This Week’s Search Roundup: AI moves into workflows as crawl, controls and testing tighten

This Week in One Sentence

Platforms are wiring AI directly into consumer and business workflows while formalising the controls (flags, toggles, directives) that determine when AI appears and what it can use. The operational reality is that SEO, local, and paid teams are being pushed towards governance, testing discipline, and resilience planning at the same time.

This Week’s Exec Summary

• AI is becoming a workflow layer (device assistants, business connectors), not just a search feature; governance and permissions are now part of SEO and local operations.
• Crawling and AI usage are converging: documentation that explicitly ties crawled data to AI increases the value of precise snippet directives and header controls.
• User control is a platform theme: experiments and previews show both pushes towards AI entry points and mechanisms to suppress AI results, complicating “one SERP” assumptions.
• Automation is being paired with more structured testing in ads; teams should be ready to supply and manage creative variants as experiment inputs.
• Reliability remains a constraint: outages are a practical reminder to design fallbacks when assistants sit inside production workflows.

Apple Intelligence meets web reality: model dependencies and rollout constraints

Apple’s AI announcements didn’t just introduce new experiences; they immediately raised questions about how “on-device” positioning intersects with external model claims, and what that means for availability by region and device. For marketers, the story is less about feature lists and more about the structural dependencies that may shape reach, consistency, and data flows.

Apple announces Apple Intelligence and new Siri AI; Google says it is powered by Gemini

Apple announced Apple Intelligence and a new AI-powered Siri with cross-app capabilities, presented as spanning devices and contexts. After the keynote, Google published a post stating Apple Intelligence is powered by Gemini, and included timing and regional limitations. Some details around restrictions and privacy implications were discussed as unclear or not fully explained.

• Apple unveiled Apple Intelligence and a new Siri AI at WWDC.
• Apple described “Apple Foundation Models” across devices including phones, watches, and Apple Vision Pro.
• Apple said the models handle image, voice, and text, with “personal context”, world knowledge, on-screen awareness, and orchestration across apps.
• An “agentic” feature was mentioned that can change compromised passwords and update the password manager.
• Google published a post after the keynote stating Apple Intelligence is powered by Gemini.
• Availability mentioned: beta now, public beta next month, and a fall release on new devices; EU/China limitations were reported but not fully explained.

Why it matters: Treat “powered by” and regional availability as operational constraints: monitor where experiences actually ship, keep messaging and support aligned to what users can access, and assume feature parity may differ by region where rollout is limited (as reported).

Applebot: crawling, indexing, and AI usage clarified

Apple also updated crawler documentation in ways that intersect directly with SEO fundamentals: how crawling is managed, how snippets are controlled, and how content may be used for AI purposes. The net effect is less ambiguity—paired with more responsibility for teams to align directives, headers, and server expectations with what the documentation now states.

Apple updates Applebot documentation to cover AI use and new crawling/indexing controls

Apple updated Applebot documentation to state that crawled data may be used for AI purposes and AI models. The docs also added or clarified multiple controls and behaviours, including no-snippet guidance and handling of X-Robots-Tag. Some statements (such as crawl delay handling) are particularly relevant for technical SEO and server management.

• Applebot documentation now says crawled data may be used for AI purposes and AI models.
• New or expanded sections cover no-snippet controls.
• The documentation adds crawl delay notes, including a statement that Applebot does not follow crawl delays because it is “really efficient”.
• A new section was added around X-Robots-Tag.
• The documentation includes notes around content being accessible for free.

Why it matters: Update your technical assumptions to match the documentation: re-check snippet controls and X-Robots-Tag usage across templates, and review crawl-management plans in light of the stated position on crawl delays.

AI in the browser: defaults, flags, and user control

Two updates pointed in the same direction: AI answers are being tested as a closer-to-default search entry point, while product work also acknowledges that some users want explicit ways to reduce or remove AI surfaces. For SEO measurement and forecasting, this pushes audience segmentation and SERP interpretation further away from “one results page” assumptions.

