Google Algorithm History

Every modern Google ranking decision is the residue of fifteen years of algorithm wars. If you cannot name what each update was punishing, you cannot diagnose why your traffic dropped today. The history is the diagnostic toolkit.

TL;DR

• The shape of Google’s algorithm is layered, not replaced. Panda’s content-quality classifier still runs underneath the Helpful Content system; Penguin still evaluates link spam inside the core. Updates accumulate; they do not retire.
• Named updates ended in 2022 for spam, 2023 for HCU. Since the March 2024 core update, almost everything is folded into broad core updates and unannounced spam systems. Naming is dead; the signals are not.
• Recovery requires identifying the era of damage. A site demoted by HCU in September 2023 needs different surgery than a site hit by the March 2025 core update or a 2024 site-reputation-abuse spam action.

The mental model

Google’s algorithm is like geology. Each major update is a sedimentary layer that never gets removed; it gets buried under newer layers and continues to exert pressure from below.

When you read about “the March 2024 core update merging the Helpful Content System into core,” that is not the HCU disappearing. It is HCU’s classifier becoming part of the bedrock — running every day, without an announcement, against every page in the index.

This is why old advice still works and old penalties still bite. A thin-content farm built in 2014 is still being demoted by Panda-derived signals in 2026, even though Google has not said the word “Panda” since 2016. The classifier was integrated, not retired.

The corollary: when you debug a traffic drop, you are doing geological survey work. Which era of signals failed? Was it content quality (Panda lineage), link profile (Penguin lineage), query understanding (RankBrain/BERT/MUM lineage), or holistic site quality (HCU lineage)? Get the era right and the playbook is obvious.

Deep dive: the 2026 reality

The chronological timeline, with what each update actually punished and where its signals live today.

Year	Update	Target	Status in 2026
Feb 2011	Panda (US English)	Thin content, content farms, scraped duplicates	Folded into core (Jan 2016); still active as a content quality classifier
Apr 2012	Penguin	Manipulative link building, exact-match anchors, link networks	Real-time inside core since Sept 2016; still demotes spammy link patterns
Aug 2012	Pirate	DMCA-heavy domains	Still applied; updated periodically without announcement
Aug 2013	Hummingbird	Whole-query semantic understanding (vs. keyword matching)	Foundation of modern ranking; subsumed
Jul 2014	Pigeon	Local search distance, prominence, relevance	Still the local ranking spine
Apr 2015	Mobilegeddon	Non-mobile-friendly pages on mobile SERPs	Replaced by mobile-first indexing (default 2020)
Oct 2015	RankBrain	ML-based query interpretation for unseen queries	Always-on; evolved into Google’s neural matching
Sept 2016	Possum	Local diversification, filtering near-duplicates	Active in local pack
Mar 2017	Fred (unconfirmed)	Ad-heavy, low-value affiliate sites	Folded into broader spam/quality signals
Aug 2018	Medic core update	YMYL re-evaluation; trust signals weighted up	Established E-E-A-T as a frame
Oct 2019	BERT	Bidirectional NLP, prepositions and intent in long-tail	Always-on; foundation for natural-language queries
May 2020	May 2020 core update	Broad re-evaluation during COVID query shifts	Standard core
May 2021	MUM announced	Multimodal, cross-language understanding	Used in featured snippets, AIO, AI Mode
Jul 2021	Spam update July 2021	Doorway pages, scraped content	Series continues unannounced
Aug 2022	Helpful Content Update v1	Site-wide signal demoting “search-engine-first” content	Merged into core March 2024; classifier still runs
Dec 2022	December 2022 link spam update (SpamBrain)	AI-detected unnatural links	SpamBrain runs continuously
Apr 2023	Reviews update generalized	Insufficient first-hand product/service reviews	Folded into core
Sept 2023	HCU September 2023	Sledgehammer: small/independent sites lost 50–90%	Most cited recovery cohort still recovering or pivoting
Mar 2024	March 2024 core update	45-day rollout; HCU folded into core; “scaled content abuse” defined	Reduced “low-quality content” by ~40% per Google
May 2024	Site reputation abuse policy	Parasite SEO via authoritative domains	Manual + algorithmic enforcement ongoing
Aug 2024	August 2024 core update	Partial recovery cohort for HCU-affected sites	Standard
Nov 2024	November 2024 core update	Broad ranking reshuffle	Standard
Mar 2025	March 2025 core update	Re-weighting of forum/UGC quality signals	Standard
Jun 2025	June 2025 core update	First update with explicit AIO citation overlap signals	Standard
Mar 2026	March 2026 core update	Currently rolling out as of writing	In flight

A few cross-cutting systems deserve their own callouts:

Google Discover is its own ranking engine for the mobile feed. It runs on a different signal mix — visual engagement, freshness, personalization — and has its own update cadence. The December 2022 product reviews update and the September 2023 HCU both visibly hit Discover-heavy publishers, but Discover also has its own unannounced refreshes that move publisher traffic without core updates.

