Google Search updates 2025-2026: what changed and what to actually do

Google ran six broad core updates and three spam updates between March 2025 and June 2026, retired rich results for FAQ, HowTo, and several other schema types, removed AMP from its documentation entirely on 1 July 2026, and shipped a wave of new AI fetchers. None of it changes the fundamentals. This page tracks the dated events and separates what moved rankings from what only makes your templates look stale.

The update calendar

Every entry below is a confirmed rollout. Core updates reassess the whole ranking system; spam updates run through SpamBrain across all languages and target on-page and content spam, not links.

Date	Type	Notes
Mar 13-27, 2025	Core	Standard broad core update
Jun 30-Jul 17, 2025	Core	~17-day rollout
Aug 26-Sep 22, 2025	Spam	SpamBrain, all languages
Dec 11-29, 2025	Core
Feb 2026	Core (Discover only)	Scoped to Discover
Mar 24-25, 2026	Spam	Sub-20-hour rollout
Mar 27-Apr 10, 2026	Core	Very volatile: 79.5% of top-3 positions moved
May 21-Jun 2, 2026	Core	Large, typical core update; YMYL hit hard
Jun 24-26, 2026	Spam	On-page/content spam, not link spam

Two things are worth internalizing. First, the March 2026 core update was unusually violent — roughly four in five top-three positions shifted — so if you saw movement then, it was systemic, not a penalty. Second, the two 2026 spam updates explicitly did not target link spam or enforce site-reputation abuse. They hit scaled content, cloaking, doorways, sneaky redirects, scraped content, and hidden text. If your traffic dropped during a spam update, look at content production practices, not your backlink profile.

Recovery from a core update does not come from a quick fix. It comes with a later update, after genuine quality improvement. There is no toggle that reverses a core-update loss.

Structured-data deprecations

Google has been steadily removing rich-result features. The markup is harmless to leave in place — it may still feed entity understanding — but it earns nothing in the SERP, so stop investing effort in it and treat it as a signal of stale templates.

• FAQPage — rich results effectively removed in May 2026 (stopped showing 7 May 2026); the Search Console report and Rich Results Test dropped it in June 2026, with API support ending August 2026. Only authoritative government and health sites still see FAQ results. It had already been restricted to large publishers since August 2023.
• HowTo — removed before 2025; zero effect today.
• Sitelinks Searchbox — removed 21 November 2024.
• Practice Problems (Quiz) — removed January 2026.
• Course Info, Estimated Salary, Learning Video, Special Announcements — removed September 2025.
• Book Actions and Vehicle Listing — deprecated June 2025.
• AMP — the AMP Viewer, AMP Cache, and Signed Exchange references were removed from Google's documentation on 1 July 2026. <link rel="amphtml"> and AMP page markup are now legacy.

Not everything is shrinking. Google added Loyalty Program markup (June 2025), Merchant priceType and validForMemberTier (February 2025), hasAdultConsideration (May 2026), and Merchant category / CategoryCode (July 2026). The direction is clear: commerce and merchant markup is expanding while informational rich results are being culled. For the full picture of what still earns results, see the structured data guide.

New fetchers and bot verification

Google shipped a run of new user agents you should recognize when reading server logs or writing robots rules:

• Google-NotebookLM (October 2025), Google-Pinpoint (November 2025), Google-CWS (November 2025), Google-Messages (January 2026).
• Google-Agent (March 2026) — AI agents browsing the web on a user's behalf. It is user-triggered and behaves like a browser, using the identity agent.bot.goog.
• Web Bot Auth (experimental, May 2026) — an IETF-draft cryptographic scheme for verifying bot identity, a response to the ease of spoofing user-agent strings.

The AI bot crawling policy guide covers how to decide which of these to allow.

What the crawler documentation clarified

Two 2026 documentation updates matter for how you build pages. The March 2026 "Inside Googlebot" post confirmed that the Web Rendering Service fetches every HTML-referenced resource — excluding media, fonts, and exotic types — and that each resource has its own per-URL byte counter that does not count against the parent page. The WRS cache TTL runs up to 30 days and ignores your HTTP cache headers. This is the concrete reason blocking CSS or JavaScript in robots.txt hurts: Google needs those resources to render and understand the page. Crawlinx flags this with robots.blocks_css_js.

Google also confirmed (December 2025) that JavaScript executes even on non-200 status codes, clarified canonicalization for JS-rendered content, and reiterated that dynamic rendering is officially deprecated as a workaround — use server-side rendering, static generation, or hydration. The JavaScript SEO guide has the detail.

The AI-optimization guide says the quiet part loud

Google's AI Optimization guide, updated 10 July 2026, is the clearest anti-hype source available. Google states plainly that these do nothing for AI Overviews or AI Mode: llms.txt, content chunking, special AI markup, and AEO-specific structured data. What works, per Google, is unremarkable: original content with a unique point of view, standard technical SEO that keeps pages crawlable and indexable, clear semantic HTML, high-quality multimedia, and accurate structured data. The independent evidence agrees — the strongest measurable correlate of AI citation is off-site brand mentions, not any on-page trick. The GEO and AEO guide covers this honestly.

At Google I/O in May 2026, AI Overviews and AI Mode were unified worldwide, Gemini 3.5 Flash became the default AI Mode model, and a redesigned multimodal search box and Search Agents (arriving summer 2026) were announced. A June 2026 opt-out toggle (UK first) lets you exclude content from AI features without affecting organic rank, and Search Console gained Generative AI performance reports — impressions only, no clicks yet.

What to actually do

Nothing on this page requires a special tactic. Keep pages indexable (index.noindex), stop blocking CSS and JavaScript, remove dead schema from templates as you touch them (Crawlinx flags schema.faq_deprecated, amp.deprecated, and schema.deprecated_type), and build genuine quality rather than chasing a recovery that only a future update can grant. The updates change the scenery. The work is the same.