Google June 2026 spam update

Google’s June 2026 spam update (June 24-26) applied global pressure across all languages via SpamBrain, running two days after Google’s May 15 policy extension that brought AI answer manipulation within the spam enforcement boundary. Google has not confirmed the two are directly linked. In the same window, Google launched dedicated AI impression reports inside Search Console, UK-first, giving practitioners their first native visibility into AI Overviews and AI Mode performance. Critical gap: impressions only, no click data. The Munich Regional Court issued a preliminary ruling on May 28 establishing Google as directly liable for false AI Overview claims, treating generated summaries as Google’s own speech rather than third-party aggregation. SE Ranking’s 2026 referral study puts AI search at 0.32% of total web traffic, a figure 134 times smaller than organic. The measurement infrastructure is arriving before the traffic justifies panic or the data justifies certainty.

What Happened This Week

Four distinct signals arrived in the same seven-day window, and the pattern is worth reading carefully.

Google rolled out a global spam update with timing that aligns with its May policy revision extending enforcement to AI answer manipulation, though Google has not confirmed the connection. Google launched the first dedicated visibility reporting for AI search surfaces inside Search Console, UK-first, with publisher opt-out controls attached. A German court issued a preliminary ruling that AI Overviews are Google’s own speech and that Google is directly liable for false claims they contain. And the publisher licensing market tightened further, with Microsoft closing its first APAC content deal while Perplexity absorbed two separate lawsuits from publishers.

Read each of these in isolation and you get four separate news items. Read them together and you get a structural shift: the AI search ecosystem is formalising. Enforcement, measurement, legal accountability, and commercial negotiation are replacing the frontier period where tactics went untested and everything was provisional. The informal era is ending faster than most practitioners have prepared for.

The Findings

What did the June 2026 spam update actually change?

Google’s June 2026 spam update ran from June 24 at 9:00 AM PT to June 26 at approximately 11:00 AM PT, a total of two days and one hour confirmed via the Google Search Status Dashboard. It applied globally across all languages and is the second spam update of 2026, following the March update which completed in under 20 hours.

What Google confirmed. The update is a tuning of SpamBrain, Google’s AI-based spam prevention engine, which has been operational since 2022. Barry Schwartz at Search Engine Roundtable verified directly with Google that the update does not target link spam, site reputation abuse, or several other specific policy categories. No companion blog post was published and no new spam policies were announced with this rollout.

What is not confirmed. Industry commentary has widely inferred that this update enforces Google’s May 15, 2026 policy revision, which extended the definition of spam to include attempts to manipulate generative AI responses in Google Search. The timing is consistent with that inference. Google has not confirmed it. Specific tactics are being discussed in the community as probable targets: recommendation poisoning (engineering content so AI models treat a domain as authoritative) and biased listicles structured to influence AI citations. These are community-reported observations, not confirmed algorithmic targets. Treat the AI-manipulation connection as a working hypothesis, not a confirmed fact.

Recovery. Google’s documentation is explicit that automated systems may require months to reassess a domain’s compliance status after a spam update. There is no faster path. If your site saw ranking drops in the June 24 to 26 window, attribution is complicated by overlap with the May 2026 core update, which completed June 2. Annotate June 24 in Search Console and do a clean 14-day comparison before and after before drawing conclusions.

Practitioner implication: Do not restructure until you have separated spam-update movement from core update residue. Both updates ran close enough together to make mid-June volatility genuinely ambiguous. If you are running scaled, templated content or systematic brand mention acquisition of any kind, audit it now regardless of whether this update touched you. The May 15 policy gives Google explicit framework to act on AI-manipulation tactics in any future rollout.

Google can now measure your AI search visibility, but only halfway

On June 3, 2026, Google launched dedicated Generative AI performance reports inside Search Console, confirmed via the Google Search Central Blog. The announcement was authored by Hillel Maoz (Search Ecosystem Engineering Manager) and Moshe Samet (Product Manager Lead, Search Console). The rollout is UK-first to a subset of site owners. No global expansion date has been confirmed.

