All issues
Discovery Digest · June 8, 2026

Issue 02. The week search infrastructure rewired for agents

Five platforms moved this week, all betting on the same thing: AI agents, not humans, are the next primary consumers of search. Apple rebuilt Siri on Google Gemini, Google shipped the first native AI search visibility data, Perplexity let agents write their own search code, Microsoft launched Bing-grounded evidence APIs for agent workloads, and xAI put first-place image-to-video generation into the API. Here is what each update means for brands navigating the discovery layer.

Issue 02. The week search infrastructure rewired for agents
01 · AI Search

Apple rebuilds Siri on Google Gemini at WWDC 2026

What
Apple unveiled a rebuilt Siri at WWDC 2026, powered by a custom version of Google Gemini and launched as a dedicated standalone app across iOS 27, iPadOS 27, and macOS 27. The rebuilt assistant replaces Spotlight as the system-wide search surface, adds persistent chat history synced via iCloud, personal-context access to emails, photos, and files, and on-screen awareness. A developer extension system opens Siri to third-party AI integrations.
When
Announced at the WWDC 2026 keynote, Cupertino, Mon 8 June 2026.
How it shifts discovery
The distribution implications are substantial. Brands that have been building AI search visibility for Google's AI Overviews and AI Mode are now also building for the same underlying model on a completely different surface, one with personal context, calendar access, and cross-app actions that Google's own search interface does not yet have. Brand entity clarity on the Gemini model is now a requirement for Apple discoverability, not just Google discoverability.
Questions to ask
  • Which of our product categories is Siri most likely to be queried about by iPhone users, and are we visible on the underlying Gemini model for those prompts?
  • How does our brand entity read across Apple's personal-context layer and Google's web index, given that Gemini-powered Siri now relies on both?
  • Are our existing AI Mode and AI Overview visibility strategies extensible to the Siri surface, or do we need a separate audit before iOS 27 ships?
Sources
  • Apple. WWDC 2026 keynote, 8 June 2026
  • MacRumors. WWDC 2026: Gemini-Powered Siri, iOS 27, macOS 27 and More, 8 June 2026
02 · Google Search

Google adds AI visibility reports to Search Console with opt-out controls

What
Google launched dedicated Search Generative AI performance reports in Search Console on 3 June 2026, giving site owners data on impressions inside AI Overviews and AI Mode across Search and Discover. Reports cover URLs, countries, devices, and date ranges. A separate opt-out toggle, rolling out in parallel, lets publishers remove their content from AI Overviews and AI Mode without affecting organic rankings.
When
Announced on the Google Search Central blog, Wed 3 June 2026, currently rolling out to a subset of websites in the UK first.
How it shifts discovery
Search practitioners have been managing AI Overview visibility with no first-party data since AI Overviews launched at scale. The reports make AI search impressions measurable inside the same tool used for organic performance, removing the most common blocker to building a coherent AI search measurement framework. The opt-out toggle is the more consequential announcement for content strategy: every publisher will have to make an explicit, evidence-based policy decision on whether to remain in AI surfaces.
Questions to ask
  • What share of our impressions are coming from AI features rather than standard organic results, and is that proportion growing month on month?
  • For which pages is appearing in AI Overviews or AI Mode net-positive for assisted attribution vs. a net-negative for direct click volume?
  • Does our team have an explicit policy on whether to opt out of AI surfaces, and is that decision owned by search, content, or commercial?
Sources
03 · AI Search

Perplexity ships Search as Code: AI agents now write their own search routines

What
Perplexity launched Search as Code on 7 June 2026, replacing fixed search API calls with AI-generated Python workflows where agents write bespoke search logic for each retrieval task rather than calling a standardised endpoint. The architecture is now the default inside Perplexity's Computer agent platform and available via the Agent API. Perplexity reports the approach outperforms OpenAI and Anthropic on key retrieval benchmarks while reducing token costs by up to 85 per cent.
When
Published on the Perplexity research blog, Sun 7 June 2026.
How it shifts discovery
If retrieval is increasingly handled by AI agents writing custom search code rather than humans entering queries, the optimisation target for brands changes: being the most retrievable, verifiable, parseable source when a model constructs its own search logic replaces being the best answer to a human query. The criteria an agent applies when writing a search routine (trustworthiness, factual density, source structure) may differ substantially from the criteria that surface content for human searchers. Brands that do not understand how they perform in agent-driven retrieval are operating on assumptions.
Questions to ask
  • How does our content perform in agent-driven retrieval tasks vs. standard search, and do we have any data on that split today?
  • What signals do AI agents use to select and rank sources when constructing bespoke search code, and are we optimising for those signals?
  • Is our search strategy team tracking the shift from human-query optimisation to agent-retrieval optimisation as a distinct workstream?
Sources
04 · AI Search

