When organic traffic falls hard, teams often jump straight to “we were hit.” That phrase hides two different problems. One is a human reviewer applying a manual action. The other is an automated system reducing visibility because the site, or parts of it, no longer meet quality, relevance, or spam thresholds. Treating them as the same thing wastes months.
This guide is a diagnostic sequence, not a pep talk. You will gather evidence, classify the drop, and choose a recovery path that matches the classification. If you need a structured findings register for this work, a penalty and recovery audit is built for exactly that separation.
Start in Search Console before you invent a story
Open Search Console and check Manual actions and Security issues first. A manual action is explicit: spammy free hosts, unnatural links to or from your site, thin content with little value, user-generated spam, cloaking, or similar named reasons. If one is listed, that is your primary diagnosis. Document the date it appeared, the scope (sitewide or partial), and the exact wording.
If Manual actions is clean, you still do not assume “algorithm update” as a vibe. You keep looking. Algorithmic suppression is a working label for a sustained visibility loss that lines up with quality systems, spam systems, or core updates, without a human message in the console. The label only becomes useful when evidence supports it.
- Export the Manual actions history if anything was ever resolved or still open.
- Note Security issues separately; malware and hacking need a different playbook.
- Capture screenshots and dates so later stakeholders cannot rewrite the timeline.
- Check whether the property covers the right hostname and protocol before you conclude “nothing is listed.”
Read the traffic shape, not just the percentage drop
Manual actions often produce a sharp, durable cliff once the action is applied, especially when sitewide. Partial actions can look like a section disappearing while other templates hold. Algorithmic effects vary: some arrive as a cliff on a known update date, others as a grind over weeks as rankings for commercial queries weaken while brand queries hold.
Compare Google organic sessions against non-Google channels, direct traffic, and brand query demand. If every channel fell after a product outage or pricing change, search systems may not be the villain. If only Google organic fell, and Search Console impressions fell with it, you are in search-visibility territory.
| Signal | More consistent with manual action | More consistent with algorithmic suppression | Needs more evidence |
|---|---|---|---|
| Search Console Manual actions | Named action present | Clean | Wrong property verified |
| Drop timing | Near action date or reconsideration denial | Near core/spam update windows or gradual decay | Deploy, migration, or tracking change same week |
| Scope | Matches action scope (site or section) | Hits thin, scaled, or low-trust templates hardest | One URL lost a featured snippet |
| Brand queries | Can fall if sitewide trust is damaged | Often hold while non-brand commercial terms fall | Brand campaign or PR shock |
| Recovery after cleanup | Needs successful reconsideration for the action | Needs sustained quality change and recrawl | Cleanup incomplete or wrong cause fixed |
Build the evidence trail for links, content, and tech
Classification without cause is theatre. Once you know whether a manual action exists, map the likely cause classes: unnatural links, scaled low-value content, scraped or doorway patterns, hacked injections, or technical accidents that look like cloaking or sneaky redirects.
Link evidence
Export referring domains from at least two sources and look for bursts, identical anchors, sitewide footer schemes, foreign-language spam on unrelated domains, and networks that repeat across many properties. Do not rush to disavow everything unfamiliar. Wrong disavows remove neutral or helpful links and distract from the real problem. Pair this work with a careful backlink audit reading guide before you upload any file.
Content evidence
Sample the templates that lost impressions. Are they affiliate doorway pages, auto-translated shells, thin location pages with the same paragraph, AI-scaled posts with no first-hand detail, or useful pages that simply aged? Algorithmic suppression often concentrates on the weakest cluster while strong product or docs pages survive. Manual thin-content actions tend to name the problem more directly in the console.
Technical evidence
Rule out self-inflicted cliffs: robots.txt blocks, accidental noindex on money templates, canonical loops after a migration, hreflang pointing at the wrong locale, or a JavaScript render failure that left Google with empty shells. These are not “penalties,” but they mimic them in analytics. A technical SEO audit is the right lens when the drop lines up with a release.
Recovery diagnosis worksheet (fill before choosing tactics)
Property: ______________________________
Drop start date (analytics): ____________
Drop start date (Search Console): _______
Manual action present? Y/N Reason: _____
Scope: sitewide / partial / URL pattern: _
Deploy or migration within 14 days? Y/N _
Templates hit hardest: _________________
Brand vs non-brand change: _____________
Link anomalies (dates, anchors): _______
Content clusters implicated: ___________
Technical accidents ruled out: _________
Working classification: ________________
Primary fix track this month: __________
Reconsideration needed? Y/N ____________Choose a recovery path that matches the classification
If a manual action is present, cleanup must address the named reason, then you request reconsideration with a factual summary of what was wrong, what you removed or fixed, and how you will prevent recurrence. Vague promises fail. Incomplete cleanup fails. Re-submitting weekly without new work fails.
If there is no manual action and evidence points to algorithmic suppression, there is no reconsideration button that magically restores rankings. You improve the underlying quality issues, remove or rewrite the scaled junk, repair link risk where it is real, and give systems time and crawl cycles to reassess. Expect measurement in weeks to months, not overnight screenshots.
- Freeze new risky acquisition tactics (purchased links, scaled doorway publishing) while diagnosis is open.
- Fix technical accidents that could explain the cliff before you rewrite the whole content library.
- For manual link actions: remove what you control, document outreach, disavow only residual risk you can justify.
- For content quality problems: prune or rewrite the worst clusters; keep pages that still deserve a job.
- Set monitoring: Manual actions panel, impressions by template, and a short list of head terms.
- Write a stop-rule: what evidence would make you revise the classification.
What “recovered” should mean in plain language
For a manual action, recovered means the action is cleared in Search Console and the affected URLs are eligible again. Rankings may still take time to rebuild. For algorithmic suppression, recovered means the affected templates regain impressions and clicks toward a pre-damage baseline, adjusted for seasonality and demand, without a console trophy to screenshot.
Do not declare victory because one branded query bounced. Do not declare permanent failure because week two is quiet. Recovery is a measured trend, not a mood. Share the classification and the evidence with whoever owns content, links, and engineering so the next “quick win campaign” does not recreate the same cause.
Common misreads that keep teams stuck
- Calling every drop a penalty when Manual actions is empty.
- Disavowing thousands of domains because a spreadsheet sorted by “spam score.”
- Deleting valuable pages that lost traffic for seasonal or SERP-feature reasons.
- Ignoring a migration or robots change that landed the same week as a core update.
- Mixing recovery cleanup with unrelated redesign work so you cannot tell what helped.
If you want the wider process context for how findings get sequenced after diagnosis, see how to prioritise SEO audit findings. Recovery work often jumps the normal hygiene queue, but it still needs clear tickets and owners.
The useful question is not “Were we penalised?” It is “What evidence classifies this drop, and what fix path matches that class?”