, strong pages are what assistants cite, and the answer layer becomes where the lost clicks went. Start with proof rather than theory: a free AI visibility audit shows where your traffic actually went, with screenshots, on a 30 minute call. And once you extend, measure it properly: how to know if your GEO and AEO efforts are working.