What the reports show. Impressions: how often a URL from your site appeared inside a generative AI feature on Google Search. Broken down by pages, countries, devices, and dates, with hourly to monthly granularity. Report data begins from May 18, 2026.

What the reports do not show. Clicks. CTR. Average position. Query-level data. This is the most important thing to understand before you use this data. You can see that your content appeared inside AI Overviews or AI Mode. You cannot determine whether that appearance drove any traffic. Google has said it will add metrics over time. No timeline has been confirmed. These AI impressions were always counted in your existing Search Console totals. This launch creates a dedicated view of data you already had, not new traffic.

Alongside the reports, Google launched a publisher opt-out toggle within Search Console settings, governing whether a domain’s content is used to generate responses across AI Overviews, AI Mode, and generative AI features in Discover. The toggle went live June 3 for the UK cohort under a binding requirement from the UK’s Competition and Markets Authority. It took effect from June 17, 2026.

The toggle does not extend to the standalone Gemini application. Opting out via Search Console has no effect on how your content surfaces in Gemini responses. Data from a Seer Interactive longitudinal study (published April 2026, tracking 2.43 billion organic impressions across 53 brands from January 2025 to February 2026) found that pages cited in AI Overviews receive 120% more organic clicks per impression than uncited pages on the same results pages, though both groups trail queries where no AI Overview is present at all, where click rates run 38% higher. Making an opt-out decision without click attribution data from the new GSC report means making a high-stakes call blind.

On June 24, Google confirmed a data logging bug affecting the Discover performance report and the Generative AI in Discover report. Annotate this date in any reporting dashboard to avoid misreading the anomaly as a traffic or engagement drop.

Practitioner implication: If you have access to the UK cohort, baseline your AI impressions now. Do not make the opt-out decision until click data is available. You do not have enough information to justify removing yourself from surfaces where being cited measurably outperforms not being cited. The one exception is if you have a specific regulatory or editorial reason to withhold content that outweighs the visibility trade-off.

A German court ruled that Google is directly liable for false AI answers

On May 28, 2026, the Regional Court of Munich issued a preliminary injunction (case 26 O 869/26) prohibiting Google from repeating false statements about two Munich-based publishers through its AI Overviews. The ruling was first reported by The Decoder on June 9 and has since been confirmed via the published decision text.

The case centred on AI-generated search summaries that linked the publishers to scams, subscription traps, and dubious business practices. Google’s AI had combined information about other companies with the plaintiffs and generated connections that appeared in none of the cited sources.

The court’s reasoning. AI Overviews evaluate, rewrite, and combine source material into new substantive statements, unlike traditional search results which direct users to external sources. Standard German case law shielding search engines from direct liability does not apply when the platform is generating original synthesis. Google’s defence that users could verify information by checking cited sources was rejected. The court found that an AI Overview presents a self-contained answer, users read it without clicking through, and the ability to disprove a claim does not remove liability for making it.

What this ruling is not. A final judgment. A binding European precedent. A ruling that settles AI liability globally. This is a regional court preliminary injunction in Germany. Google has said it is reviewing the decision. It can be appealed.

What it may become. The court’s reasoning that AI synthesis differs legally from search result aggregation is coherent and has been applied in a second German case since. If it survives appeal, the same logic reaches every AI answer engine that generates novel responses from source material, not just Google. The Munich court explicitly noted the ruling may have international reach.

Separately, in the US, independent musicians filed Kogon v. Google alleging Google trained its Lyria 3 and ProducerAI models on YouTube’s Content ID database without permission. Google moved to dismiss on June 9, 2026, arguing YouTube’s Terms of Service grant sufficient usage licences. Both proceedings are in preliminary stages and no financial determinations have been made.

