Skip to content

Google I/O 2026 Search Changes: A Local Business Guide

Google's 2026 I/O event introduced a redesigned search box and new AI search guidance. For local businesses, the practical question is not whether AI search is coming but which fundamentals to strengthen now and which agent-readiness work to start preparing. This guide summarises the announced changes and the expert commentary published alongside them, with a decision checklist, a worked example and a field map for AI booking agents.

What Google announced

Two announcements matter most for local search. First, Google described a new AI-powered Search experience as its biggest shift in search in 25 years. Second, Google published updated guidance on how websites can succeed in AI-driven results, which gave marketers clearer direction on what matters and what does not.

The redesigned search box, which Google calls the Intelligent Search Box, expands as you type so you can ask longer, more conversational questions. It accepts text, images, files, video and open Chrome tabs. Suggestions go beyond autocomplete, and you can move from a standard search into a fuller AI conversation. Google also described standing agents that keep watching on your behalf, such as alerting you when a followed band announces a nearby show.

For users, this means less guessing at keywords and more describing what they want. For businesses, it reshapes the queries you need to appear in and likely means fewer but higher-intent visits rather than many low-intent clicks.

The fundamentals still carry the weight

The updated AI search guidance reinforces that the foundations of SEO have not changed. What changes is how you execute them for AI Overviews and AI Mode. The guidance also clears away some noise: special text files such as llms.txt are not required to appear in AI search (they do no harm and are simple to add, but they are not the point), and chopping content into tiny chunks for AI to read is more myth than value.

What matters is what always has: genuinely useful content, a clean and well-structured site, and real depth on what you do, where you do it and how good you are at it. An accurate, active Google Business Profile and a steady flow of reviews remain central.

One caveat worth keeping in view: Google's own guidance and its Chrome tools did not speak with one voice on llms.txt in the same month, which is a sign of how new this area still is. Treat the guidance as direction, not a settled rulebook.

The genuinely new part: AI agents that book

The guidance includes a short section on AI agents that act on behalf of a user, such as booking a table or comparing products, and on emerging standards such as the Universal Commerce Protocol that will let them do more. Google treats this as early and optional. Myles Anderson, quoted in the published commentary, argues for giving it more weight: if an agent cannot read your prices in detail, see your availability and book into your calendar, it cannot complete the job for the customer and will move on to a site that can. That is where the highest-intent demand is heading.

This is the part of the announcement that most changes what a local business needs to publish. It is not about adding a special file; it is about making the commercial facts an agent needs machine-readable and current.

How the published expert commentary frames the shift

The commentary published alongside the announcements, quoting named SEO practitioners, converges on a few points and diverges on others.

On the convergence side, Rachel Ellen of Croud argues that local search is moving from keyword matching toward understanding context, intent and reputation. A 4.8-star rating tells Google people like a business; AI search is increasingly able to interpret why they like it. Customer photos, video, reviews and forum discussions combine into a richer picture than ratings alone. Success therefore depends less on optimising a single profile and more on a consistent reputation wherever customers share experiences.

On the divergence side, David Mihm of Near Media reads the event as a short-term impact of minimal significance for local search, with bigger effects coming from UX and SERP feature changes tied to monetisation, and with Local Pack evolution toward an AI Overview-style carousel likely to be tested heavily first. Celeste Gonzalez of Lastmile Retail instead expects the expansion of personal intelligence to be a real shakeup once fully rolled out in regular results, pushing businesses toward more holistic, ideal-customer-focused content and stronger evidence of who they are and what they offer.

Mihm also cautions against taking Google's mythbusting of page chunking and schema.org at face value, noting that Bing published near-simultaneous guidance highlighting chunking and historically encouraging schema, and that other model providers' methods are not fully known. These are disagreements in emphasis, not contradictions of the announced facts.

Worked example: a cafe's rating versus its signals

Take a cafe with a 4.8-star rating. The rating alone tells an AI that people like the place. Rachel Ellen's commentary illustrates how richer signals change the picture: a customer photo might show dogs relaxing inside, a video might reveal a calm atmosphere suited to remote working, reviews might mention people travelling across town to visit, and forum discussions might recommend the oat milk lattes.

