HCU & Core update recovery: the field playbook

Since the Helpful Content Update waves began, we've taken +35 sites through full recoveries — including sites that had lost 70%+ of their Discover distribution and were considered dead by their previous agencies. Recovery is not luck and it's not waiting. It's a sequence. Here is the playbook, phase by phase.

Phase 0 — Triage: is it actually the update?

Half the "HCU victims" we diagnose weren't hit by the HCU at all. Traffic drops get misattributed constantly: a quality-tier demotion, a technical regression after a redesign, a cluster that simply went cold. Before any recovery work, we run the drop against our monitoring data — 75,000+ analysis points — to date it precisely and match it to a known update window. Wrong diagnosis, wasted year. Our 60-second triage tool gives you a first read.

Phase 1 — Forensic diagnosis (weeks 1–2)

Full-corpus classification: every URL scored on engagement history, topical fit, E-E-A-T signals, and whether it earned its traffic or rode a trend. The output is a map of what the site actually is in Google's eyes — usually very different from what the publisher believes it is.

Phase 2 — Pruning & consolidation (weeks 2–6)

• Prune the corpus segments that drag the sitewide classifier down — thin, unengaged, off-cluster content. Noindex or 410, case by case.
• Consolidate overlapping articles into fewer, stronger URLs with proper redirects.
• Reinforce author signals, sourcing, and first-hand experience markers on what remains.

Phase 3 — Signal reconstruction (weeks 4–12)

A pruned site is a clean slate, not a recovered one. Now the sitewide signals need positive evidence: new content published on the strongest clusters only, engineered engagement waves on each release (this is where Étincelle™ and a well-segmented push base earn their keep), technical layer brought to reference standard.

Phase 4 — Re-entry (months 3–6)

Classifier reassessments are periodic, tied to Core update cycles. When the reassessment comes, a site that spent months accumulating clean engagement evidence flips back — in our experience typically within 3 to 6 months, sometimes above its pre-update baseline. One women's magazine we handled went from −70% to +115% of baseline in four months. The GSC curves are here.

What doesn't work

Waiting ("it'll come back" — it won't on its own). Publishing more of the same content that caused the demotion. Mass "AI cleanup" rewrites that change words but not signals. And switching domains — you're abandoning your history, not your classifier.

Recovery is the hardest discipline in Discover work and the one where experience compounds fastest. If your curve flatlined, it can restart — but every month of waiting extends Phase 4.