Practitioner implication: The immediate operational implication for most practitioners is limited. This is a legal and publishing risk story, not an optimisation change. The medium-term signal is worth tracking: if AI platforms face direct liability for hallucinated claims about businesses, the commercial pressure to improve factual accuracy in AI answers increases. That is a different quality signal to anything practitioners currently optimise for. Watch whether this ruling influences how Google handles AI Overview corrections at scale.

Google updated recipe display in AI Mode: what is confirmed and what is probable

On June 30, 2026, Robby Stein, VP of Product for Google Search, announced an update to recipe display in Google’s AI Mode conversational interface.

What is confirmed. Visual link cards now appear at the top of the AI Mode response for relevant recipe queries, showing the creator name, star ratings, ingredient count, and a visual thumbnail. This replaces the previous March 2026 interface that required a user to tap a dish name before publisher links became accessible. Rajan Patel at Google confirmed on X that Google is incorporating publisher feedback on this experience. Community reporting confirms the card treatment is live in AI Mode.

What is not confirmed. Whether Schema structured data is a requirement to qualify for the card layout. Google has not stated this. Whether the update applies globally or is limited by device, region, or language. Google has not confirmed scope.

On Schema as a requirement: balance of probabilities. Google has not confirmed Schema markup as a technical prerequisite for the new card layout. However, the data fields displayed (creator name, star ratings, ingredient count, thumbnail) map precisely to Schema.org properties: author, aggregateRating, recipeIngredient, and image. Google’s AI search surfaces have consistently rewarded structured, machine-readable content across every rich result format introduced since 2018. On balance, the probability that complete Recipe Schema improves eligibility for this card treatment is high. This should be treated as probable, not confirmed. Auditing and completing Recipe Schema carries minimal downside risk regardless of whether it directly determines card eligibility, because complete Schema improves eligibility for standard rich results independently of AI Mode.

Publisher response has been mixed. Creators welcomed the prominent link placement but noted that Google’s AI still generates a full recipe summary below the cards and presents it as authoritative, capturing the majority of user attention before any publisher link is reached. This cannibalization of the read is unresolved, and Google’s response has been to note they are listening to feedback rather than committing to a structural fix.

Practitioner implication: Audit Recipe Schema completeness now. Specifically: author (publisher byline, correctly formatted), aggregateRating (ratingValue and reviewCount, not empty), recipeIngredient (full structured ingredient list, indexable), and image (high-resolution, multiple aspect ratios). If you manage food, travel, or lifestyle content with recipe-adjacent intent, this is a low-cost audit with potential upside in both AI Mode card eligibility and standard rich results. You cannot resolve the cannibalization problem from your end. That is a platform decision.

Google is testing AI inside paid search placements, and advertisers have no opt-out yet

On July 1, 2026, Google confirmed it is running a limited test injecting AI-generated summaries directly beneath the descriptions of sponsored search results. The feature was first identified by digital marketer Darcy Burk. Google has confirmed the test is running but has not confirmed the generation mechanism or whether it draws from landing page copy, broader index data, or both. Google has also not confirmed whether advertisers will receive opt-out controls.

Each injected block carries a disclaimer: “Google AI responses are generated independently and can make mistakes, so double-check responses.”

On the same date, Google announced that AdSense ad intent units will display Gemini-generated content alongside related ads when a user interacts with a unit.

In both cases, Google’s systems are deciding what supplementary content appears alongside an advertiser’s paid placement without advertiser input or control.

The risk in regulated industries. Financial services, healthcare, legal, and other regulated categories have specific compliance requirements around how products and services can be described. If Google’s AI-generated summary misrepresents a product, which the disclaimer acknowledges is possible, the advertiser has no current mechanism to correct or suppress it. Google’s disclaimer places verification responsibility on the user, not accuracy responsibility on Google.