Chrome Canary flag routes address-bar searches to AI Mode; Google says it was an error and not planned as default

A Chrome Canary flag was observed sending address-bar searches directly to Google AI Mode. Later the same day, Google stated the behaviour was an error and said there are no plans to make AI Mode the default for Chrome searches. The report referenced a Friday timing, but did not specify date/timezone in the notes.

• A Chrome Canary flag was added that routed address-bar searches directly to AI Mode.
• The behaviour was reported publicly as observed in testing.
• Later that Friday (reported around 2:30 p.m.), it was stated the behaviour was an error.
• Google said there are no plans to make AI Mode the default for Chrome searches.

Why it matters: Track entry points into AI experiences as a distinct variable in performance analysis; even when described as an error, the test location (address bar) is a reminder that the funnel into AI answers can shift quickly.

Bing previews browser extension to toggle AI features and adds “-AI” query operator

Bing released a preview browser extension for Chrome and Edge that lets users toggle off AI features, such as Copilot answers in search results. It also introduced a “-AI” query option to avoid receiving an AI result. Both were described as a preview intended to gather feedback and guide broader integration.

• Bing released a preview extension for Chrome and Microsoft Edge.
• The extension lets users toggle off AI features like Copilot answers in results.
• Bing added the ability to append “-AI” to a query to avoid AI results.
• The changes were positioned as a preview to collect feedback from dedicated users.

Why it matters: Build reporting that can accommodate “AI-suppressed” behaviours in your audience: assume a portion of users will actively avoid AI answers and continue interacting with classic results patterns.

Gemini as an operational dependency: connected actions and reliability risk

This week showed both sides of assistant adoption: deeper integration into business tooling, and a reminder that outages can disrupt workflows when teams treat AI as an always-on layer. The practical response is governance plus resilience—without assuming that assistants behave like stable infrastructure.

Gemini introduces business connectors starting with Google Business Profiles

Google announced a new Gemini business feature that connects Gemini directly to business tools, beginning with Google Business Profiles. The connector is expected to roll out “in the coming weeks” (timing not specified beyond that). Once connected, Gemini can support common profile and management tasks using a command-style prompt.

• Google said Gemini will connect directly to tools businesses already use.
• The first supported connection is Google Business Profiles.
• Rollout timing was described only as “in the coming weeks”.
• Usage example: type “@Google business profiles” then make requests.
• Example tasks: set business hours, show insights, help with a food menu, and help with customer reviews.

Why it matters: Prepare local operations for AI-assisted edits by tightening permissions, workflows, and review checks; treat any assistant-made updates as changes that still require accuracy and brand compliance monitoring.

Gemini experiences outage and service is later restored

Gemini experienced an outage that affected many users, though not everyone. A company executive said restoration work was underway and later posted that service was back up and running. The incident underlines the fragility of AI dependencies when used as an operational layer.

• Gemini experienced an outage affecting a large number of users.
• Not all users were affected.
• A company VP posted that restoration work was in progress.
• Several hours later, it was stated that everything was back up and running.

Why it matters: Put fallbacks in place for AI-dependent workflows (drafting, reporting, support triage) so outages do not stall production, and document “manual mode” procedures for critical tasks.

Paid media platform changes: more controlled testing and shifting access rules

On the paid side, the pattern was practical rather than flashy: more formal experimentation tools where automation dominates, and policy-driven eligibility shifts in regulated verticals. The teams that win here are the ones that can test systematically and keep compliance tight across markets.

Google rolls out asset experiments for Performance Max campaigns

Google is rolling out new asset experiments for Performance Max campaigns. The feature is designed to test how different creative assets and asset groups affect results, using multiple comparison approaches. It includes comparisons between entirely new groups, incremental asset additions, and seasonal versus evergreen creative.

• Google is rolling out asset experiments for Performance Max.
• The goal is to test how creative assets impact performance.
• Advertisers can compare entirely new asset groups.
• Advertisers can measure the impact of adding individual assets.
• Advertisers can test seasonal creative against evergreen content.

