SCT / USA: ready
The April 3, 2025 Correction

How to Fix Your GSC Impressions in 2026

Your CTR did not collapse. Google corrected an under-counting bug. Here is how to re-baseline, isolate the artifact, and stop reporting fake decline.

April 3
2025 Cut Date
-30%
Typical Apparent CTR Drop
0
Actual Traffic Lost

The summary, in one paragraph

On April 3, 2025, Google fixed a known under-counting bug in how AI Overview and SGE impressions were logged in Search Console. Sites that did nothing wrong saw impressions step up sharply while clicks held flat — making CTR look like it had collapsed by 20-40% overnight. The traffic did not change. The denominator did. If you have been reporting CTR decline year-over-year against a pre-April-3 baseline, your report is wrong.

The 4-step re-baseline

Step 1

Cut the timeline

Treat April 3, 2025 as a hard cut. Build two CTR baselines: pre-cut and post-cut. Do not compare across the boundary without normalizing.

Step 2

Confirm clicks are flat

If your raw click count and analytics sessions are steady across the date boundary, you saw a measurement correction, not a traffic loss.

Step 3

Re-anchor expected CTR

The post-fix CTR curve is roughly 30% lower at every position. Update internal CTR targets and dashboards to the new curve.

Step 4

Isolate AI Overview share

The new impressions are mostly AI Overview placements. Separate them from blue-link impressions so the rest of your CTR analysis stays clean.

What Search Console Tools does automatically

  • Auto-detects the April 3, 2025 step-change in your data and labels it on every chart.
  • Applies pre/post baselines to CTR analysis so striking-distance scoring is not skewed by the artifact.
  • Flags pages where the AI Overview impression share is now >50% so you can deprioritize those queries.
New Guide

The 3-Part Data Integrity Audit

Day 1: The Impression Correction. Day 2: AI Answer-to-Click Ratio. Day 3: The 5 strike-distance queries one title tag away from page 1 on your site.

No spam. 1-click unsubscribe anytime.