Google announced three updates to Search and AI Mode this week, Roger Montti reported for SEJ. Reading his article prompted me to examine these updates, the broader pattern, and their implications for this year’s search.
If we look at this in detail, it seems that the updates are pushing most of what was happening on the results page to the completion of the task.
Announced by Google
Google has launched individual hotel price tracking in Search, now available worldwide for signed-in users searching in English and Spanish. Email alerts notify users of rate changes within selected dates.
Additionally, in March, Canvas travel planning with AI Mode went from Labs preview to general US availability, allowing users to define trips and find custom itineraries with flights, hotels, and attractions that automatically save. Agent-powered store calling, first introduced in classic search, will soon be released in AI mode, which will allow Google’s AI to call nearby stores, check inventory, using Gemini and Duplex models.
Rose Yao, Product Lead at Search, posted updates on X. More details are available in Google’s blog post.
The pattern
These updates reflect Google’s product direction seen in research, patents, and executive statements since January.
In January, Google published a SAGE research paper on agents training four-step reasoning chains, laying the foundation for multi-step operations in search.
Photosi’s April interview made the language public. Photosi said, “Most of the queries will be Search Agents.” Our dive followed how his language changed from “search will change” to specific definitions of task completion.
Earlier this month, Montti argued that task-based agent search is already changing SEO, citing Google’s global rollout of restaurant reservations as proof that the future in Photosi’s language was overdue for the product.
Last week, the US Patent Office published a continuation patent for Google titled “Automatically providing search results after the fact” (our story). Installation describes a system that waits for answers if they are not immediately available, and then delivers them later through the interaction of the assistant.
These updates continue in the same way. Canvas is moving from Labs preview to wider US availability, roughly five months after its initial launch in November. Shoplifting has been introduced with AI Mode following its debut in Search last November. Additionally, hotel price tracking is now available in Search at the single location level.
Microsoft’s latest news fits a similar pattern. Sumit Chauhan, President of Microsoft’s Office Product Group, wrote on the company’s blog that Copilot’s agent capabilities are now available in Word, Excel, and PowerPoint:
“Copilot creates far more value when doing work — formatting, reordering, creating visualizations, and transforming data — than just suggesting steps.”
Default features for Microsoft 365 Copilot and Premium subscribers, and available in Personal and Family plans. It’s unclear whether businesses will get the same reporting in agent-driven environments, a point not addressed in Microsoft’s post.
Vocabulary Not yet strong
Google uses “agent” in its product language and announcements, describing features like calling and AI Mode as task-oriented. The SeatGeek partnership was called the “Google Agetic AI Search Experience.” Other companies also use the same agent framework language.
Photosi describes the ‘agent manager’ as Google’s role and envisions a future where Search becomes an ‘agent manager’ that manages various tasks. It positions Google as an orchestration layer over agents rather than a direct competitor.
Montti used “task-based search” in his recent SEJ coverage, sometimes shortened to TBAS. That’s his shorthand for the beat, not industry standard terms.
“Agentic” describes ability. “Agent manager” refers to a specific architectural role required by Google. “Activity-based” focuses on the user’s intent. When three different labels appear in one month, the market is still working out what to call this.
Why This Matters to Search Professionals
The features introduced this week are changing the definition of visibility in several business segments.
Local retailers are now experiencing a new discovery environment. When a store calls through AI Mode, Google agents, instead of users, will contact businesses to confirm stock and details. Google did not disclose which stores its agents will contact first, how eligibility is determined, or if specific business information influences the process.
An analysis of 68 million AI crawler visits to 858,457 sites managed by Duda shows that sites linked to Yext, Google Business Profile, and review systems were indexed more often than those without. These findings describe browser behavior, not agent calls. It is not known if the same signals influence which stores are contacted.
Hotels and tourism businesses are now faced with the rental price of each property. Itineraries are based on the Canvas selection concept. There is no report showing if a hotel appeared in a Canvas program, triggered an alert, or was named in an AI Mode response.
Publishers are facing constant pressure from AI-driven shortening. The Index Exchange analyzed 1,200 publishers on its exchange, and found that 69% had a year-over-year decline in ad opportunities, for an average decline of 14%.
Rejection varies across specifics. Health and careers publishers saw a 40-50% drop in ads, while news and politics publishers saw a 7% drop.
Vanessa Otero, Founder and CEO of Ad Fontes Media, told Index Exchange on the same episode:
“When it is important enough that you want accurate and complete information about a major international, national, or local event, a high-quality news site is still a better experience than asking an AI chatbot, which may give a generalized or inaccurate answer. AI users already know this, which is why most news consumers still go directly to their trusted sites. News has always worked well for advertisers, and if this site will be strong, this site will have a strong role, and if this the site will have the power to be critical to the open web of the future.”
Travel publishers are under pressure as Canvas covers itineraries without citing sources, making it impossible for publishers to know if their coverage is influencing the travel plans.
Ecommerce sellers can’t see where the stores are called, so they can’t determine if the inventory, listing accuracy, or Google business profile signals are working.
Multi-platform installation complicates the strategy. Google agents like structured data and verified profiles. Perplexity Computer routes all 19 models with various retrieval preferences. ChatGPT Atlas scrapes browser content directly. The OpenAI Operator uses a GUI view to interact with the rendered pages.
A single business has multiple acquisition channels with different technical requirements. Developing a single strategy no longer covers all areas.
What is not seen yet
As our coverage has flagged the measurement gap, it has grown.
Search professionals still can’t see if their business has been included in the Canvas itinerary. They cannot see if the agent called them. They cannot see if their hotel has been exposed to a price tracking alert. And they can’t see how often their content has been used to compile someone else’s travel plan.
No new reporting areas have been posted alongside the updates. Alphabet reported $63.1 billion in Google Search and Other advertising revenue for Q4 2025, up 17% year-over-year, crediting executives with search and cloud acceleration and the benefits of using AI. No new reporting tools have emerged to help businesses track their role in AI search.
The pattern holds across fields. ChatGPT transmission data is limited to what OpenAI shares. The appearance of a confused quote is within Perplexity. Google agent locations do not include a clean map of the Search Console.
Academic research on agent training continues to develop. Two April 2026 papers on arXiv show the pace. CW-GRPO, from Junzhe Wang and colleagues, proposes reinforcement learning optimization for adaptive search agents. SKILL0, developed by Zhengxi Lu and colleagues at Zhejiang University, trains agents to internalize skill packages. The result is agents that operate without the overhead of an instruction during guesswork.
The training pipeline is developing faster than the pipeline businesses depend on. Search professionals cannot bridge that gap alone. Google, OpenAI, Perplexity, and Anthropic will all need to provide equivalent agent reporting. No one has publicly committed to doing so.
Looking Forward
Photosi said 2027 will be “an important turning point.” He mentioned non-engineering workflows and specific agent business processes. Our installation went along that timeline.
May brings Google I/O and Microsoft Build. Both companies are likely to expand their agent locations at those events, making the report a more urgent thing to watch. If businesses can’t see their role in job-based search, they can’t promote it or argue over who should pay for it.
Two lingering questions sit behind that. Pay per click is activated when users click on links. Store calls, canvassing, and price tracking do not generate clicks, and no platform has defined replacements. Schema.org is designed for search engine visibility, not for agents who need real-time inventory, booking availability, and action endpoints. Agent-readable enterprise data standards have also not been met.
What happens next depends on whether or not the platform builds reporting capabilities. So far, no one has explained how it will happen. Until that changes, businesses will be preparing for places they can’t see. The next signs are coming to I/O and Build in three weeks.