Why it matters: Treat creative as a testable input, not a one-off deliverable: plan asset variants (copy, imagery, video) with clear hypotheses so experiments can run without slowing campaign iteration.

Microsoft expands Audience Ads eligibility for cryptocurrency exchanges in approved markets

Microsoft expanded eligibility for cryptocurrency exchanges to run Microsoft Audience Ads across approved crypto advertising markets. Eligibility depends on compliance with Microsoft’s cryptocurrency advertising policy and applicable local regulations, and requirements vary by market. This is a policy and access change rather than a creative format update.

• Microsoft expanded Audience Ads eligibility for cryptocurrency exchanges.
• The change applies across all approved crypto advertising markets.
• Advertisers must comply with Microsoft’s cryptocurrency and related products policy.
• Compliance with local regulations is required.
• Eligibility requirements and policy details vary by market.

Why it matters: Treat market-by-market eligibility as a planning input: re-check approvals before scaling, and keep landing pages and disclosures aligned with policy expectations to reduce preventable disruptions.

Local SEO: review replies, AI templating, and rejection patterns (as reported)

A data-heavy report suggests that moderation outcomes may correlate with detectable patterns in review replies—especially when language looks templated and replies are delayed. For teams using AI to scale responses, the operational requirement is to reduce repeatable footprints and tighten process discipline.

Report finds patterns in rejected Google review replies, including AI-templated language and timing

A report from GMBapi analysed 12,752 rejected Google review replies and identified patterns associated with removals. The report highlighted a high share of removals tied to five-star reviews, commonality of templated AI-style language, and long delays between the original review and the business reply. The report also described a sharp increase in rejections over time, peaking with a large jump in 2024.

• The report analysed 12,752 rejected review replies.
• 93% of removed replies were replies to five-star reviews.
• Many rejections were associated with AI-tool replies using templated enthusiastic language.
• Rejections were reported as low in 2022, increased in 2023, then jumped roughly 10x (or more) in 2024.
• The average rejected reply was written 1,221 hours (roughly 50 days) after the review was posted.

Why it matters: If you use AI for review replies, tighten tone guidelines, reduce boilerplate, and improve response timeliness to minimise patterns that appear automated (as reported) and may increase moderation risk.

Google Search Profiles: a new creator surface (reported thresholds)

Search appears to be experimenting with more profile-like spaces that aggregate creator content across platforms. The reported thresholds and observed beta insights hint at a potentially new visibility layer, but elements were described as varying by rollout and account.

Google Search Profiles described, including eligibility thresholds and beta insights

Google Search Profiles were described as shareable creator spaces that aggregate linked accounts and surface content across platforms. Reported eligibility thresholds were follower-based across major social networks, and a beta insights view was described as showing Search performance metrics. Some elements were observed in the wild and may vary by rollout and account.

• Search Profiles are described as a shareable space to highlight content across platforms in Search.
• Reported eligibility: 100,000 YouTube followers, 100,000 Instagram followers, 100,000 X followers, or 300,000 TikTok followers.
• An example profile path mentioned was profile.google.com/rustybrick.
• creators.google.com/@rustybrick was mentioned as a possible URL (reported/observed, not confirmed as general availability).
• Accounts can be linked so the profile aggregates content in one place.
• Prior knowledge panels may be translated into a search profile that can be claimed (reported as a possibility).

Why it matters: Monitor branded queries and entity visibility for signs of Search Profiles appearing; where they do, treat them as a new click path that could sit alongside traditional results and knowledge panels (as reported/observed).

What to watch next week

Key follow-ups and practical checks for the coming week:

• Whether Apple or Google clarifies the “powered by Gemini” claim and how it aligns with Apple’s on-device and privacy positioning (still partially unclear based on the notes).
• Any additional Applebot documentation edits or enforcement signals around no-snippet, X-Robots-Tag, and AI usage of crawled data.
• Rollout status of Gemini’s Google Business Profiles connector, including which accounts/regions get access first and what permissions it requires.
• Early advertiser feedback on Performance Max asset experiments: what testing options are actually available in accounts and how results are reported.