Practitioner implication: If you run paid search for a regulated category, begin manually monitoring your ad placements for AI-injected content now. Document any instance where the generated summary misrepresents your product or service. You will need that record for any compliance discussion. There is no opt-out confirmed. This is a watch item, not an action item, until scope and controls are announced.

CDN-level AI crawler controls are here, but compliance is still voluntary

On June 23, 2026, Cloudflare rolled out a managed robots.txt feature that dynamically generates directives for known AI crawlers, removing the need for publishers to manually write and maintain individual bot rules.

Concurrently, Cloudflare is testing a content-use signalling extension within robots.txt that defines four permission fields: search, ai-input, ai-train, and a content-reuse signal specifying what a crawler may retain after access.

The compliance ceiling. Cloudflare acknowledges robots.txt is voluntary. There is no enforcement mechanism beyond potential legal exposure under copyright or computer access law. Publisher executives reported to Digiday that Perplexity’s crawlers actively bypass robots.txt directives. Perplexity denies this. The practical value of the Cloudflare feature depends entirely on whether the crawlers you most want to control choose to respect it.

The categorisation that matters most for practitioners is the distinction between training crawlers and retrieval crawlers. Model training crawlers (GPTBot, ClaudeBot) consume content with no traffic return. AI search and retrieval crawlers (OAI-SearchBot, PerplexityBot) scrape live content to generate answers with citation links back to the publisher. These two categories should not be treated identically. Blocking training crawlers selectively protects proprietary content without affecting AI search visibility, while blocking retrieval crawlers removes that visibility entirely.

Practitioner implication: If you have not reviewed your robots.txt for AI crawler exposure, do it now. Use Cloudflare’s managed feature if you are already on Cloudflare, as it reduces maintenance overhead. Before writing any blocking rules, classify your target crawlers by purpose: training vs. retrieval vs. core search indexers. Blocking Googlebot or OAI-SearchBot has direct consequences for organic and AI search visibility that most sites cannot absorb.

Competitor Moves

OpenAI

The high-profile December 2025 Disney licensing deal, which granted Sora access to more than 200 Disney, Marvel, Pixar, and Star Wars character IPs alongside a $1 billion equity investment, was cancelled in March 2026 when OpenAI shut down Sora. The $1 billion investment was never consummated. This matters for practitioners as a signal: what appeared to be the most significant IP licensing deal in AI video generation collapsed when the underlying platform was discontinued. Long-term strategic bets on individual AI platforms carry platform risk.

OpenAI did confirm a multi-year global display licensing agreement with Getty Images, announced June 21, 2026, allowing ChatGPT to surface credited, attributed visual assets directly within answers. This is a live deal. The Getty agreement covers retrieval and display, not generative use of the IP, which distinguishes it structurally from the Sora arrangement.

On June 26, OpenAI simplified model picker controls for ChatGPT Enterprise and Edu users, replacing previous Thinking mode labels with Instant, Medium, High, Extra High, Pro Standard, and Pro Extended tiers (confirmed via OpenAI Help Center release notes).

On July 2, the Financial Times reported (unconfirmed) that CEO Sam Altman proposed offering the US government a 5% equity stake in OpenAI to reduce political pressure. OpenAI has not confirmed this. Treat it as unverified.

Perplexity

CNN filed a lawsuit against Perplexity in the Southern District of New York on May 28, 2026, alleging unauthorised copying of over 17,000 articles, real-time verbatim reproduction via retrieval-augmented generation, and wilful infringement for ignoring a December 2025 cease-and-desist letter.

A separate Houston-based wiretapping class action alleges Perplexity’s tracking technology intercepted users’ questions and responses and shared them with Meta and Google without consent, including in incognito mode. A preliminary settlement range of $5 million to $150 million was reported by a secondary source. The sourcing methodology for that figure is not credible and the dollar range should be treated as unverified.