The Product Reviews Update ran as a named series from April 2021 through November 2023 (six numbered iterations). It targeted thin Amazon-affiliate content lacking original photography, hands-on testing, and comparison tables. It is now generalized into core as the Reviews System and applies to product, service, business, media, and place reviews.

Spam updates ran as named events from June 2021 through October 2023 (about eight major rollouts). Since 2024, SpamBrain runs continuously and updates are unannounced. The exception: large-scale enforcement waves like the May 2024 site-reputation-abuse announcement.

The Helpful Content System in its 2022–2023 form was a binary site-wide classifier. After the March 2024 core merge, it operates per-page and per-cluster, which is why some affected sites saw partial recoveries on specific URL clusters in August/November 2024.

Visualizing it

flowchart TD
  A["Query"] --> B["Hummingbird semantic parse 2013"]
  B --> C["RankBrain ML interpretation 2015"]
  C --> D["BERT natural language 2019"]
  D --> E["MUM multimodal 2021"]
  E --> F["Core ranking signals"]
  F --> G["Panda content quality classifier"]
  F --> H["Penguin link spam classifier"]
  F --> I["Helpful Content classifier"]
  F --> J["Reviews System"]
  F --> K["SpamBrain continuous"]
  G --> L["Final ranked results"]
  H --> L
  I --> L
  J --> L
  K --> L
  L --> M["AI Overviews layer 2024"]
  L --> N["AI Mode layer 2025"]

Bad vs. expert

The bad approach

Most SEOs treat algorithm history as trivia and “fix it” by aimlessly improving the site. Typical traffic-drop response:

Traffic dropped 38% on March 12.
Action plan:
- Add more content
- Build more backlinks
- Improve Core Web Vitals
- Update meta descriptions
- Run a content refresh

This fails because every action targets a different signal layer. If the cause was the March 2024 core merge of HCU, “more content” actively made things worse. If the cause was a December 2022 SpamBrain pass on a toxic backlink, Core Web Vitals are irrelevant. Generic site improvements waste 90% of the recovery budget.

The expert approach

Pin the date, identify the update, and target the specific signal class.

// algorithm-correlation.js — minimal traffic-drop diagnostic
const drops = [
  { startDate: '2024-03-05', endDate: '2024-04-19', updateId: 'mar-2024-core' },
  { startDate: '2024-08-15', endDate: '2024-09-03', updateId: 'aug-2024-core' },
  { startDate: '2024-11-11', endDate: '2024-12-05', updateId: 'nov-2024-core' },
  { startDate: '2025-03-13', endDate: '2025-03-27', updateId: 'mar-2025-core' },
  { startDate: '2025-06-30', endDate: '2025-07-16', updateId: 'jun-2025-core' },
];

function classifyDrop(siteDropDate, dropPercent, affectedClusters) {
  const match = drops.find(u =>
    siteDropDate >= u.startDate && siteDropDate <= u.endDate
  );
  if (!match) return 'unaligned-investigate-other-causes';
  if (match.updateId === 'mar-2024-core' && dropPercent > 30) {
    return 'hcu-merged-into-core: rewrite-or-prune-thin-pages';
  }
  if (affectedClusters.includes('product-reviews')) {
    return 'reviews-system: add-first-hand-evidence';
  }
  if (affectedClusters.includes('forum') || affectedClusters.includes('ugc')) {
    return 'site-reputation-abuse-or-ugc-quality';
  }
  return 'broad-core-quality-rebalance';
}

This works because it forces correlation before action. You diagnose the era, then apply the era-specific playbook documented in the next module.

Do this today

1. Open Google Search Console → Performance → Search results. Compare the last 16 months. Note every drop greater than 15% with its exact start date.
2. Cross-reference each drop date against Google’s official Search status dashboard at status.search.google.com/products/rGHU1u87FJnkP6W2GwMi/history. Flag matches within 7 days of any “Ranking” event.
3. For every confirmed match, classify which lineage applies — content (Panda/HCU), links (Penguin/SpamBrain), reviews, site-reputation-abuse, or query-understanding.
4. In Ahrefs or Semrush, run a “lost keywords” report for each drop window. Filter to URLs that lost positions; group by URL pattern (e.g., /blog/, /reviews/, /compare/) to find the affected cluster.
5. In GSC → Pages, look at indexed page count over the same window. A cliff in indexed pages signals a quality-classifier action; a stable count with declining clicks signals a ranking re-weighting.
6. Pull the URL list of the top 50 lost pages into Screaming Frog and crawl them. Tag each with one of: thin (<300 words), AI-without-edit, no-author, no-original-research, link-spam-target, parasite-on-subdomain.
7. Subscribe to the Google Search Status Dashboard RSS feed and Search Engine Roundtable’s update tracker so you have alerts for the next event.
8. Build a one-row-per-update spreadsheet for your site: update name, date, percent change, affected clusters, hypothesis, action taken, recovery outcome. This becomes your institutional memory.