Microsoft launches Web IQ at Build 2026: search infrastructure built for AI agents

What
Microsoft announced Web IQ at Build 2026 on 2 June, a suite of AI-native grounding APIs that connects AI agents to fresh web intelligence from pages, news, images, and video. Where Bing returns full pages, Web IQ returns structured evidence objects: extracted passages, ranked relevance, and packaged intelligence at sub-165 millisecond response times, nearly 2.5 times faster than comparable alternatives. The system is model-agnostic and MCP-native.
When
Announced at Microsoft Build 2026, San Francisco, Tue 2 June 2026.
How it shifts discovery
Web IQ is the clearest statement yet that Microsoft views Bing's primary role as infrastructure for the agent era rather than a consumer search surface. The shift from returning pages to returning structured evidence objects rewrites the requirement for publisher content: being indexable is table stakes, being returnable as a verifiable evidence object is the new threshold. Brands that publish dense, attributable, citation-worthy content will outperform brands relying on broad topical coverage when agent workloads use Bing as their grounding layer.
Questions to ask
  • Is our content structured to be returned as a verifiable evidence object, or is it readable only at full-page length by a human?
  • Do we know how our brand's factual claims are represented in AI agent reasoning workloads that use Bing as a grounding layer?
  • As agent traffic through Bing-grounded systems grows, are we tracking it as a distinct channel in our attribution model?
Sources
05 · xAI

xAI releases Grok Imagine Video 1.5 via API, takes first place in image-to-video benchmarks

What
xAI released Grok Imagine Video 1.5 as an API preview via api.x.ai on 3 June 2026. The model converts still images into cinematic video using natural-language motion prompts, supports up to 720p, and preserves the look, lighting, and atmosphere of the source image. On release it entered first place in the Artificial Analysis image-to-video arena; pricing is $0.08 per second at 480p and $0.14 per second at 720p.
When
Released as an API preview via x.ai, Wed 3 June 2026.
How it shifts discovery
First-place image-to-video generation available via a per-second API signals that cinematic video from brand images is now a commodity workflow, not a specialist production. When agentic workflows can animate any still image into a video, the source image becomes a discoverability and brand-control problem. The quality, licensing status, and on-brand accuracy of brand images used as source material will directly determine what AI-generated brand video looks like across every platform that integrates this API.
Questions to ask
  • Does our brand have a licensed, on-brand image library ready for use as source material in AI-driven video generation workflows?
  • How does our visual brand identity translate when animated from still images: is it recognisable, consistent, and on-strategy?
  • Are we monitoring AI-generated brand video for accuracy and alignment, and do we have a policy for when it is wrong?
Sources

Key takeaways

What to walk away with this week

  1. Apple's distribution plus Google's model means the Gemini AI surface reaches every iPhone user this autumn. Audit your Gemini visibility and add it to your measurement stack before iOS 27 ships.

  2. Google's AI visibility reports are the first native data tool for AI Overview and AI Mode impressions. Establish a baseline now and build an evidence-based opt-out policy before the rollout expands beyond the UK.

  3. Perplexity's Search as Code and Microsoft's Web IQ are both built for AI agents, not human users. Structured, extractable, citation-worthy content is no longer the differentiator: it is the threshold requirement for agent-driven retrieval.

  4. First-place image-to-video generation is now an API commodity. Build a licensed brand image library before agentic workflows start animating your brand from source images you do not control.

  5. Five platforms rewired for the agent era in one week. Teams adjusting measurement and content strategy now are a full cycle ahead of teams waiting for consensus.

Previous issue

Issue 02. Measuring the AI Search Era

Read issue

The Discovery Digest · Every Friday

Stay ahead of AI Search

Five updates a week across ChatGPT, Claude, Gemini, Perplexity, Copilot, Grok and Google AI Overviews, with the questions worth asking.

Free5 updates weeklyUnsubscribe anytime