Combining those signals lets AI describe the cafe in terms a rating cannot: dog-friendly, work-friendly, worth a trip, known for a specific drink. The practical implication is that a business should treat reviews, forums, communities and social discussions as sources of customer insight, not just marketing channels. Identify recurring themes and act on them; those themes are what AI systems use to understand and describe the business.

Decision checklist: what to do now versus what to prepare

Use this checklist to separate work that pays off now from work that prepares for the agent era.

Do now:

  • Publish genuinely useful, in-depth content about what you do, where you do it and how good you are at it.
  • Keep the site clean and well-structured.
  • Keep the Google Business Profile accurate and active, with a steady flow of reviews.
  • Monitor reviews, forums, communities and social discussions for recurring themes, and act on them.
  • Track brand searches as an AI search visibility indicator, since answer volatility and format variation make AI rankings problematic.

Prepare next:

  • Make prices, availability and booking details readable by an AI agent, not just by a human.
  • Decide whether to add optional files such as llms.txt; they do no harm but are not the mechanism.
  • Do not spend effort chopping content into tiny chunks on the assumption that AI requires it.
  • Watch emerging standards such as the Universal Commerce Protocol, which Google treats as early and optional.

Skip or deprioritise:

  • Treating llms.txt or chunking as the lever for AI visibility.
  • Optimising a single profile in isolation from the wider reputation picture.

Field map: what an AI booking agent needs to read

The published commentary names the commercial facts an agent needs to complete a booking. Use this map to check whether your site exposes them.

  • Prices: readable in detail, not only as a headline range.
  • Availability: current and specific enough for the agent to judge whether a slot exists.
  • Calendar: bookable into, so the agent can complete the reservation rather than hand it back.
  • Products and services: what you offer, stated clearly enough for comparison.
  • Location: where each offering is available.
  • Differentiators: what distinguishes you from competitors, so the agent can justify a recommendation.

If any of these is missing or stale, the agent cannot finish the job and will move to a site that can.

A useful failure distinction

Two failure modes look similar but need different fixes.

If your business is not described accurately in AI answers, the likely cause is weak or inconsistent reputation signals: thin content, an inactive profile, few reviews, or themes in reviews and forums that contradict your positioning. The fix is fundamentals and reputation consistency.

If an AI agent cannot complete a booking, the likely cause is missing machine-readable commercial data: prices, availability or calendar access. The fix is agent-readiness, not more content.

Confusing the two leads to wasted effort, such as adding llms.txt when the real gap is an unreadable price list.

Frequently asked questions

Do I need llms.txt to appear in AI search?

No. The published guidance says special text files such as llms.txt are not required to show up. They do no harm and are simple to add, but they are not the point. Note that Google's own Chrome tools began checking for llms.txt in the same month, so the signals are mixed; treat it as optional.

Should I chop my content into small chunks for AI to read?

The published commentary calls this more myth than value. What matters is genuinely useful content, a clean and well-structured site, and real depth on what you do, where you do it and how good you are at it. David Mihm cautions that Bing published near-simultaneous guidance highlighting chunking, so the picture across providers is not settled.

What is the Intelligent Search Box and why does it matter for local businesses?

It is Google's redesigned search box, described as its biggest change to the box in 25 years. It expands as you type, accepts text, images, files, video and open Chrome tabs, and lets you move from a standard search into a fuller AI conversation. For businesses, it reshapes the queries you need to appear in and likely means fewer but higher-intent visits.

What does an AI agent need to book a customer for my business?

The published commentary names prices readable in detail, current availability, and calendar access so the agent can complete the booking. It also points to clear product and service descriptions, location information and differentiators, so the agent can compare and recommend. If any of these is missing, the agent will move to a site that can complete the job.

How should I measure visibility in AI search?

David Mihm suggests tracking brand searches as an AI search visibility KPI, because answer volatility and format variation make AI rankings problematic. He also notes that Local AI Overviews are starting to recommend specific businesses, whether or not in a Google Business Profile carousel format.

Will the Local Pack change into an AI Overview?

David Mihm expects the Local Pack to evolve toward something closer to an AI Overview or summarised carousel in the mid-term future, roughly the next year, but expects Google to test heavily before a broader rollout. This is a forecast, not an announced change.