Google Connected Apps: The Google Base Moment for AI Search

Google has opened Connected Apps to developers, letting brands run their own services inside AI Mode and Gemini. Like Google Base twenty years ago, the early movers are likely to get outsized visibility. Here is where to start.

Chris Liversidge CEO of QueryClick & Corvidae By Chris Liversidge · 29th October, 2026 · SEO

A New Front Door Into Search

In July Google started letting US users connect third-party apps to AI Mode in Search. The launch partners were Instacart, Canva and YouTube Music: build a barbecue shopping list and drop it into your Instacart cart, ask Canva for flyer templates, have AI Mode curate a playlist and save it to YouTube Music, all without leaving the results page.

That was the demo. This week came the part that matters for everyone else. On 28 October Google opened a Connected Apps platform for developers in preview, with one integration that can reach the Gemini app, AI Mode in Search, or both. Adobe, Experian, Localiza and Peloton are already live in Gemini and are due in Search in the coming weeks.

In other words, the list of brands that can act inside Google's AI surfaces is no longer a list Google picks. You can apply to be on it.

Why This Feels Like Google Base

Some of us remember when Google Base opened up and merchants could push product data straight into Google. Most businesses spent the first couple of years deciding whether it mattered. The ones who uploaded early, and kept their data clean, picked up visibility far out of proportion to their size, because there was so little competition for it. By the time product listings became a paid Shopping product, the early movers had the data, the processes and the learning curve already behind them.

Connected Apps has the same shape. A new surface, a new integration standard, a phased preview and, for now, very few apps competing for any given task. Google says users can call an app by name, using "@" or simply mentioning it in a query, and that AI Mode can offer a list of available apps after a general request or default to the last app a user chose for that kind of task. Google has not said how that list is ordered. In a category with two connected apps, that question barely matters. In a category with twenty, it will matter a great deal.

I would expect the early window to reward the brands that turn up, as Google Base did. I would also be honest that nobody outside Google knows how long that window lasts, and even Google's own John Mueller has described it as extremely early.

What Google Has Actually Built

  • One integration, two surfaces. You build once and choose whether the app appears in Search, the Gemini app, or both.
  • MCP at the core. Apps connect through a cloud-hosted Model Context Protocol server, the open standard for exposing tools to AI models. Android apps can use AppFunctions to run tasks on the device instead.
  • Account linking. Users link their account with your service once through OAuth, then use it across Google's AI surfaces.
  • A console and a review. Registration, testing, review and publishing run through Google Connected Console, which also reports traffic, latency and error rates once you are live.
  • Rules with teeth. No third-party advertising inside your app, no deceptive tool descriptions written to steer the model, no near-duplicate apps. Breaches can mean rejection or removal.
  • US English in Search, for now. AI Mode connected apps are available in the US in English. Gemini availability is set app by app, and Google has not named its preview markets.

Who This Is For

This is the important filter. A connected app is not a page Google cites. It is a service Google calls on the user's behalf, usually against an account the user already holds with you. That makes it a strong fit for businesses where customers log in to get something done, and a weak fit for brochure sites with nothing behind the login.

The early partners make the pattern clear: grocery baskets, design templates, credit information, car rental, fitness. Practical candidates include:

  • Retailers with accounts and repeat purchase - reorder, build a basket, check an order, find what is in stock near me.
  • Travel, hospitality and rental - check availability, hold a booking, change a reservation.
  • Financial services - quotes, eligibility checks, balances and applications, within whatever your compliance team will sign off.
  • Subscriptions and SaaS - create, edit, schedule or report on something the user already does in your product.
  • Utilities, telco and services - account status, plan comparison, appointment booking.

If your business has no logged-in task worth delegating, connected apps are not your next move. Keep investing in being the source AI Mode cites, which is a separate visibility layer and still the one most brands should be working hardest on.

Six Practical Steps to Move Early

  1. Pick the job. List the five tasks your customers most often log in to do. The best first app does one of them very well, rather than everything badly.
  2. Check what sits behind it. If those tasks already run on an internal or public API with proper OAuth, most of the hard work is done and an MCP server is a wrapper around it. If they do not, that is the real project, and it is worth scoping now.
  3. Apply for the preview. Access is being opened in phases through developers.google.com/connected-apps. Being in the queue costs nothing. If you trade in the US, there is a live route into AI Mode today. If you only trade in the UK and EU, the build is identical, and the work you do now is the work you will need on launch day.
  4. Write your tool descriptions like landing page copy. The model decides when and how to use your app from your server instructions and tool descriptions. Clear, specific, honest descriptions are the new on-page optimisation. Inflated ones are explicitly against the policy.
  5. Build the brand people will name. Users invoke apps by name. Brand demand, existing customer accounts and a reason to say "@yourbrand" will matter more here than in classic rankings. This is where brand and performance budgets stop being separate conversations.
  6. Measure it from day one. When a task completes inside AI Mode, the session that produced it never reaches your site, so your web analytics will not see it. Tag every call that arrives through your MCP server, tie it back to revenue in your own systems, and treat Google's console figures as one view rather than the whole picture. Our analytics and attribution team can help you set that up properly.

The Bigger Picture

Put this next to Universal Cart and the Universal Commerce Protocol and the direction is hard to miss. Google wants the task, and increasingly the transaction, to happen inside its own surfaces. You can treat that as a threat to your traffic, and in part it is. But a connected app is the one version of that future where your brand is the thing doing the work, by name, inside Google's interface.

The businesses that moved early on Google Base did not do it because they knew how big it would become. They did it because the cost of turning up was low and the cost of arriving late was high. Connected Apps looks like the same bet.

If you want to work out whether your business has a task worth putting in front of AI Mode, and what it would take to build, talk to our search and AI optimisation team.

QueryClick: the last agency you will ever need.
Passionate About Performance

QueryClick combine best in class performance marketing strategy with unique, globally patented data insights & AI to help our client partners eliminate wasted media spend and grow.

Author
Chris Liversidge CEO of QueryClick & Corvidae
Chris Liversidge
Founder and CEO of QueryClick & Corvidae
Client Experience
Our Accreditations & Partnerships
Google Partner Premier 2026

Contact Us

Area of interest

Newsletter Signup