Perplexity’s Publisher Program, launched January 2026, had enrolled over 2,400 partner publishers by Q1 2026. Revenue sharing is reported at $8 to $15 per 1,000 citations with up to a 50% quality multiplier for high entity density content. These figures come from a secondary source (Digital Strategy Force) and have not been confirmed by Perplexity directly.

Microsoft

On July 3, 2026, Microsoft and Nine Entertainment announced a content-licensing agreement giving Microsoft Copilot access to Nine’s journalism beyond paywalled previews. The agreement covers The Australian Financial Review, The Sydney Morning Herald, The Age, Brisbane Times, and WAToday. Copilot will display snippets, headlines, and summaries with links directing users back to Nine’s publications. No financial terms were disclosed by either company. Secondary reporting suggests a figure under $25 million based on the absence of an ASX disclosure requirement. Treat that as inference, not confirmed data.

This is Microsoft’s first AI content deal with a major Australian publisher and its first in the APAC region. For practitioners managing content in Australian financial, business, or news verticals, Nine’s publications carry significant topical authority. This deal increases the probability that Copilot draws on Nine’s content when answering queries in those categories, which has implications for citation strategy.

On June 16, 2026, Bing Webmaster Tools globally released four new AI Performance metrics in preview: Intents, Topics, Citation Share, and Compare. A surge in citation data from June 1 was confirmed by Microsoft’s Krishna Madhavan as a routine data backfill artifact, not an organic shift.

What This Means For You

The week’s events are not four separate stories. They are one structural shift viewed from four angles.

Enforcement is real. The May 15 policy revision gave Google explicit framework to act on AI manipulation tactics. The June spam update applied pressure. Targets remain unconfirmed, but the timing and community data are consistent. The Munich ruling adds legal accountability alongside algorithmic enforcement. The combination is compressing the space for tactical manoeuvring that relied on the absence of rules.

Measurement arrived before the money did. GSC AI Performance reports are a genuine step forward and the most-requested Search Console feature the SEO community has wanted since AI Overviews launched. But impressions without clicks means you can see whether you are cited without knowing whether it matters commercially. Do not make the opt-out decision on impression data alone.

The publisher licensing architecture is being built now. The Nine-Microsoft deal is the first in APAC. Getty-OpenAI covers attributed visual display. Perplexity is absorbing lawsuits from publishers it has not paid. The commercial terms being negotiated in 2026 will set the template for how AI search platforms compensate content creators for the next several years. For publishers and brands that produce original content with genuine authority, being early and deliberate about what you will and will not license matters.

The data on AI search as a traffic channel is sobering and should recalibrate expectations. SE Ranking’s 2026 study (tracking referral traffic across a multi-million domain dataset from 2024 to 2026) found AI search engines drove 0.32% of total website traffic, up from 0.02% in 2024, a 16x increase, but still 134 times smaller than organic search. ChatGPT drives 74.78% of that AI referral share. AI search is growing fast from a very small base. Anyone representing AI search as a replacement for organic traffic in 2026 is overclaiming.

What to Watch Next

Google Search Console Generative AI Performance Reports: In limited rollout. UK-first cohort only. Global expansion unconfirmed. The missing metric to watch for is click data.

UK CMA publisher opt-out toggle: In limited rollout. UK-only. Global expansion unconfirmed. If your site is in the UK cohort, the June 17 effective date has passed and the setting is already active.

Google AI-generated summaries in sponsored search results: In testing. No confirmed geographic scope, no confirmed opt-out mechanism for advertisers.

Google AI Overviews web-only toggle: In testing. Allows users to strip all generative features and revert to standard blue-link results. Wide adoption would affect AI Overview impression volumes across the board.

Google AI Mode autocomplete in Ask Anything: In testing. Autocomplete and AI-powered suggestions in the AI Mode follow-up prompt box.

Google AI Search Information Agents: Announced, not yet live. Persistent background agents monitoring the web for user-defined tasks. Scheduled for Google AI Pro and Ultra subscribers in Summer 2026.