How we tell a redesign, a Google update and AI apart
We diagnose a traffic drop in a fixed order: confirm it’s real, date it, test what your site changed, overlay Google’s confirmed updates, then check where the clicks went, including to AI answers. Order matters. Matching the drop to an update date first would have blamed Google, wrongly, in two of the four public cases we checked.
The steps, in order
Rule out a false alarm
We rule out tracking breaks, seasonality and falling demand first, so we don’t diagnose a drop that isn’t there.
- Tracking breaks: we compare Search Console clicks with your analytics. If analytics fell and Search Console didn’t, the tag broke, not the traffic.
- Seasonality: we compare each month with the same month a year earlier. In our heating-company case, April 2026 was 57% below April 2025, far outside that business’s usual spring dip.
- Falling demand: we check impressions. If impressions fell while average position held, fewer people searched.
Date the drop
We date the drop, the time of death, because every later test is matched against that date. We look for the month clicks fell 30% or more below the trailing three-month high and stayed there for two months. A sustained fall of 20% or more is still worth a check; 30% or more for two months is what we treat as a clear drop.
Test what your site changed
We test your site’s own changes before Google’s, because they’re fixable and the most expensive to miss. We pull pages archived before the drop from the Wayback Machine and request each one live, recording 404 errors, redirects to the homepage and redirect chains. We also check domain redirects, noindex tags, canonical tags, robots.txt rules, and whether your content appears without JavaScript.
Overlay Google’s confirmed updates
We overlay every confirmed Google update on the drop date, to see which ones could be involved. The source is Google’s Search Status Dashboard. See the Google update calendar.
Find where the clicks went
We compare position, impressions and clicks by page and query, to see whether the loss follows the pages that changed, spreads across pages that didn’t, or sits on queries where rankings held. For queries where rankings held and clicks fell, we check for AI Overviews and AI Mode answers, and who they cite.
Name the cause and grade the odds
We name the cause and grade each part of the loss, so you know what can come back. The grades:
- Recoverable. The cause is on your site, and fixing it is in your hands.
- Partly recoverable. Part of the loss can come back. The evidence says which part.
- Waits on Google. You can fix what’s on your side, but whether traffic returns depends on Google reassessing the site, possibly not until a later update.
- Structural. The search results changed, usually because an AI answer now takes the click. Fixing your site won’t bring those clicks back.
Price the loss
We price the loss in dollars a month, by page and query, so you can see which parts are worth acting on. We multiply the clicks you lost, from Search Console, by the cost per click advertisers pay for the same queries. That figure is a stand-in for what the traffic was worth. It isn’t your revenue.
What evidence do we use?
For the free Cause Check (outside data only)
- Third-party estimates of your monthly visits from US search and your rankings over time.
- Google’s Search Status Dashboard, for update dates.
- The Wayback Machine’s archive, for your old URLs.
- Live requests to your URLs, robots.txt and page source.
- What you tell us changed, and when.
Added for the Autopsy
- Your Google Search Console data: clicks, impressions and positions by page and query; indexing reports; the Manual actions report.
- The live search results for your lost queries, including AI answers and who they cite.
- Advertisers’ cost per click for the same queries, to price the loss.
What can’t we see from outside?
- Small sites. Third-party ranking data can’t see sites with fewer than roughly 1,000 US visits a month. In our forum test, six of ten sites were too small, or outside the US, to see. We say so instead of guessing.
- Estimates. Outside traffic figures are estimates. The Autopsy replaces them with your Search Console data.
- Gaps in the archive. The Wayback Machine doesn’t capture every page. We test the old URLs it has, and the count says how many.
- Platform switches. We can’t yet reliably detect a platform change from outside. That’s why the Cause Check form asks what changed.
- Inside Google. Nobody outside Google can see why a specific page was re-ranked. We can show what changed, where, and when.
- Chat assistants. Clicks lost to chat assistants leave no record in Search Console.
- Overlaps. When a redesign and an update hit the same pages at the same time, the split isn’t always clean. The verdict says when it isn’t.
The method applied
We checked ten public threads from Google’s help forum this way, from the outside, where owners had asked for help after a traffic drop or a redesign. Six sites were too small, or outside the US, to see. In two of the four we could see, the update date alone would have blamed Google for something the site did. Four cases is a small sample, and people who post in forums tend to have confusing cases. It was enough to fix the order of our steps.
The four case files. The method, applied to your site: the Autopsy. The three suspects: your redesign, Google and AI.
FAQ
Why three business days?
The questions are narrow and the evidence is structured: what changed, when, and which pages and queries lost the clicks. Three business days is enough to answer that properly.
Can we run these checks ourselves?
Many of them, yes. View the source of a key page for “noindex”, open an old URL and see where it lands, compare your drop date with Google’s update calendar. The Autopsy does all of it at once, against your Search Console data, and prices the result.