Google Information Agents Confirmed for Home Services: What They Read

When someone sets up a Google Information Agent to find an HVAC contractor, no query lands in your analytics. No referral fires. No impression appears in Search Console. An AI process reads your service pages -- at any hour, with no human present -- and either puts you on a shortlist or doesn't.

We've been tracking this surface since Google announced Information Agents at I/O 2026. This week, for the first time, we have practitioner-level confirmation that home services are a supported use case. Three findings from our August 2026 research change the infrastructure picture.

What information agents do (and when they do it)

Google Information Agents -- called Gemini Spark inside the Gemini app and Workspace -- are persistent background processes that monitor web content on standing topics and push synthesized briefings to subscribers. A user sets a task: "find me a reliable roofer in my area." The agent monitors continuously, and when it identifies candidates, it notifies the subscriber with source links.

In our August 7, 2026 research (Scout session 100, `knowledge/platform-citation-behaviors.md`), we documented the first practitioner-level characterization of how these agents handle home services queries. From marketingcode.com's Google I/O 2026 coverage: "Gemini Information Agents allow users to set up a background agent that scans for a roofer, an HVAC tech, an electrician -- whatever -- and notifies them when the right shop shows up."

AI SEO practitioner Chris Raulf put the business reality plainly: "When Gemini Spark evaluates [local businesses] at two in the morning while I am asleep, the agent reads what is machine-readable, decides what matters, and either includes you in the shortlist it hands its human, or it does not."

The citation mechanism is the same as any other AI surface -- a source link in a push notification is a referral visit. What's different is that no user types a query, and no query time appears in your logs.

The subscription gate dropped

Our July 31, 2026 investigation of Information Agent audit scope (`methodology-recs/2026-07-31-google-information-agents-audit-scope.md`, Scout session 93) identified the primary reason not to prioritize this surface: access was gated behind Google's $100/month AI Ultra subscription. The consumer population was too small to matter.

That changed. In our August 5, 2026 research (Scout session 98), we confirmed that Google expanded Information Agents to AI Pro subscribers in the US on July 25, 2026. AI Pro is $20/month. Two independent sources -- techmymoney.com and androidheadlines.com -- confirmed the expansion date; the feature remains excluded from EEA, Switzerland, UK, and Nigeria.

A $20/month tool reaches a meaningfully different population than a $100/month one. Professionals, homeowners with home management tools, and property managers -- the same demographic that books contractors, hires renovators, and signs service agreements -- are now the addressable user base. This surface moved from theoretical to actively relevant between July and August 2026.

Why it won't appear in GA4

Here's what makes Information Agents structurally different from every other citation surface we audit: when an agent reads your site during its background monitoring cycle, no session fires in GA4.

Our August 6, 2026 research (Scout session 99, `knowledge/platform-citation-behaviors.md`) documented two independent practitioner sources confirming this. SEOHandbook.co.uk on Gemini Spark: "For publishers the significance is that Spark monitors topics and compiles digests from web content, so a site can be read, synthesised, and acted on with no click and no signal in GA4, Search Console, or server logs." Digital Applied: "Information agents don't generate sessions -- when an agent monitors a topic for weeks and fires a notification that converts, there is no UTM, no session."

This is a different problem from the AI traffic attribution gap we already track -- where AI-driven sessions exist but arrive as Direct/(none) in GA4. The Information Agent read phase generates no session at all. You can be evaluated, shortlisted, and handed to a subscriber without anything appearing in your analytics until the moment they tap a source link.

GA4's native AI Assistant channel, launched May 13, 2026, classifies traffic from ChatGPT, Gemini, DeepSeek, Copilot, and Grok under the `ai-assistant` medium. It recognizes gemini.google.com referrals from direct browser clicks. It does not cover Information Agent push-notification clicks, which originate from a notification delivery mechanism rather than a gemini.google.com URL. As of August 7, 2026, no practitioner has published confirmed data on what referral string (if any) an Information Agent click carries.

The practical implication: you cannot optimize your way to a measurable Information Agent citation count. What you can do is make sure the agents that evaluate your business find your infrastructure readable.

What agents actually read

When an Information Agent evaluates a local contractor, it reads what is machine-readable. Chris Raulf's framing from our August 7 research: "Your site does not need to impress an agent; it needs to be parseable by one. Clear service information, clean schema, and consistent business details that match your Google Business Profile -- with real pages for real services in real locations."

From SEOHandbook.co.uk's characterization of Gemini Spark's local business behavior: "Local-business citation runs through Google Business and Maps, not generic editorial." This is the same infrastructure that drives standard AI Mode citations. GBP completeness -- business categories, service areas, hours, service descriptions -- is the primary input. Schema on your service pages reinforces the GBP signal; it doesn't replace it.

There's also a citability hierarchy worth understanding. Our session 99 research documented Position.digital's finding that Information Agents have "a significantly higher bar for citability" than standard AI Mode responses. A business not appearing in standard AI Mode responses is, per their characterization, extremely unlikely to appear in Information Agent briefings. The fix for Information Agent visibility runs through standard AI Mode first -- which means the infrastructure work is the same in both cases.

The three things that need to be in order

None of this requires a new optimization track. The infrastructure that makes a contractor visible to Information Agents is the same infrastructure that moves standard AI Mode scores:

**Google Business Profile completeness.** Every service category selected, service area defined, hours current, service descriptions populated. GBP is the primary data source agents pull from for local business evaluation -- both confirmed in this research and consistent with what Sourcepull audits show across AI Mode generally.

**Parseable service pages.** Real pages for real services in real locations. H2 headings that use the language a homeowner would search. Entity data -- business name, address, phone, service categories -- in accessible HTML rather than JavaScript-rendered markup. A service page an agent can't parse is a page that doesn't exist in its evaluation.

**Consistent entity signals.** Business name, category, and address identical across GBP, your website's LocalBusiness schema, and every directory where you're listed. An inconsistent entity -- different name strings across Yelp, BBB, and your schema -- creates disambiguation ambiguity that causes agents to default to more clearly established competitors.

These are the same gaps Signal Check audits surface: how AI systems currently describe your business, which platforms cite you, and where the entity infrastructure is broken. If you want to understand where you stand before the next information agent evaluation runs, that's the starting point.