Microsoft Brand Agents: Should You Turn On Clarity’s AI Agent on Shopify?

Microsoft Brand Agents looks like a chatbot sitting on your store. It is first a distribution channel: turning it on publishes your catalog to Bing and Copilot.

I activated the feature in early September 2026 on a Shopify store I manage. What the terms say does not match what the interface advertises. Here is the detail.

Disclosure: I am a co-founder of Cockpyt AI. I state it every time the tool appears in my content.

What is Microsoft Brand Agents?

Brand Agents is a Microsoft service that installs a conversational agent on your online store to guide shoppers from product discovery to checkout. It runs inside Microsoft Clarity, Microsoft’s free behavioural analytics tool, and relies on the models that power Copilot.

Two things separate it from a standard chatbot. The agent ingests your product catalog, your prices, your availability and your shipping and return policies. It generates answers on the fly instead of drawing from a fixed answer bank. And it cross-references those answers with the behavioural signals Clarity already collects on your site.

The scope stays tightly framed as I write this:

• Status: beta, free, waitlist registration required.
• Platforms: Shopify and WooCommerce. Other CMS platforms are not eligible.
• Surfaces: websites only, no mobile applications.
• Project type: retail and ecommerce, where purchase is the primary conversion.
• Languages: the agent converses in every language and shoppers can switch mid-conversation.

On the store where I turned it on, the configuration interface is fully localised. The tab appears directly in the Shopify admin, under the Microsoft Clarity application.

What actually happens when you accept the terms?

You consent to a copy of your catalog being published to Microsoft Merchant Center. Microsoft states it plainly in its documentation: accepting the Brand Agents terms enables catalog publication to MMC, making your products discoverable across Bing, Copilot and other Bing-grounded AI assistants.

That clause changes the nature of the decision. You are not only choosing to add a chat bubble to your site. You are opening a product discovery channel into the Microsoft ecosystem, without configuring a Merchant Center feed manually, without media budget, and with updates reflected in real time across both surfaces.

The timing is not neutral. According to the AI Traffic Trends report published by Adobe Digital Insights in June 2026, AI-sourced traffic to US retail sites grew 138% year over year in Q2 2026, and those visitors convert 54% better than non-AI traffic. Adobe states it measures more than one trillion visits. These figures remain vendor-owned data, covering the US market, without independent audit: I read them as a directional trend, not as a guarantee transposable to a European store.

One point deserves attention: Microsoft states that your Brand Agents catalog remains the primary source for on-site agent interactions, and that the copy published to MMC extends visibility without taking the catalog away from you. The wording is reassuring. It says nothing about how Copilot arbitrates between your listing and a competitor’s on the same query.

On-site agent or discovery channel: two levers, two dashboards

Brand Agents does two different things that call for different metrics. Confusing them leads to misjudging the setup after three weeks.

The last row is the one I look at first on a client account. The chat switches off in one click. Distributing a catalog into a third-party ecosystem is a contractual decision, not an interface setting.

How do you turn on Brand Agents on a Shopify store?

Activation runs through six steps, from the Shopify App Store to publishing the agent. Budget around thirty minutes of configuration, plus catalog ingestion time, which ranges from minutes to hours depending on how many SKUs you carry.

1. Install the application Microsoft Clarity: AI Insights from the Shopify App Store. If Clarity already runs on your store, skip to the next step.
2. Enable the app embed block Clarity Agents JS in your Shopify theme settings. Without that block, waitlist registration stays blocked.
3. Join the waitlist from your Clarity dashboard and accept the Brand Agents terms. This is the step where you consent to catalog publication in Merchant Center.
4. Configure the welcome message, the bubble text and the widget position on mobile and desktop. The preview updates live in the right-hand panel.
5. Describe your offering and your tone. Those two fields drive the agent’s personality and the way it presents your products.
6. Test in preview, then publish. The agent becomes visible to your shoppers. You can edit the settings and republish as many times as needed.

Once the agent is live, a dedicated tab appears in Clarity with performance metrics. Microsoft’s documentation notes that data can take a few hours to appear after installation.

Five settings to fix before you publish your agent

The setting that matters most sits in the status screen, not the customisation screen. It is the traffic percentage exposed to the agent.

By default, my interface showed “All traffic”. Microsoft explicitly states in that same screen that a lower percentage lets you test with a smaller share of visitors. That option exists for a reason: a generative agent talking about your products to 100% of your visitors on day one, with no conversation reviewed, is a bet on the quality of a system still in beta.

The four other settings to handle before publishing:

• The bubble text. The default label points to AI-assisted shopping. A label aligned with your category vocabulary earns better open rates than a generic formula.
• Widget position. On mobile, a bottom-centre widget often overlaps the add-to-cart button in Shopify themes. Check on a real phone, not in the preview.
• The offering description. This field conditions what the agent asserts about your products. A vague description produces vague answers.
• The privacy policy. Microsoft requires a statement in your site policy, detailed below.

