agentificial
case study · anonymised
·Diagnosis· Jul – Sep 2026

Diagnosing an 80% organic-traffic collapse from 17 months of Search Console data

A medical-tourism group's website lost four-fifths of its monthly search clicks. We found the cause, retracted two of our own repair recommendations, and set up weekly monitoring. Traffic has not recovered yet.

−80.8%
monthly clicks: 4,082 (May 2025) to 782 (Aug 2026)
35,504
clicks over 17 months, from 5,753,802 impressions
34
clicks in 90 days for the page a model said drove 40%

The problem

Search clicks had fallen steeply and the cause was unclear. Early readings pointed in different directions: a URL migration in which 95 old URLs mapped to 57 targets (24.9% recovery), and a third-party model estimating that a single article drove 40% of traffic.

Each explanation implied a different fix, so the first job was to find out which one the data supported.

What we built

  • An archive of the full Search Console export: 17 months of clicks and impressions by date, page and query.
  • Analysis that checks clicks, not only impressions. That check is what overturned two of our own repair recommendations.
  • A weekly regression monitor with dated runs on 31 August and 7, 14, 21 and 28 September 2026, flagging page-level cliffs and decays.
  • Technical repairs: five service pages that had returned HTTP 500 for ten months (a corrupted line from a bulk find-and-replace) and 16 malformed internal links across 15 posts.
  • A gated content pipeline with word-count, citation and E-E-A-T checks. 107 articles were prepared (13 pillar, 94 supporting).

Result

  • Monthly clicks fell from 4,082 (May 2025, the peak) to 782 (August 2026): −80.8%. Across the 17 months: 35,504 clicks and 5,753,802 impressions.
  • On 1 September 2026 the cause was resolved as algorithmic (the May 2026 core update and June 2026 spam update), not the URL migration.
  • The model's "40% of traffic" article earned 34 clicks in 90 days in Search Console. What the data did show: queries naming individual people made up 60.9% of non-branded clicks.
  • A baseline reporting zero non-branded rankings was an artefact of its seed keyword list. Measured data showed 429 ranked keywords, 24 of them non-branded in the top 20.
  • The 28 September 2026 monitor run flagged 2 page cliffs and 96 page decays.

Scope and limits

  • Diagnosis and monitoring, not growth. The trend was still down at the last monitor run and we claim no traffic recovery.
  • The 107 articles are unpublished, held pending a strategy decision.
  • The client's Search Console data is not shared beyond these aggregates.
· Could be you next

Same problem,
different stack.

Tell us the workflow that should already be done by the time your team logs in. We'll come back with a one-page assessment.