What Microsoft does not put forward

Four points appear in the documentation without appearing in the activation flow.

A mandatory statement in your privacy policy

The Brand Agents terms require merchants to insert a specific disclosure in the site privacy policy. It must state that you partner with Microsoft to deliver the service, that order management data may be captured through first-party and third-party cookies, and that conversation content and interaction events are shared with Microsoft to operate, secure and improve the service. Microsoft publishes a dedicated page with the exact sample wording. That obligation is flagged nowhere in the activation journey.

Answers are generated, therefore fallible

The agent does not recite a FAQ. It produces contextual answers from your catalog. On products with regulated characteristics, food supplements, cosmetics, medical devices, alcoholic drinks, financial products, the margin for error is not symmetrical: a loose reformulation can turn into a non-compliant claim. In those categories I treat conversation review as a duty of care, not a best practice.

An unsourced commercial claim

The interface announces that Microsoft’s partners have doubled their conversion rate. No methodology, no sample, no reference period accompanies that statement in the configuration screen. I treat it as a sales argument, not as data you can build a business case on.

A beta service

Beta means scope changes, no service level commitment and limited reporting filters. The documentation specifies that the dashboard currently supports only time frame, day of week, country and device filters.

Which metrics should you track once the agent is live?

The Brand Agents dashboard exposes four families of indicators. I rank them in this order on client accounts.

Adoption comes first: sessions with the agent available, conversation sessions, unique engaged users, and above all the engagement rate, meaning the share of eligible visitors who actually interact. An agent nobody opens produces no usable data, whatever conversion rate it displays.

The funnel comes next: bubble shown, chat used, add to cart, checkout, purchase. Microsoft compares that path between agent-assisted and organic sessions and calculates an uplift percentage. That comparison carries a classic bias: visitors who open a chat are already more engaged than average. The displayed uplift blends the agent’s effect with a selection effect.

Sales: total sales value from assisted sessions, average order value, average sales uplift. Same methodological caveat.

Engagement depth: average chat session duration, total complete turns, average turns per conversation. This is the indicator I read to judge the agent’s real quality, ahead of any monetary metric.

The blind spot sits elsewhere. This dashboard measures what happens on your site and inside the Microsoft ecosystem. It does not tell you whether ChatGPT, Perplexity or Gemini cite your brand when a shopper asks them for a product recommendation. In my own bot activity analysis, OpenAI and Perplexity crawlers each accounted for 2.38% of requests against 11.90% for Microsoft: the blind spot is measurable inside the tool itself.

Disclosure reminder: Cockpyt AI is the tool I co-founded. I mention it here because it covers exactly the blind spot described above, and you can compare it with market alternatives before choosing.

Should you turn on Brand Agents now?

Turn it on if three conditions hold: your catalog is clean and current, someone on your side can review conversations every week, and you accept catalog publication into the Microsoft ecosystem without reservation.

Wait if any one of those three is missing. A catalog with stale prices or wrong delivery times will not produce a mediocre agent, it will produce an agent that repeats your errors out loud, in natural language, with the apparent authority of an official assistant.

The edge case concerns regulated categories. There, the question is not “does it convert” but “who signs off on what the agent is allowed to assert”. Until that answer exists internally, activation stays premature, even at 10% of traffic.

Frequently asked questions about Microsoft Brand Agents

Is Brand Agents paid?

The service is free during the beta phase, like the rest of Microsoft Clarity. Microsoft has published no pricing for what comes next. I treat the free access as temporary and advise against building an economic model that depends on it.

Do you need a Microsoft Clarity account to use Brand Agents?

Yes. Brand Agents is a feature inside Clarity, not a standalone product. Installation runs through the Clarity application on Shopify or WooCommerce, then through activation from the Clarity dashboard.

Does Brand Agents work on PrestaShop, Magento or WordPress without WooCommerce?

No. The documentation limits the service to Shopify and WooCommerce. Microsoft mentions more platforms coming without a public timeline.

Can I activate the agent without publishing my catalog to Microsoft Merchant Center?

The documentation presents catalog publication as a consequence of accepting the terms, not as a separate option. If that distribution raises a contractual or competitive issue in your category, have the terms reviewed before activating.

Does the agent speak languages other than English?

Yes. Microsoft states the agent supports all languages and that shoppers can switch language mid-conversation. The admin interface is also localised.

How do you measure the effect of Brand Agents on your AI visibility?

The Clarity dashboard does not measure it. It tracks the agent on your site. To know whether your brand and products appear in answers from Copilot, ChatGPT, Perplexity or Gemini, you need a generative visibility tracker querying those engines directly.

What happens if I uninstall the application?

The agent disappears from your site. What becomes of the catalog copy published to Merchant Center depends on withdrawing your consent, a point the documentation handles less precisely than activation. Document the exit path before you enter.